The vertebrate brain is highly conserved topologically, but less is known about neuroanatomical variation between individual brain regions. Neuroanatomical variation at the regional level is hypothesized to provide functional expansion, building upon ancestral anatomy needed for basic functions. Classically, animal models used to study evolution have lacked tools for detailed anatomical analysis that are widely used in zebrafish and mice, presenting a barrier to studying brain evolution at fine scales. In this study, we sought to investigate the evolution of brain anatomy using a single species of fish consisting of divergent surface and cave morphs, that permits functional genetic testing of regional volume and shape across the entire brain. We generated a high-resolution brain atlas for the blind Mexican cavefish Astyanax mexicanus and coupled the atlas with automated computational tools to directly assess variability in brain region shape and volume across all populations. We measured the volume and shape of every grossly defined neuroanatomical region of the brain and assessed correlations between anatomical regions in surface fish, cavefish, and surface × cave F2 hybrids, whose phenotypes span the range of surface to cave. We find that dorsal regions of the brain are contracted, while ventral regions have expanded, with F2 hybrid data providing support for developmental constraint along the dorsal-ventral axis. Furthermore, these dorsal-ventral relationships in anatomical variation show similar patterns for both volume and shape, suggesting that the anatomical evolution captured by these two parameters could be driven by similar developmental mechanisms. Together, these data demonstrate that A. mexicanus is a powerful system for functionally determining basic principles of brain evolution and will permit testing how genes influence early patterning events to drive brain-wide anatomical evolution.
brain atlasAstyanaxcavefishneuroanatomyneurodevelopmentevo/devoResearch organismOtherhttp://dx.doi.org/10.13039/100000002National Institutes of HealthR15MH118625DuboueErik Rhttp://dx.doi.org/10.13039/100000002National Institutes of HealthR01GM127872KeeneAlex Chttp://dx.doi.org/10.13039/100000002National Institutes of HealthR35GM138345KowalkoJohanna Ehttp://dx.doi.org/10.13039/100000002National Institutes of HealthR15HD099022KowalkoJohanna Ehttp://dx.doi.org/10.13039/100000002National Institutes of HealthR21NS122166KowalkoJohanna EKeeneAlex Chttp://dx.doi.org/10.13039/100000001National Science Foundation1923372KowalkoJohanna EKeeneAlex CDuboueErik Rhttp://dx.doi.org/10.13039/100000001National Science Foundation2202359KowalkoJohanna Ehttp://dx.doi.org/10.13039/501100000854Human Frontier Science ProgramRGP0062KeeneAlex Chttp://dx.doi.org/10.13039/100000002National Institutes of HealthDE026446AlbertsonCraigThe funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.Author impact statementGenetic analyses reveal that neuroanatomical areas that are developmentally related co-evolve with one another.Introduction
Regional topology of the brain has remained remarkably conserved in vertebrate lineages across evolution (Holland and Holland, 2021; Charvet et al., 2011). While the arrangement for subregions of the brain remains constant, the overall size and shape of individual brain regions can vary considerably, even among closely related groups. (Hoops et al., 2017; Pereira-Pedro et al., 2020; Neubauer et al., 2020; Briscoe and Ragsdale, 2018). Comparative neuroanatomical studies suggest that novel anatomical changes are built upon the conservation of ancestral brain anatomy, with anatomical changes resulting in the development of new regions that likely expand function and functional repertoire (Woych et al., 2022; Wolff et al., 2017). Furthermore, the convergence of similar functional elaborations of the vertebrate brain, despite independent lineages of evolution, suggest some regions of the brain are primed for anatomical and ultimately functional diversification (Schumacher and Carlson, 2022; Emery and Clayton, 2004; Northcutt, 2002). Although these comparative studies have provided a wealth of insight into the evolutionary history of neuroanatomy, pioneers in the field strongly advise diversifying our animal model pool to permit functional tests to theories underlying the evolution of the vertebrate brain (Northcutt, 2002; Striedter, 1998).
Two central hypotheses are thought to drive anatomical brain evolution; the first suggesting that subregions brain-wide tend to evolve together, with selection operating on mechanisms that govern the growth of all regions, and the second positing that selection can act on individual brain regions, and that regions which are functionally related will anatomically evolve together independent of other brain regions (Barton and Harvey, 2000; Herculano-Houzel et al., 2014; Montgomery et al., 2016). While data supporting each hypothesis exist, the large divergence times and poor understanding of evolutionary history in most comparative models makes generalizing these theories difficult. Additionally, these theories have largely been applied to studies analyzing anatomical size (Barton and Harvey, 2000; Wartel et al., 2019; Axelrod et al., 2018), overlooking how 3D shape of brain regions are impacted by evolutionary processes. Moreover, the relationship between the evolution of the size and shape of distinct anatomical regions is poorly understood, and it is unclear how these two important aspects of neuroanatomy explain the evolution of the brain.
While volume and shape are known to govern anatomical variation across the brain, there is still uncertainty of whether similar or distinct mechanisms govern both parameters (Reardon et al., 2018). However, most comparative studies tend to focus on either volume or shape, with some volume to shape analyses comparing trends across independent studies (Gómez-Robles et al., 2014; Sansalone et al., 2020; Montgomery et al., 2021). Current models to explore mechanisms driving volume and shape rely on non-model systems that lack experimental approaches, or model organisms that lack genetic diversity, creating an impediment for investigating basic principles of brain evolution. (Ponce de Leon et al., 2021; Mitchell, 1977). Non-traditional models can bridge this experimental gap, including experimental approaches to determine whether similar genetic and developmental mechanisms underlie general principles of evolution.
The blind Mexican cavefish Astyanax mexicanus provides a powerful model for directly testing how genetic variation impacts brain-wide anatomical evolution (Mitchell, 1977; Jeffery, 2008). A. mexicanus exists as a species with two distinct forms: river dwelling surface fish and cave dwelling populations that have independently evolved troglobitic phenotypes (Bradic et al., 2012; Gross, 2012). This separation has led to high genetic diversity between populations which underlies the stark differences in phenotypes between surface and cave populations (Borowsky, 2021; Warren et al., 2021). Importantly, surface × cave hybrid offspring are biologically viable, allowing us to exploit the genetic differences between each population, and ultimately identify the genetic underpinnings of neuroanatomical evolution in the cavefish brain (O’Gorman et al., 2021; Duboué et al., 2011). Therefore, a hybrid population analysis using novel neurocomputational tools can be used to study covariation of neuroanatomy across a well-annotated atlas, and directly test whether cavefish brains exhibit support for either the developmental or functional constraint hypothesis. Finally, the relationship between brain region shape and volume can be analyzed in comparative and direct analyses, which will be critical in understanding how brain regions evolve in relationship to one another.
In the current study, we generated a brain-wide neuroanatomical atlas for A. mexicanus and applied new computational tools for assessing brain-wide changes in both brain region volume (Gupta et al., 2018) and shape (Conith et al., 2019). We then applied this atlas to hybrid brains to make associations between naturally occurring genetic variation of wildtype populations and neuroanatomical phenotypes. Our surface × cave hybrid data reveal that brain-region volume and shape are genetically specified to regulate brain-wide anatomical evolution in A. mexicanus. Furthermore, volume and shape exhibit similarities in brain-wide anatomical covariation, suggesting that these two parameters share developmental mechanisms that are causing cavefish brains to contract dorsally and expand ventrally. These results suggest that selection may be operating on simple developmental mechanisms, that likely impact early patterning events to modulate the volume and shape of brain regions.
ResultsGeneration of a single brain-wide atlas for all <italic>Astyanax</italic> morphs
To analyze regional variation in brain anatomy, we created a single atlas for all A. mexicanus morphs to provide neuroanatomical comparisons across surface, cave, and surface × cave hybrid populations (Figure 1a–c). A neuroanatomical analysis pipeline from zebrafish that performs automated segmentation of brains was then adapted and tested on A. mexicanus brains (Gupta et al., 2018; Figure 1b and c and Figure 1—figure supplement 1a–c). This tool provides a single atlas that can be continually segmented through brain regions to identify neuroanatomical differences across various molecularly and functionally defined sub-nuclei (Figure 1c, Figure 1—figure supplement 1d, e). The segmentation accuracy was confirmed by a pairwise cross correlation of tERK staining and manual to automated segmentation overlap (Figure 1—figure supplement 2a, b; >98% tERK cross-correlation and >78% segmentation cross-correlation). Previous brain atlases for model organisms were constructed using molecular markers that are known to demarcate specific subregions, such as transgenic lines and antibody labeling (Gupta et al., 2018; Randlett et al., 2015; Kunst et al., 2019). To determine whether our atlas maintains similar molecular accuracy, we developed a dual staining technique that combines RNA hybridization chain reaction (HCR) in situ hybridization with immunohistochemistry (IHC), resulting in automated segmentation of RNA in situ probes via total-ERK antibody registration. With this approach, we were able to confirm the accuracy of larger segments identified through automated segmentation, including subregions of the hypothalamus and optic tectum, with accurate segment bounding of insulin gene enhancer protein (ISL-1) (Figure 1d, Randlett et al., 2015; Kunst et al., 2019; Sanek and Grinblat, 2008; Langenberg and Brand, 2005) and orthodenticle homeobox 2 (otx2) RNA labeling, respectively (Figure 1—figure supplement 3a, French et al., 2007; Paridaen et al., 2009; Diotel et al., 2015). We then tested the accuracy of smaller regions, such as the dorsal subpallium, medial preoptic region, and thalamus, via gastrulation brain homeobox 1 (gbx1), oxytocin (oxt) and nitrous oxide 1 (nos1) RNA labeling, respectively (Figure 1—figure supplement 3b–d, Randlett et al., 2015; Kunst et al., 2019; Blechman et al., 2007; Gutierrez-Triana et al., 2014). Finally, we confirmed the accuracy of the smallest subregions of the brain that can be defined molecularly, such as the locus coeruleus and dorsal raphe, using tyrosine hydroxylase (TH) and 5-hydroxytryptamine (5-HT) antibody labeling, respectively (Figure 1—figure supplement 3e,f, Gupta et al., 2018; Randlett et al., 2015; Sittaramane et al., 2009; Kidwell et al., 2018; Oikonomou et al., 2019; Ulhaq and Kishida, 2018).
Developing a single <italic>A</italic>. <italic>mexicanus</italic> atlas to perform direct brain-wide morphometric analyses across all populations.
(a) Map showing the 29 independently evolved cave populations (black dots) of the El Abra region in Mexico. The Pachón cavefish population used for this project is marked as a red dot. Scale bars = 0.5 cm (full fish, 1 year adult) and 0.5 mm (larvae). (b) Schematic showing registration and atlas inverse registration method used to create an A. mexicanus atlas for cross-population segmentation and analysis. (c) Sagittal and transverse (i–iii) sections of the 26 region surface fish and cavefish atlas. (i) Habenula (pink), pallium (blue), ventral thalamus (purple), and preoptic (light green). (ii) optic tectum neuropil (sky blue), optic tectum cell bodies (green), tegmentum (light purple), rostral hypothalamus (dark blue), posterior tuberculum (gold), statoacoustic ganglion (beige). (iii) Cerebellum (dark purple), prepontine (light green), locus coeruleus (brown), raphe (beige), intermediate hypothalamus (dark brown), and caudal hypothalamus (bright red). (d) Islet1/2 antibody segmentation following ANTs inverse registration of cavefish atlas. Islet positive neurons exhibit the same segmentation in the preoptic, rostral, and caudal portions of the hypothalamus that have been reported islet positive in zebrafish. Scale bars (b–d) = 80 µm.
Surface × Pachón F<sub>2</sub> hybrid brain atlas in nifty format.
Whole-brain segmented z-stack that outlines 180 defined brain regions. Z-projections of atlas stack were used to generate images.
Pipeline for immunohistochemistry, automated segmentation, and volumetric comparisons of individual fish larvae.
(a) Larval fish were imaged at 5 dpf to measure external anatomical features (e.g. standard length). (b) Larvae were fixed at 6 dpf, immunostained with total-ERK (tERK) and imaged on a two-photon microscope (2P). (c) All larvae were registered to a surface × cave F2 hybrid reference brain, followed by an inverse registration of the segmented cavefish atlas. (d) The 180-region ZBB brain atlas was transformed into scalable atlases, 4 major subdivisions, 13 developmentally defined and 26 molecularly and functionally defined regions. (e) Segmented larval brains were run through CobraZ to volumetrically measure and statistically compare each brain region. Scale bar = (a) 500 µm, (b and c) 100 µm, (d) 50 µm, and (e) 25 µm.
Cross-correlation analysis between hand and automated segmentation of total-ERK-defined brain segments.
(a) Optical sections (coronal) through the and optic tectum (white matter light blue and gray matter red) and pallium (dark blue). Hand segmentations are in color, while automated versions are shown in white outline. (b) Cross-correlation analysis comparing the percent of overlap between hand and automated segments for ERK-defined regions. Pixel correlation percentages did not vary between surface and cavefish. Abbreviations, S=surface fish and P=Pachón cavefish. Cross-correlation percentages were compared using a standard t-test. Scale bars = 25 µm. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
RNA probe and antibody analysis of segmentation accuracy across the <italic>Astyanax</italic> atlas.
Segmentation of (a) otx2 positive neurons in the optic tectum, (b) gbx1 neurons in the subpallium, (c) nos1 neurons in the thalamus, (d) oxt positive neurons in the medial preoptic regions, (e) anti-TH neurons in the locus coeruleus, and (f) 5-hydroxytryptamine (5-HT) positive neurons in the superior raphe. Hand segmentations are in color, while automated versions are shown in white outline. Inset panels show outline of segment with positive expression of marker (top) and overlay with full 180-region atlas (bottom). Scale bars = 50 µm for full field, 20 µm for inset images.
Determining neuroanatomical variation brain-wide for surface and cave populations
To analyze and compare the relative volume of brain regions in surface fish and cavefish populations, volumetric data was measured and analyzed for variation between surface and cave brains (Figure 2, Figure 1—figure supplement 1e). The atlas was applied to individuals from the Pachón cavefish, Molino cavefish, Rio Choy surface fish, Pachón to Rio Choy F1 and F2 hybrid, and Molino to Rio Choy F2 hybrid populations, via immunolabel-based brain registration and inverse registration, allowing us to address the evolutionary mechanisms underlying variation in brain anatomy. Importantly, Pachón and Molino cavefish are independently evolved populations (Herman et al., 2018), that allow us to determine whether the process of evolution impacts neuroanatomy convergently in cave environments, despite differences in the standing genetic variation of the two populations.
Volumetric variation in wildtype populations reveal convergent dorsal contraction and ventral expansion in the brain of two cavefish populations.
(a) Volumetric comparison of the diencephalon in surface fish, Pachón and Molino cavefish. Percent total brain volume represents pixels of segment divided by total pixels in the brain. Sagittal sections show diencephalon (purple). (b) Volumetric comparisons of the dorsal diencephalon (green) and hypothalamus (orange). (c) Volumetric comparisons of the habenula (gold), ventral thalamus (teal) and dorsal thalamus (burnt orange) of the dorsal diencephalon; along with the preoptic (cyan), intermediate zone (purple), and caudal zone (red), of the hypothalamus. Sample size = surface (16) and Pachón cavefish (24). (d) Colorimetric model depicting size differences in brain regions between surface fish and cavefish. A larger volume in surface fish results in blue coloration, while a larger volume in cavefish results in a red coloration. Horizontal optical sections depicting (i) dorsal, (ii) medial, and (iii) ventral views of the brain. p-Value significance is coded as: *=p < 0.05, **=p < 0.01, ***=p < 0.001, ****=p < 0.0001. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010). Scale bars = 80 µm (a), 25 µm (b).
Volumetric values for the diencephalon of wildtype larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Volumetric values for the dorsal diencephalon of wilid-type larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Volumetric values for the habenula of wildtype larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Volumetric values for the ventral thalamus of wildtype larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Volumetric values for the dorsal thalamus of wildtype larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Volumetric values for the hypothalamus of wildtype larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Volumetric values for the preoptic region of wildtype larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Volumetric values for the intermediate zone of wildtype individual larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Volumetric values for the caudal hypothalamus of wildtype individual larvae.
Three columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish. Populations were analyzed via one-way ANOVA.
Variation in segment volume between surface and cavefish populations.
(a) Major brain divisions of the vertebrate brain and box plots comparing the volume of each region between surface and cave. Volumes are reported as percentage of total brain volume, calculated by dividing total pixels in a segment by total pixels within the brain. (b) Columns segmenting each major brain division into brain regions defined by developmental, molecular, and functional categories. Each box plots brain segment is color-coded to the corresponding atlas picture at the top of each column. All segments were statistically analyzed using a Student’s t-test and Holm’s corrected for multiple comparisons. p-Value significance is coded as: *=p < 0.05, **=p < 0.01, ***=p < 0.001, ****=p < 0.0001. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010). Scale bars = 40 µm (Tele, Dien, Mesen), 80 µm (Rhomb).
To survey volumetric variation across populations, we progressively segmented subdivisions within our brain atlas from larger gross anatomical segments to molecularly defined subregions. This progressive segmentation provided an analytical tool for defining regional variability through sub-nuclei, with localization of variability increasing as we scaled through each level of the atlas (Figure 2a–c, Figure 2—figure supplement 1). Our initial analysis of numerous brain regions revealed results that support previously published studies for the four major brain regions and larger subdivisions of those brain regions (Figure 2—figure supplement 1a, b; Loomis et al., 2019; Jaggard et al., 2020). While this comparison found volumetric differences that were previously reported in broad developmental regions (Jaggard et al., 2020; Menuet et al., 2007), such as the hypothalamus (Figure 2b; F=9.252, surface to Pachón p=0.0154, surface to Molino p=0.003), our atlas was able to determine that the intermediate and caudal hypothalamus were enlarged in cavefish populations (Figure 2c, Figure 2—figure supplement 1b; intermediate – F=19.11, surface to Pachón p<0.0001, surface to Molino p<0.0001; caudal – F=10.98, surface to Pachón p<0.0001, surface to Molino p<0.0001). In addition, we also discovered novel volumetric differences, including contraction of the dorsal diencephalon in cavefish (Figure 2b; F=39.89, surface to Pachón p=0.0025, surface to Molino p<0.0001), that we localized to the dorsal thalamus in Pachón cavefish and across the thalamus and habenula in Molino cavefish (Figure 2c, dorsal thalamus – F=16.64, surface to Pachón p<0.0001, surface to Molino p<0.0001; Habenula – F=16.64, surface to Molino p<0.0001). Overall, we were able to use this single atlas to pinpoint discrete differences between brain regions of surface fish and cavefish, while also creating a brain-wide model for Pachón and Molino cavefish that highlights a convergent dorsal-ventral remodeling of the brain in both populations (Figure 2d).
Analysis of hybrid animals defines neuroanatomical associations brain-wide
Lab-generated hybridization between surface and cave populations provides a powerful system for determining how high genetic diversity of natural populations contributes to phenotypic diversity. To define anatomical relationships volumetrically between wildtype populations and larval offspring, we quantified relative volume for each brain region of surface, cave, and surface × cave F1 and F2 hybrids (Figure 3, Figure 3—figure supplement 1). Hybrid brain regions show variability that appears consistent with different modes of inheritance, including surface dominant, cavefish dominant, and surface × cavefish intermediate anatomical forms (Figure 3—figure supplement 1a–c). We then investigated regional variability in surface × cave F2 hybrid brain regions by further segmenting down to local sub-nuclei (Figure 3a–d, Figure 3—figure supplement 2a–c). These analyses revealed molecularly defined regions that account for segment variability, such as the ventral sub-nuclei of the optic tectum stratum periventricular (Figure 3d; ventral optic tectum stratum periventricular – F=34.61, surface to surface × Pachón F2 p<0.0001, Pachón to surface × Pachón F2 p<0.0001, surface to surface × Molino F2 p<0.0001, and Molino to surface × Molino F2 p=0.0005) and pallium (Figure 3—figure supplement 2c, ventral pallium – F=43.56, surface to surface × Pachón F2 p<0.0438, Pachón to surface × Pachón F2 p<0.0001, surface to surface × Molino F2 p<0.0001, and Molino to surface × Molino F2 p=0.0002). These results reveal that brain-wide anatomical variation is likely genetically heritable in cavefish and that this genetic relationship can be resolved at the sub-nuclei level across the brain.
Scalable segmentation of the tectum identifies high variability in the ventral sub-nuclei of the optic tectum’s cell layers.
(a) Volumetric comparison of the mesencephalon in surface fish, Pachón cavefish, Molino cavefish, surface × Pachón F2 hybrid (SPF2), and surface × Molino F2 hybrid (SMF2) larvae. Sagittal sections showing the mesencephalon (green). Percent total brain volume represents pixels of segment divided by total pixels in the brain. Segment tree abbreviations, M – mesencephalon, TeO – optic tectum, Tg – tegmentum, R – rostral, D – dorsal, V – ventral. (b) Volumetric comparisons of the optic tectum (yellow) and tegmentum (purple). (c) Volumetric comparisons of the optic tectum white (neuropil; forest green) and gray matter (cell bodies; orange). (d) Volumetric comparisons of rostral (royal blue), dorsal (purple), and ventral (lime green) segments of the optic tectum gray matter. All segments were statistically analyzed using a standard ANOVA and Holm’s corrected for multiple comparisons. p-Value significance is coded as: *=p < 0.05, **=p < 0.01, ***=p < 0.001, ****=p < 0.0001. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010). Scale bars = 80 µm (a), 25 µm (b), 50 µm (c and d).
Volumetric values for the mesencephalon of wildtype and F<sub>2</sub> hybrid larvae.
Five columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish, and F2 surface × cave populations. Populations were analyzed via one-way ANOVA.
Volumetric values for the tegmentum of wildtype and F<sub>2</sub> hybrid larvae.
Five columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish, and F2 surface × cave populations. Populations were analyzed via one-way ANOVA.
Volumetric values for the optic tectum of wildtype and F<sub>2</sub> hybrid larvae.
Five columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish, and F2 surface × cave populations. Populations were analyzed via one-way ANOVA.
Volumetric values for the optic tectum gray matter of wildtype and F<sub>2</sub> hybrid larvae.
Five columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish, and F2 surface × cave populations. Populations were analyzed via one-way ANOVA.
Volumetric values for the optic tectum white matter of wildtype and F<sub>2</sub> hybrid larvae.
Five columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish, and F2 surface × cave populations. Populations were analyzed via one-way ANOVA.
Volumetric values for the rostral optic tectum of wildtype and F<sub>2</sub> hybrid larvae.
Five columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish, and F2 surface × cave populations. Populations were analyzed via one-way ANOVA.
Volumetric values for the dorsal optic tectum of wildtype and F<sub>2</sub> hybrid larvae.
Five columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish, and F2 surface × cave populations. Populations were analyzed via one-way ANOVA.
Volumetric values for the ventral optic tectum of wildtype and F<sub>2</sub> hybrid larvae.
Five columns of normalized volumetric values for wildtype surface, Pachón and Molino cavefish, and F2 surface × cave populations. Populations were analyzed via one-way ANOVA.
Volumetric variability in hybrid larvae reflects wildtype genetic diversity through dominant and intermediate phenotypes.
Representative regions for hybrid larvae that show (a) an intermediate brain size between wildtype surface and cave values, (b) genetic dominance with Pachón cavefish (dorsal diencephalon) or surface fish (subpallium) wildtype populations, and (c) no difference between hybrid and wildtype populations. All segments were statistically analyzed using a standard ANOVA and Holm’s corrected for multiple comparisons. p-Value significance is coded as: *=p < 0.05, **=p < 0.01, ***=p < 0.001, ****=p < 0.0001. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
Hybrid brains link genetic variation in wildtype populations to anatomical variation in distinct sub-nuclei of the olfactory bulb, subpallium, and pallium.
(a) Volumetric comparison of the telencephalon in surface fish, Pachón cavefish, Molino cavefish, and surface × cave F2 hybrids. Percent total brain volume represents pixels of segment divided by total pixels in the brain. (b) Sagittal sections show subdivisions of the telencephalon, pallium (green), subpallium (light blue), and olfactory bulb (red). (c) Volumetric comparisons of discrete regions of the pallium, dorsal (sky blue) and ventral (purple), and olfactory bulb, rostral (gold) and caudal (green). All segments were statistically analyzed using a standard ANOVA and Holm’s corrected for multiple comparisons. p-Value significance is coded as: *=p < 0.05, **=p < 0.01, ***=p < 0.001, ****=p < 0.0001. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010). Scale bar = 80 µm (a), 50 µm (c).
Covariation of F<sub>2</sub> hybrid brain regions reveals brain-wide anatomical tradeoff impacting distinct developmental clusters
To determine which brain regions covary and whether anatomical variation provides support for either the developmental or functional constraint hypothesis, we looked for pairwise anatomical associations for regional volume across all subregions of the brain in surface to Pachón and surface to Molino F2 hybrids. These results were then run through a hierarchical cluster analysis to gain insight into brain-wide evolutionary mechanisms driving anatomical change in cavefish brains. For this dataset, clusters constitute regions that share volumetric associations, wherein F2 brain regions are either getting larger or smaller together (Figure 4a–d). The clustering analysis for the 180-brain region scale revealed six large clusters for surface to Pachón F2 hybrids (Figure 4b) and twelve clusters for surface to Molino F2 hybrids (Figure 4d), with each cluster showing strong positive volumetric associations among subregions in that cluster. We also found strong negative correlations between cluster groups (Figure 4b and d), suggesting that these regions have the potential to co-evolve by similar genetic mechanisms, with one group getting larger as the other gets smaller. Surprisingly, these clusters map onto the brain in well-defined dorsal to ventral positions, with positive associations being found across dorsal clusters and ventral clusters, and negative associations between dorsally and ventrally positioned clusters (Figure 4e, f). This first analysis suggests that small subregions of the brain are clustering as larger modules and exhibiting brain-wide volumetric associations that suggest an anatomical tradeoff along the dorsal-ventral axis, where some areas become reduced in size at the expense of other areas’ increasing volumes.
Volumetric covariation and clustering of hybrid brain regions reveals convergent associations across the dorsal-ventral axis.
(a) Cross-correlation analysis of surface to Pachón F2 hybrids for the 180 segmented Astyanax brain atlas. (b) Cluster analysis array showing six clusters exhibiting positive volumetric associations. Positive relationships are color-coded light red, negative dark red (n=37). (c) Cross-correlation analysis of surface to Molino F2 hybrids for the 180 segmented Astyanax brain atlas. (b) Cluster analysis array showing 12 clusters exhibiting positive volumetric associations a. Positive associations are color-coded light red, negative dark red (n=37). Clusters color-coded and mapped onto the surface × cave F2 hybrid reference brain for both (e) surface to Pachón F2 hybrids and (f) surface to Molino F2 hybrids. The rainbow gradient represents depth along the z-plane, blue shifted (dorsal) to red shifted (ventral). Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
Correlation coefficient matrix for 180-brain region atlas volumetric comparisons of surface × Pachón F<sub>2</sub> hybrid larvae.
Pairwise correlation coefficients for pairwise comparisons between each brain region within the F2 population. Each individual coefficient represents the relationship between two individual brain regions within the F2 population.
Correlation coefficient matrix for 180-brain region atlas volumetric covariation of surface × Molino F<sub>2</sub> hybrid larvae.
Pairwise correlation coefficients for pairwise comparisons between each brain region within the F2 population. Each individual coefficient represents the relationship between two individual brain regions within the F2 population.
Covariation of brain region size reveals developmental tradeoff between dorsal-ventral clusters brain-wide in independently derived cavefish populations.
(a) Pairwise correlation and (b) clustering matrices comparing covariation for surface to Pachón F2 hybrids between 13 brain segments. (c) Illustrations depicting clustered segments: cluster 1 includes the subpallium (blue), hypothalamus (orange), posterior tuberculum (olive green), tegmentum (purple), and prepontine (light blue) cluster 2 includes the optic tectum (green), cerebellum (yellow), pons (light purple), reticulopontine (orchid purple), and medulla oblongata (light orange). (d) Pairwise correlation and (e) cluster matrices for surface to Molino F2 hybrids. (f) Illustration depicting surface to Molino hybrid clustered segments for clusters 1 and 3 correspond to regions found in surface to Pachón hybrid clusters 1 and 2, respectively. Note the statoacoustic ganglion (Sg) is not depicted in the illustrations of the midline sagittal section. Sample size, surface to Pachón F2 hybrids n=37 and surface to Molino F2 hybrids n=40. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
To help map these brain-wide volumetric associations to larger developmental regions, we reduced our segmentation to 13 ontologically defined regions (e.g. hypothalamus, cerebellum, etc.), we then performed pairwise correlation and cluster analyses on our 13 brain region scale atlas for the two populations of surface × cave F2 hybrids (Figure 4—figure supplement 1a–d). This developmental cluster analysis revealed three clusters for both populations (Figure 4—figure supplement 1c, d), with positive associations between neuroanatomical areas within a cluster. We then mapped neuroanatomical regions with the clusters back on the brain and found that loci within each cluster were physically localized together (Figure 4—figure supplement 1e). The large clusters for both populations encompassed the same broad anatomical regions, with one cluster comprised of the dorsal and caudal areas of the brain (e.g. optic tectum and cerebellum), while the second cluster was predominantly made up of the ventral brain (e.g. hypothalamus and subpallium; Figure 4—figure supplement 1e, f). When compared to surface to Pachón hybrids, only the statoacoustic ganglion clustered differently in surface to Molino hybrids (Figure 4—figure supplement 1c, d). Therefore, we found that the two large portions of the brain exhibit the same dorsal ventral volumetric associations as the smaller clusters found in our 180-brain region analysis. To further analyze the data statistically, we added up the correlation values of the clusters and ran a pairwise comparison across clusters (clusters 1 and 2, t=18.48, p<0.0001; clusters 1 and 3, t=13.82, p<0.0001; clusters 2 and 3, t=5.802, p=0.0011) that revealed statistical significance across all clusters displaying negative volume associations. Taken together, these analyses suggest that a general feature of neuroanatomical evolution in cave-derived populations of A. mexicanus may be a developmental tradeoff between ventral expansion and dorsal contraction, as evinced by parallel findings in two independently evolved populations.
Geometric morphometrics provide an analytical tool for understanding the relationship between shape and volume during brain-wide evolution
Previous studies examining variation in the brain have mostly focused on volume (Hoops et al., 2017; Eliason et al., 2021) or shape (Pereira-Pedro et al., 2020; Watanabe et al., 2021), with few providing a comparison of how shape and volume vary brain-wide (Axelrod et al., 2018; Reardon et al., 2018; Sansalone et al., 2020). We sought to examine whether shape variation follows similar patterns as volume, which could suggest shared genetic or developmental origins underlying variation, or whether shape and volume were unrelated. To determine morphological variation in shape across the brain, we employed shape analysis (i.e. geometric morphometrics) approaches previously used in assessing shape variation among whole brain and brain regions (Conith et al., 2020; Conith and Albertson, 2021). We first examined whether shape showed variation between populations for regions with no variation in volume, then how volume and shape relate within specific regions and finally whether shape variations follow the same brain-wide patterns seen in volumetric variation.
To begin evaluating how subregion shape varies between surface fish and cavefish brains, we chose to characterize the pineal and preoptic region because they show no volumetric variation across populations, yet play functional roles in behaviors that are highly variable across Astyanax populations. We characterized pineal and preoptic shape among the three populations using landmark-based geometric morphometrics. We performed a principal component analysis (PCA) to reduce our landmark data to a series of orthogonal axes that best represent shape variation within our brain regions. We found significant differences in shape of the pineal and preoptic region among wildtype surface fish, cavefish, and surface × cave F2 hybrids (Figure 5—figure supplements 1 and 2). Importantly, surface × cave hybrids have a range of phenotypes that likely exhibit additive (preoptic, Figure 5—figure supplement 1a, F=5.076, Pr(>F)=0.001) and genetically dominant (pineal, Figure 5—figure supplement 2a, F=4.159, Pr(>F)=0.0001) modes of inheritance, suggesting that the differences in shape may be driven by genetics. Taken together, we find that despite a lack of volumetric differences among populations, population-level variation in regional brain shape can occur, and the specific variation we observed likely impacts adaptive behaviors discovered in previous studies.
Shape and volume relationships within regions can exhibit similarities or variation on a region-to-region basis
To determine whether features of shape and volume of regions covary in relation to brain-wide anatomical evolution in cavefish, we chose to analyze two regions from our volumetric cluster 1, the optic tectum and cerebellum, and two regions from cluster 2, the hypothalamus and tegmentum (Figure 5a, b). These regions allowed us to test whether shape exhibits covariation patterns similar to volume, including positive relationships within and negative relationships across dorsal and ventral clusters. First, we analyzed shape variation across F2 individuals for each brain region. To that end, we again performed a PCA to characterize shape variation and identify what aspects of shape were driving differences within the F2 hybrid population for each of the four brain regions (Figure 5a, Figure 5—figure supplements 3–6).
Shape covariation suggests volume and shape share brain-wide mechanism of brain evolution.
Representatives of shape for the cerebellum and hypothalamus of pure populations (a) principal component 1 (PC1) and (b) principal component 2 (PC2). (c) Correlation matrix comparing the covariation of shape between regions from volumetric covarying cluster 1, cerebellum (Ce) and optic tectum (TeO), and cluster 2, hypothalamus (Hyp) and tegmentum (Tg). Sample size, n=37. (d) A cluster analysis of covariation grouped regions into two clusters as predicted by volumetric covariation. Scale bars = 100 µm. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
Correlation coefficient matrix for 13 brain region atlas shape covariation of surface × Pachón F<sub>2</sub> hybrid larvae.
Each individual coefficient represents the relationship between two individual brain regions within the F2 population.
Shape variability of the preoptic region in hybrid larvae display an intermediate phenotype between wildtype populations.
(a) Principal component analysis (PCA) capturing 54% of the variation across surface, cave, and surface × cave F2 hybrids. PC1 describes preoptic width, while PC2 describes length. (b) Illustrations of the median shape for each population. Top row provides an anterior view, bottom row provides a side view. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
Pineal shape variation in hybrids exhibits a cavefish dominant phenotype.
(a) Principal component analysis (PCA) capturing 50% of the variation across surface, cave, and surface × cave F2 hybrids. PC1 describes pineal length, while PC2 describes pineal width across populations. (b) Illustrations of the median shape for each population. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
Optic tectum shape variation is characterized by width curvature and thickness.
(a) Principal component analysis (PCA) capturing 52% of the variation across surface × cave F2 hybrid optic tectums. PC1 describes optic tectum curvature and length (b), while PC2 describes optic tectum width (c) across surface × cave F2 hybrid. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
Cerebellar shape variation is characterized by depth of curvature and thickness.
(a) Principal component analysis (PCA) capturing 39% of the variation across surface × cave F2 hybrid cerebellums. PC1 describes cerebellar curvature and length (b), while PC2 describes cerebellar width (c) across surface × cave F2 hybrid. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
Hypothalamic shape variation is characterized by length, central thickness, and posterior width.
(a) Principal component analysis (PCA) capturing 39% of the variation across surface × cave F2 hybrid hypothalamus. PC1 describes hypothalamic length and center depth (b), while PC2 describes hypothalamic posterior width (c) across surface × cave F2 hybrids. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
Tegmentum shape variation is characterized by depth and length.
(a) Principal component analysis (PCA) capturing 30% of the variation across surface × cave F2 hybrid tegmentums. PC1 describes tegmental length and center depth (b), while PC2 describes tegmental posterior width (c) across surface × cave F2 hybrids. Statistical tables can be found in the Dryad repository associated with this study (Portella et al., 2010).
We then assessed the degree of association among these four brain regions using partial least squares (PLS), a method which permitted the complete landmark configuration of each individual to be used in assessing the degree of covariation. We found that our shape × shape associations broadly match patterns observed in the volumetric data, as we observed significant covariation in shape between the optic tectum and cerebellum (Z=2.48, p=0.004) alongside covariation between the hypothalamus and tegmentum (Z=4.026, p=0.0001). However, we also observed covariation in shape between the hypothalamus and cerebellum (Z=3.146, p=0.0001), indicating that the regulation of shape and volume is likely under distinct development control.
Given that certain regions of the brain could exhibit differences in shape among populations (i.e. preoptic, pineal) independent of volume, we directly assessed the degree of association between shape and volume for each brain region. We found significant associations between volume and shape in three of four brain regions. There were strong associations between volume and shape in the cerebellum, tectum, and tegmentum, while we found no association in the hypothalamus (tegmentum, Z=2.686, Pr(>F)=0.0041; optic tectum, Z=2.06, Pr(>F)=0.0178; cerebellum, Z=2.752, Pr(>F)=0.0021). These results show that the relationship between volume and shape varies among individual brain regions, suggesting that shape and volume can exhibit either shared or distinct mechanisms.
Shape and volume variation follow the same covariation pattern brain-wide suggesting shared developmental mechanisms of brain evolution
We sought to compare regional shape variation across the brain to better understand the similarities and differences between how shape and volume evolve in the brain. To determine whether shape and volume were modulated by distinct or similar mechanisms, we extracted the first principal component from our optic tectum, cerebellum, hypothalamus, and tegmentum (Figure 5a, b). By extracting a single variable we could assess associations among the shapes of brain regions using the same cluster-based methodological approaches that were applied to the volumetric data. As a result, we could determine whether variation in shape and volume are modified by distinct or varying developmental mechanisms. We found that covariation of anatomical shape clusters the same as volume in a dorsal-ventral fashion, with cluster 1 and cluster 2 showing positive relationships within clusters, and negative relationships across clusters (Figure 5c, d). This initial shape analysis suggests that mechanisms in support of the developmental constraint hypothesis may be impacting both anatomical volume and shape to reorganize the dorsal-ventral development of cavefish brains. However, future efforts looking at all brain regions will be needed for a stronger conclusion.
Discussion
Here, we establish a laboratory model of anatomical brain evolution that utilizes an innovative, molecularly defined neuroanatomical atlas and applied computational tools, which can be used to assess mechanisms underlying anatomical brain evolution. The application of this atlas and these computational approaches to F2 hybrid fish permits a brain-wide dissection of how neuroanatomy changes, and a powerful analysis of not only how different neuroanatomical areas evolve but also which areas co-segregate together. These studies suggest that brains of cave-adapted populations of A. mexicanus are anatomically evolving via the developmental constraint hypothesis (Herculano-Houzel et al., 2014; Montgomery et al., 2016), with a brain-wide dorsal-ventral relationship, that suggests expansion of ventral regions is directly related to contraction of dorsal regions. Finally, this study is one of the first to directly assess how the volume and shape of brain regions relate to one another across genetically and phenotypically diverse populations of a single species.
Previous studies examining how the brain evolves have largely been restricted to comparative analyses between closely related, albeit different species, and these studies have revealed gross differences in neuroanatomy, connectivity, and function between derived animals (Reardon et al., 2018; Gómez-Robles et al., 2014; Sansalone et al., 2020; Montgomery et al., 2021). However, the A. mexicanus model system provides a powerful tool for assessing how the brain evolves in a single species with multiple divergent forms and an extant ancestor (Jeffery, 2008; Gross, 2012; Jeffery, 2001). Moreover, because surface and cave forms are the same species, the ability to produce surface/cave and cave/cave F2 hybrid fish permits a powerful dissection of functional principles underlying brain evolution (O’Gorman et al., 2021; Duboué et al., 2011). We previously published population-specific neuroanatomical atlases for this species and used these to examine how gross neuroanatomy differs between surface and cave fish, and how physiology relates to behavior (Loomis et al., 2019; Jaggard et al., 2020). The current study extends applications of this model to further understand how the brain evolves, and includes a single atlas for all populations to functionally compare neuroanatomy in pure and hybrid offspring, automated brain segmentation for 180 annotated sub-populations of neurons, and the application of computational approaches for a complete whole-brain assessment of the evolution of the brain. In future studies, we will be able to utilize the genetic diversity of these populations to map anatomical traits and then functionally test how natural genetic variation in parental populations (e.g. surface, Pachón, etc.) impact the anatomical evolution of the cavefish brain.
Two competing hypotheses exist to explain brain evolution: one theory suggests that the majority of the brain evolves anatomically via changes to shared developmental programs, whereas others have suggested that more discrete regions will independently evolve based on shared function (Herculano-Houzel et al., 2014; Montgomery et al., 2016). Our data from two independent cavefish populations provides evidence that supports the notion that the developmentally related regions of the brain co-evolve, with the dorsal-caudal areas of the brain evolving together, and that regions such as the optic tectum and the cerebellum, two areas that constitute a large proportion of the dorsal-caudal region shrink in size. In contrast, rostral-ventral areas co-evolve together, such as the hypothalamus and subpallium, that are enlarged in cavefish. Importantly, we find in F2 hybrid fish that reduced optic tectum and cerebellum are concomitant with an enlarged hypothalamus and subpallium, suggesting that expansion of some regions come at the expense of others. This anatomical outcome may suggest that early brain patterning genes and developmental mechanisms could be influencing the establishment of the dorsal-ventral axis of the brain (O’Gorman et al., 2021; Duboué et al., 2011; Gupta et al., 2018), leading to an asymmetrical shift in overall brain mass. Moreover, finding this in two separate populations suggests that these changes are common principles for fish evolving in a cave environment.
Early brain patterning is tightly regulated by genetic networks and developmental mechanisms that include organizing centers conserved across bilaterians (Sylvester et al., 2011; Stoykova et al., 2000; Blaess et al., 2008; Denes et al., 2007; Holland et al., 2013). These networks and mechanisms are controlled by both morphogen pathways, such as sonic hedgehog (shh) and bone morphogenic protein (bmp), and transcription factors (Molina et al., 2007; Sasagawa et al., 2002; Wilson and Maden, 2005), like the homeobox gene cluster (hox) (Hunt et al., 1991; Krumlauf et al., 1993; Spitz et al., 2001; Hatta et al., 1991), that orchestrate axis development and help regulate regional specification. Early forward genetic screens in model organisms led to the discovery of axial patterning genes, with mutants displaying drastic phenotypes impacting dorsal-ventral and anterior-posterior axis patterning. For instance, mutations in the transforming growth factor-beta and sonic hedgehog signaling pathways, in zebrafish, fly, and mice revealed severe phenotypic impacts in ventral forebrain development (Chiang et al., 1996; Maity et al., 2005; Schier et al., 1996). Our analyses suggest that two independently derived cavefish populations exhibit changes in early brain development that impact the majority of brain regions in a strictly dorsal-ventral fashion. While the divisions of the dorsal ventral axis are initially established via canonical pathways, such as shh and bmp, disruptions in downstream targets tend to be localized to specific ventral and dorsal regions (Shimamura and Rubenstein, 1997; Ko et al., 2013; Karaca et al., 2015; Portella et al., 2010; Diaz and Puelles, 2020). Therefore, our hybrid experiments may be the result of changes upstream of larger gene regulatory networks, impacting several genes that contribute en masse to the development of the dorsoventral axis (Levine and Davidson, 2005; Alexandro et al., 2021). This hypothesis would explain the concomitant dorsal contraction and ventral expansion revealed in the correlation and clustering analyses of surface to cave hybrid larva. Further genetic and embryonic analyses will be needed to answer many outstanding questions, including: do these anatomical changes reflect a developmental ‘hotspot’, and are these anatomical features a common outcome for evolving non-visual sensory dominance?
Additionally, degeneration of the eye and subsequent impact on the entire brain has not been extensively studied in cavefish. While eye formation begins in cavefish populations, the embryonic eye primordia quickly undergoes apoptosis (Jeffery and Martasian, 1998; Yamamoto and Jeffery, 2000), followed by axon degeneration and denervation of retinal ganglion cells in the midbrain (Soares et al., 2004). This loss of afferent ocular contacts contributes to a decrease in the overall growth and size of the optic tectum (Soares et al., 2004; Pottin et al., 2011). Furthermore, rescuing eye development via lens transplantation from a surface donor to a cave host results in increased tectal mass on the contralateral side of the transplanted lens (Soares et al., 2004). Recent work has also hypothesized that changes in spatiotemporal gene expression of anterior neural markers impact both the eye and brain in a pleotropic manner (Menuet et al., 2007; Pottin et al., 2011; Alié et al., 2018). Therefore, a remaining question is whether eye degeneration and changes in dorsal-ventral patterning are separate mechanisms impacting the brain, or share a common mechanism that results in the overall ventral expansion and dorsal contraction observed in these cavefish populations. While this study did not resolve these questions, we are currently utilizing our novel computational atlas to determine the overall impact of eye degeneration on brain-wide anatomy.
Past comparative neuroanatomical research has raised many unanswered questions for how neural circuits evolve, including how variation in brain development between closely related groups relate to functional convergence of specialized brain regions (Schumacher and Carlson, 2022; Emery and Clayton, 2004; Northcutt, 2002; Güntürkün, 2012; Earl, 2022). For instance, it was initially thought that the pallial regions in lobe finned fishes (including tetrapod’s) and ray finned fishes evolved independently to produce convergent functional traits, evinced by variation in developmental processes, cell types, and circuitry involved in these emergent behaviors. However, recent work utilizing a well-preserved fossil of an ancient ray finned species suggests that ray finned fish initially possessed the same developmental processes as lobbed finned fish (Figueroa et al., 2023). Therefore, functional similarity of the pallium may be a homologous ancestral state that was maintained by the two groups and not a convergent feature derived independently. Although this recent finding supports a major revision of our understanding of vertebrate brain evolution, we agree with the field that variation in cell type diversity and complexity of forebrain circuit development across these derived extant groups present unique and non-overlapping neurological traits. Neuroanatomical discoveries like this provide a prime example that creative strategies, including examination of the fossil record for soft-tissue preservation and functional studies in a diversity of non-model organisms, will be necessary to reveal unique and generalized principles of neural evolution.
Recent neural evolutionary studies in several non-model species have shown that some brain regions provide evolutionary potential for convergent function (Schumacher and Carlson, 2022; Carlson, 2016; Earl, 2022). For example, several independently derived groups of weakly electric fish have convergently evolved electrogenerative and electroreceptive potentials through expansions of the cerebellum (Schumacher and Carlson, 2022). This rather specific structural and functional innovation suggests that the cerebellum provides an anatomical substrate, with specific gene regulatory networks and cell types (Güntürkün, 2012), that are best suited for fish to gain electroreceptive properties (Schumacher and Carlson, 2022). While convergent functional innovations are observed in these independent lineages, the secondary consequences on behavior can vary from one species to another, suggesting that the process of evolution is acting upon functional expansion (electric properties), providing a substrate for novel behaviors (Schumacher and Carlson, 2022; Carlson, 2016). In our study, two independently derived cavefish populations show convergent neuroanatomical variation across the dorsal-ventral axis, resulting in an overall expansion of specific sensorimotor regions, including the hypothalamus and subpallium. Both the hypothalamus and the subpallium have diverse functions, and many behavioral modifications in cavefish, including aggression, stress, and sleep, have been related to functional variation in these areas (Rodriguez-Morales et al., 2022; Chin et al., 2018; Jaggard et al., 2018). We hypothesize that the reduction of dorsal regions preserves the energy needed to expand this ventral substrate of non-visual sensorimotor regions as anatomical potential to engender novel behaviors (Moran et al., 2014). That these changes are found in independently evolved populations further supports this notion.
Other taxonomic groups have also experienced increases in anterior forebrain volume, leading to the formation of new cell types and layers (Woych et al., 2022; Lust et al., 2022; Briscoe et al., 2018). The functional expansion of the forebrain and flexibility of supramodal cognition in primate brains has been linked to convergent adaptations to specific subsets of the cortex (Sneve et al., 2019; Hill et al., 2010; Chaplin et al., 2013). These changes included an expansion to the size and organizational complexity of the cortex, while also developing novel forebrain circuits that permit a functional compacity to produce more complex cognitive and social behaviors. Our data points to a similar phenomenon, wherein subpallial and hypothalamic regions are expanding in these two cavefish populations that likely shifts the primary integrative processes in the optic tectum to the ventral forebrain. It will be paramount going forward to determine whether ventral anatomical expansion is leading to new cell types, and how these anatomical changes impact ancestral neural circuits in relation to cavefish behavior.
In addition to volume, evolutionary changes in shape of neuroanatomical regions have been shown to alter function of different regions. The mammalian cortex, for example, has evolved from a smoother lissencephalic cortex in more ancestral species to a folded one in more derived animals such as primates (Herculano-Houzel et al., 2014; Elias and Schwartz, 1969; Molnár et al., 2014). Folding of the cortex is thought to increase surface area and has been implicated in more complex processing of the brain (Molnár et al., 2014; Tallinen et al., 2014; Hofman and Falk, 2012; DeCasien et al., 2017; Abzhanov et al., 2006). However, we do know that shape variation has been shown to be a common adaptation in other tissues. Beak differences in Galapagos finches have been shown to change in accordance with the size of food sources, and such changes have been shown to rely on differences in bone morphogenic protein signaling (Parsons and Albertson, 2009; Kozol et al., 2021; Choi et al., 2016). Craniofacial differences in African cichlids also have been shown to vary as an adaptive quality to food availability (Gupta et al., 2018; Conith et al., 2019). Furthermore, standard methods for assessing complex shape features have been applied to studying brain shape evolution in non-model organisms, generating anatomical evolutionary hypotheses that have lacked an appropriate model for assessing functional mechanisms of anatomical evolution (Reardon et al., 2018; Gómez-Robles et al., 2014; Molnár et al., 2014). By applying these morphological measuring and analyzing methods with our hybrid volume pipeline, we were able to see that complex shape phenotypes are likely genetically encoded, evidenced in hybrid intermediate phenotypes, and that similarities in covariation of shape and volume across dorsal and ventral regions may be impacted by shared mechanisms. However, due to the labor-intensive nature of shape analyses, we acknowledge that only 4 of 13 brain regions from the volumetric covariation analysis were assessed and are working to compare the remaining regions in our ongoing studies. Additionally, some of our shape variation could be capturing biological elements that are captured in the volumetric analysis, which we cannot rule out in the current analysis. While the functional and adaptive significance of differences in shape are not known, future work relating neuronal activity and function with differences in shape in this model could help address this question.
Together, these results support the developmental constraint hypothesis of brain evolution in cave-adapted A. mexicanus fish populations, suggesting early genetic and developmental impacts reshaping neuroanatomy brain-wide. This study represents the first computational brain atlas for a single species with multiple evolutionary derived forms, and the application of the atlas to hybrid animals represents the first assessment of how different neuroanatomical areas evolved in both volume and shape. Moreover, we can now combine this atlas with a myriad of cutting-edge tools that we have generated for this model, including functional neuroimaging and genome editing, that will allow researchers to identify the genetic mechanisms that explain these changes. The strong genetic and neuronal conservation of the vertebrate brain, as well as the simplified nervous system of fish, suggests that this model offers great potential to discover the general principles of evolution that impact the brain.
Materials and methodsFish maintenance and husbandry
Mexican tetras (A. mexicanus) were housed in the Florida Atlantic Universities Mexican tetra core facilities. Larval fish were maintained at 23°C in system-water and exposed to a 14:10 hr light:dark cycle. Mexican tetras were cared for in accordance with NIH guidelines and all experiments were approved by the Florida Atlantic University Institutional Care and Use Committee protocol #A1929. A. mexicanus surface fish lines used for this study; Pachón cavefish stocks were initially derived from Richard Borowsky (NYU); surface fish stocks were acquired from Rio Choy stocks. Surface Rio Choy were outcrossed to Pachón to generate F1 hybrids, while F1 hybrid offspring were incrossed to produce F2 hybrids.
IHC and imaging
Larval IHC was performed as previously published (Avants et al., 2011), using antibodies raised against total ERK (ERK; p44/42 MAPK [ERK1/2], #4696, Cell Signaling Inc, Danvers, MA, USA), Islet-1 and Islet-2 homeobox (Islet1/2, #39.4D5, Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, USA), TH1 (AB152, Sigma-Aldrich Inc, Burlington, MA, USA), and 5-HT (AB125, Sigma-Aldrich Inc). IHC-stained larvae were imaged on a Nikon A1R multiphoton microscope, using a water immersion 25×, NA 1.1 objective.
Combined IHC and HCR in situ hybridization
To combine IHC and HCR in situ hybridization, the HCR in situ hybridization methodology for zebrafish embryos and larvae from Molecular Instruments (Kozol et al., 2022) was performed with the following exceptions: during the detection stage, larvae were incubated in probe solution for 48 hr to improve hybridization of RNA probes, and larvae were washed with 5× SSCTx (0.2% Triton X-100) instead of 5× SSCTw (0.2% Tween20) following hairpin incubation. Following HCR in situ hybridization, larvae were incubated in 5× SSCTx (0.2% Triton X-100) with 2% bovine serum albumin (BSA) at room temperature for 2 hr on a rocker (low speed). Following incubation, a primary antibody solution was added that included 5× SSCTx, 1% DMSO, 1% BSA, and 1:250 dilution of total-ERK antibody. Larvae were then incubated in primary antibody solution at 4°C on an orbital shaker set to 90 RPM for 48 hr. Primary antibody solution was then washed out three times with 5× SSCTw (0.2% Tween-20) for 10 min at room temperature on a rocker (low speed). Following primary incubation, a secondary antibody solution was added that included 5× SSCTw 1% DMSO, 1% BSA, and 1:500 dilution of goat anti-mouse IgGγ1 secondary antibody, Alexa Fluor 555 (Thermo Fisher, Waltham, MA, USA). Larvae were then incubated in secondary antibody solution at 4°C on an orbital shaker set to 16 hr and 90 RPM. Finally, secondary antibody solution was washed out three times with 5× SSCTw for 10 min at room temperature on a rocker (low speed) and subsequently imaged on a Nikon A1R confocal microscope, using a water immersion 20×, NA 0.95, long working distance objective, with 1.2× zoom.
Generation of the brain-wide <italic>A. mexicanus</italic> atlas
To generate a segmented atlas for cave and surface Astyanax, we used a previously published neuroanatomical atlas (Gupta et al., 2018) from a related fish, the common zebrafish (Danio rerio), that is neuroanatomically homologous with A. mexicanus (Jaggard et al., 2020). We first modified the zebrafish brain browser brain atlas, neuropil, and cell body mask for the existing zebrafish resource CobraZ by using previously published Advanced Normalization Toolbox (ANTs; Gupta et al., 2018; Wile, 2005) registration and inverse registration scripts. This process creates a set of computational instructions for aligning our hybrid standard brain to the zebrafish reference brain, these instructions are then reversed to map the zbb segmented atlas onto our hybrid standard brain (Figure 1b). This created a hybrid brain atlas that could be used to register brains from all four A. mexicanus populations, producing a single computational atlas for measuring brain size and shape (Figure 1c, Figure 1—figure supplement 1c–e). We validated our Astyanax segmented atlas using three distinct approaches, cross-correlation of tERK saturated pixels, automated to hand segmentation overlap, and molecular markers that label distinct neuroanatomical regions. The cross-correlation analysis between registered Astyanax brain and the zbb reference brain revealed that the two were highly correlated (rho = 0.95). Next, we hand-segmented five brains each for surface fish, Pachón cavefish, and F2 surface × cave hybrid larvae. These labeled neuroanatomical areas were then compared to automated segmentation from our brain atlas by running a custom cross-correlation script. 3D volumetric images were imported into MATLAB using the ‘imread’ function, vectorized to a 1D vector using ‘imreshape’, and then a Pearson’s correlation was performed using the ‘corr’ function (scripts are deposited in the Dryad depository for this study; Schlager, 2017). The automated segmentation to hand segmentation analysis revealed no difference in segmentation accuracy across Astyanax populations and >80% correlation between ERK-defined hand-segmented and automated segmentation (Figure 1—figure supplement 2). Finally, antibodies and RNA probes were used to test the accuracy of segment bounding for subregions that are known to outline specific molecularly defined neuronal populations (Figure 1d, Figure 1—figure supplement 3, Table 1).
Hybridization chain reaction (HCR) in situ hybridization probes.
Gene
A. mexicanuspopulation
Ensembl ID
Molecular Instruments Lot #
gbx1
Surface
ENSAMXT00000037099.1
PRO705
gbx1
Pachón
ENSAMXT00005023309.1
PRO706
otx2b
Surface
ENSAMXT00000055482.1
PRQ451
otx2b
Pachón
ENSAMXT00005060650.1
PRQ452
oxt
Surface
ENSAMXT00000041101.1
PRQ449
oxt
Pachón
ENSAMXT00005006990.1
PRQ450
Automated segmentation and brain region measurements
Surface, Pachón, and surface to Pachón hybrid larval tERK-stained brains were registered and segmented using the aforementioned ANTs scripts. The resulting brain mask and segmentation file for each larvae was then processed using the morphometric analysis suite CobraZ. CobraZ measures the size of segmented regions of the brain and calculates regional size as percent of total brain (pixels of brain region/total pixels in brain; Gupta et al., 2018). We did not amend ANT’s registration or inverse registration scripts, nor did we change the CobraZ parameters used in Gupta et al., 2018. We did additionally produce a modified segmentation file that defines larger subregions that overlap with tERK neuropil to provide cross-correlation analysis across brain regions and populations. Finally, we tested the accuracy of subregion segmentation using the HCR in situ hybridization probes and antibodies previously mentioned in the above sections. All scripts used in the analysis and generation of statistics (Supplementary Statistical Tables.xlsx) and materials in figures are archived in the Dryad submission associated with this study (see Data sharing; Schlager, 2018).
3D geometric morphometric methods to characterize shape variation
Correlations between volumes of brain regions were determined using custom-written scripts in Python. Volume data was imported from Microsoft Excel into Python using the pandas library. SciPy was then used to determine the pairwise correlation between all brain regions. The Seaborn library was then used to generate a heat map with annotations set to ‘True’ to overlay correlation coefficients on the pairwise correlation matrix. Cluster analysis of the corresponding pairwise correlation matrix was performed using SciPy toolkit. The distance matrix was first calculated from the correlation matrix and then indexed into the corresponding clusters. The correlation matrix was then clustered by grouping all regions that clustered (i.e. had the same index value). The resulting metric was again generated using Seaborn. The code for these analyses can found in the Dryad repository (see README, Kozol et al., 2021).
3D geometric morphometric methods to characterize shape variation
We used 3D geometric morphometrics to characterize shape variation in six different brain regions: preoptic, pineal, cerebellum, hypothalamus, tectum, and tegmentum. For the preoptic and pineal regions we landmarked parental and F2 hybrid populations (preoptic: Pachón [n=23], surface [n=23], F2 [n=34]; pineal: Pachón [n=15], surface [n=23], and F2 [n=36]). We landmarked only F2 hybrids for the remaining brain regions (cerebellum, n=28; hypothalamus, n=34; tectum, n=30; tegmentum, n=34). We used a combination of landmark types to best assess shape in each brain region (i.e. fixed, semi, surface), and placed landmarks onto the extracted 3D meshes using the morphometrics program LandmarkEditor (v3.0) (Baken et al., 2021). Landmark placement was manually conducted, except for the pineal, in which we utilized a semi-automated method (see below). To characterize shape in the remaining brain regions we used 16 landmarks for the preoptic (fixed LM n=16), 102 landmarks for the cerebellum (fixed LM n=2; surface semi-landmarks n=99), 34 landmarks for the hypothalamus (fixed LM n=34), 202 landmarks for the tectum (fixed LM n=2, surface semi-landmarks n=200), and 26 landmarks for the tegmentum (fixed LM n=26).
For the pineal body, we placed two fixed landmarks at the anterior and dorsal apexes of the pineal and surrounded the base of the pineal with 26 sliding semi-landmarks. We then took advantage of a procedure to automate the placement of 99 surface landmarks across the pineal region to wrap the pineal body with sliding surface semi-landmarks to best characterize the shape of this subregion among individuals. This required building a computer-aided design (CAD) template of the pineal using FreeCAD (v.0.16.6712), which we modeled as a hemisphere, and placing the fixed landmarks, sliding semi-landmarks (Figure 5—figure supplements 1 and 2), and surface landmarks on the CAD model using LandmarkEditor. We then used the R package Morpho to map the surface landmarks from the template to the pineal model of each individual specimen using the placePatch function (Adams, 2021; Conith et al., 2023).
Following landmark placement, we performed a Procrustes superimposition on our shape data for each brain region to remove the effects of translation, rotation, and scaling from all individuals using the gpagen function from the geomorph (v4.0) package in R (Schlager, 2017; Schlager, 2018). Following superimposition, we performed a PCA to reduce our landmark data to a series of axis that best reflect differences in brain shape variation with each region. We plotted the component scores from each PCA to visualize how the shape of each brain region varies among parental populations and/or within the F2 hybrids (Figure 5—figure supplements 3–6). We also extracted PC1 – the PC that explains the greatest amount of shape variation for a given brain region – from the cerebellum, hypothalamus, tectum, and tegmentum for use in a subsequent cluster analysis.
Allometry
We explicitly wanted to retain the allometric component of shape variation given that one of our major goals was to understand how a variable related to size – volume – varies among brain regions and populations. Similarly, developmental modularity may be a function of allometric scaling relationships (Conith et al., 2023) so by retaining allometry in our shape data, the results from our volumetric and shape cluster analysis should be more comparable. Despite this, we tested for allometry in our shape data by performing a multivariate regression of shape on centroid size using the procD.lm r function from geomorph and found three of our F2 surface × cave hybrid brain regions exhibited an association between shape and size (cerebellum, hypothalamus, tegmentum), while three did not (pineal, preoptic, tectum), further highlighting the complex nature of size and shape relationships within the brain (Figure 5—figure supplement 6).
Partial least squares and cluster analysis using shape data
To assess the degree of association between brain subregion shape and volume, we performed a multivariate regression of shape on volume using the procD.lm r function from geomorph. Similarly, to assess associations among brain subregion shapes, we performed a PLS analysis using the two.b.pls r function from geomorph. PC1 extracted data from the 3D geometric morphometric methods to characterize shape variation section were then run through the pairwise correlation and cluster SciPy functions described for the F2 hybrid volumetric analyses.
Statistics
All wildtype population standard t-tests were calculated using the program CobraZ (Bradic et al., 2012). For hybrid population comparisons, Prism (GraphPad Software Inc, San Diego, CA, USA) was used to run standard ANOVAs, followed by a Holm-Šídák’s multiple comparisons test to correct for comparing across statistical permutations for each figures analysis. All statistical tables for main figures and figure supplements are available in the ‘Supplementary Statistical Tables’. To evaluate covariation of F2 subregions, geometric morphometry analyses were all conducted in R (Choi et al., 2016; Avants et al., 2011) using the packages geomorph (v4.0) and Morpho (v2.6) (Kozol et al., 2022; Wile, 2005; Schlager, 2017; Schlager, 2018) to assess associations and produce morphospace plots. Sample sizes for this study were based off previous studies (Bradic et al., 2012; Sittaramane et al., 2009), and therefore power analyses were not conducted.
Data sharing
All data, statistical tables, custom code, and adapted tools have been made available on a Dryad repository, https://doi.org/10.5061/dryad.w9ghx3frw/ (Choi et al., 2016).
Additional informationCompeting interests
No competing interests declared
Author contributions
Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing - original draft, Project administration
Data curation, Formal analysis
Formal analysis
Data curation
Data curation
Data curation
Data curation
Data curation
Conceptualization, Funding acquisition, Writing - original draft, Project administration
Conceptualization, Funding acquisition, Writing - original draft, Project administration
Conceptualization, Writing - review and editing
Conceptualization, Investigation, Writing - original draft, Project administration
Ethics
Mexican tetras were cared for in accordance with NIH guidelines and all experiments were approved by the Florida Atlantic University Institutional Care and Use Committee protocol #A1929.
Additional files
README.md file that includes text describing the material in <xref ref-type="supplementary-material" rid="scode2">Source code 2</xref>, a repository of scripts used in this study.
Kozol et al. Code.zip file containing bash (.sh), fiji (.ijm), and matlab (.m) files used for this study.
Data availability
All raw and analyzed data, custom code and adapted tools have been uploaded into a Dryad repository, https://doi.org/10.5061/dryad.w9ghx3frw. Custom code and adaptive tools are also included in the supplemental material.
The following dataset was generated:
KozolRACree-NewmanATolentinoBPazAYuiskaABaneshKConithALloydEKowalkoJKeeneAAlbertsonCDuboueER2023Data from: A brain-wide analysis maps structural evolution to distinct anatomical modulesDryad Digital Repository10.5061/dryad.w9ghx3frw
Acknowledgements
We would like to thank Dr. Harry Burgess for his help in adapting the zebrafish brain browser atlas and modifying files for CobraZ to analyze A. mexicanus neuroanatomy. We thank Dr. James Jaggard for his expertise and early atlas work as a foundation for this project. To the administration and staff at the Jupiter Life Science Initiative in the Department of Biology at Florida Atlantic University, especially Peter Lewis and Arthur Loppatto for overseeing the health and care of the FAU Astyanax fish facility. This research was supported by grants from the NIH to ERD R15MH118625-0, ACK HFSP-RGP0062/1R01GM127872/, JEK R35GM138345/R15HD099022, ACK and JEK R21NS122166. This work was also supported by a grant from the NSF to ERD, JEK, and ACK #1923372 and an NSF grant to JEK 2202359.
ReferencesAbzhanovAKuoWPHartmannCGrantBRGrantPRTabinCJ2006The calmodulin pathway and evolution of elongated beak morphology in Darwin’s finchesNature44256356710.1038/nature0484316885984AdamsDCC2021Geomorph: software for Geomorphic morphometric analyses4.0.5R Package Versionhttps://cran.r-project.org/web/packages/geomorph/index.htmlAlexandroETMüllerFAndersenJSundaramLKathiriaAShcherbinaAFarhKChangHYPașcaAMKundajeAPașcaSPGreenleafWJ2021Chromatin and gene-regulatory dynamics of the developing human cerebral cortex at single-cell resolutionCell1845053506910.1016/j.cell.2021.07.03934390642AliéADevosLTorres-PazJPrunierLBouletFBlinMElipotYRetauxS2018Developmental evolution of the forebrain in cavefish, from natural variations in neuropeptides to behavioreLife7e3280810.7554/eLife.3280829405116AvantsBBTustisonNJSongGCookPAKleinAGeeJC2011A reproducible evaluation of ANTs similarity metric performance in brain image registrationNeuroImage542033204410.1016/j.neuroimage.2010.09.02520851191AxelrodCJLabergeFRobinsonBW2018Intraspecific brain size variation between coexisting sunfish ecotypesProceedings of the Royal Society B2852018197110.1098/rspb.2018.1971BakenEKCollyerMLKaliontzopoulouAAdamsDC2021geomorph v4.0 and gmShiny: Enhanced analytics and a new graphical interface for a comprehensive morphometric experienceMethods in Ecology and Evolution122355236310.1111/2041-210X.13723BartonRAHarveyPH2000Mosaic evolution of brain structure in mammalsNature4051055105810.1038/3501658010890446BlaessSStephenDJoynerAL2008Gli3 coordinates three-dimensional patterning and growth of the tectum and cerebellum by integrating Shh and Fgf8 signalingDevelopment1352093210310.1242/dev.01599018480159BlechmanJBorodovskyNEisenbergMNabel-RosenHGrimmJLevkowitzG2007Specification of hypothalamic neurons by dual regulation of the homeodomain protein OrthopediaDevelopment1344417442610.1242/dev.01126218003738BorowskyR2021Evolution: The genetics of trait evolution in cavefishCurrent Biology31R1014R101610.1016/j.cub.2021.06.01634428412BradicMBeerliPGarcía-de LeónFJEsquivel-BobadillaSBorowskyRL2012Gene flow and population structure in the Mexican blind cavefish complex (Astyanax mexicanus)BMC Evolutionary Biology12910.1186/1471-2148-12-922269119BriscoeSDAlbertinCBRowellJJRagsdaleCW2018Neocortical association cell types in the forebrain of birds and alligatorsCurrent Biology2868669610.1016/j.cub.2018.01.03629456143BriscoeSDRagsdaleCW2018Homology, neocortex, and the evolution of developmental mechanismsScience36219019310.1126/science.aau371130309947CarlsonBA2016Differences in electrosensory anatomy and social behavior in an area of sympatry between two species of mormyrid electric fishesThe Journal of Experimental Biology219314310.1242/jeb.12772026567347ChaplinTAYuHHSoaresJGMGattassRRosaMGP2013A conserved pattern of differential expansion of cortical areas in simian primatesThe Journal of Neuroscience33151201512510.1523/JNEUROSCI.2909-13.201324048842CharvetCJStriedterGFFinlayBL2011Evo-devo and brain scaling: candidate developmental mechanisms for variation and constancy in vertebrate brain evolutionBrain, Behavior and Evolution7824825710.1159/00032985121860220ChiangCLitingtungYLeeEYoungKECordenJLWestphalHBeachyPA1996Cyclopia and defective axial patterning in mice lacking Sonic hedgehog gene functionNature38340741310.1038/383407a08837770ChinJSRGassantCEAmaralPMLloydEStahlBAJaggardJBKeeneACDuboueER2018Convergence on reduced stress behavior in the Mexican blind cavefishDevelopmental Biology44131932710.1016/j.ydbio.2018.05.009ChoiHMTCalvertCRHusainNHussDBarsiJCDevermanBEHunterRCKatoMLeeSMAbelinACTRosenthalAZAkbariOSLiYHayBASternbergPWPattersonPHDavidsonEHMazmanianSKProberDAvan de RijnMLeadbetterJRNewmanDKReadheadCBronnerMEWoldBLansfordRSauka-SpenglerTFraserSEPierceNA2016Mapping a multiplexed zoo of mRNA expressionDevelopment1433632363710.1242/dev.14013727702788ConithAJLamDTAlbertsonRC2019Muscle-induced loading as an important source of variation in craniofacial skeletal shapeGenesis57e2326310.1002/dvg.2326330418689ConithAJKiddMRKocherTDAlbertsonRC2020Ecomorphological divergence and habitat lability in the context of robust patterns of modularity in the cichlid feeding apparatusBMC Evolutionary Biology209510.1186/s12862-020-01648-x32736512ConithA.JAlbertsonRC2021The cichlid oral and pharyngeal jaws are evolutionarily and genetically coupledNature Communications12547710.1038/s41467-021-25755-534531386ConithAJHopeSAAlbertsonRC2023Covariation of brain and skull shapes as a model to understand the role of crosstalk in development and evolutionEvolution & Development258510210.1111/ede.1242136377237DeCasienARWilliamsSAHighamJP2017Primate brain size is predicted by diet but not socialityNature Ecology & Evolution111210.1038/s41559-017-011228812699DenesASJékelyGSteinmetzPRHRaibleFSnymanHPrud’hommeBFerrierDEKBalavoineGArendtD2007Molecular architecture of annelid nerve cord supports common origin of nervous system centralization in bilateriaCell12927728810.1016/j.cell.2007.02.04017448990DiazCPuellesL2020Developmental genes and malformations in the hypothalamusFrontiers in Neuroanatomy1460711110.3389/fnana.2020.60711133324176DiotelNRodriguez VialesRArmantOMärzMFergMRastegarSSträhleU2015Comprehensive expression map of transcription regulators in the adult zebrafish telencephalon reveals distinct neurogenic nichesThe Journal of Comparative Neurology5231202122110.1002/cne.2373325556858DubouéERKeeneACBorowskyRL2011Evolutionary convergence on sleep loss in cavefish populationsCurrent Biology2167167610.1016/j.cub.2011.03.02021474315EarlB2022Humans, fish, spiders and bees inherited working memory and attention from their last common ancestorFrontiers in Psychology1393771210.3389/fpsyg.2022.93771236814887EliasHSchwartzD1969Surface areas of the cerebral cortex of mammals determined by stereological methodsScience16611111310.1126/science.166.3901.1114897390EliasonCMMcCulloughJMAndersenMJHackettSJ2021Accelerated brain shape evolution is associated with rapid diversification in an avian radiationThe American Naturalist19757659110.1086/71366433908824EmeryNJClaytonNS2004The mentality of crows: convergent evolution of intelligence in corvids and apesScience3061903190710.1126/science.109841015591194FigueroaRTGoodvinDKolmannMACoatesMICaronAMFriedmanMGilesS2023Exceptional fossil preservation and evolution of the ray-finned fish brainNature61448649110.1038/s41586-022-05666-136725931FrenchCREricksonTCallanderDBerryKMKossRHageyDWStoutJWuennenberg-StapletonKNgaiJMoensCBWaskiewiczAJ2007Pbx homeodomain proteins pattern both the zebrafish retina and tectumBMC Developmental Biology78510.1186/1471-213X-7-8517634100Gómez-RoblesAHopkinsWDSherwoodCC2014Modular structure facilitates mosaic evolution of the brain in chimpanzees and humansNature Communications5446910.1038/ncomms546925047085GrossJB2012The complex origin of Astyanax cavefishBMC Evolutionary Biology1210510.1186/1471-2148-12-10522747496GüntürkünO2012The convergent evolution of neural substrates for cognitionPsychological Research7621221910.1007/s00426-011-0377-921881941GuptaTMarquartGDHorstickEJTaborKMPajevicSBurgessHA2018Morphometric analysis and neuroanatomical mapping of the zebrafish brainMethods150496210.1016/j.ymeth.2018.06.00829936090Gutierrez-TrianaJAHergetULichtnerPCastillo-RamírezLARyuS2014A vertebrate-conserved cis-regulatory module for targeted expression in the main hypothalamic regulatory region for the stress responseBMC Developmental Biology144110.1186/s12861-014-0041-x25427861HattaKKimmelCBHoRKWalkerC1991The cyclops mutation blocks specification of the floor plate of the zebrafish central nervous systemNature35033934110.1038/350339a02008211Herculano-HouzelSMangerPRKaasJH2014Brain scaling in mammalian evolution as a consequence of concerted and mosaic changes in numbers of neurons and average neuronal cell sizeFrontiers in Neuroanatomy87710.3389/fnana.2014.00077HermanABrandvainYWeagleyJJefferyWRKeeneACKonoTJYBilandžijaHBorowskyREspinasaLO’QuinKOrnelas-GarcíaCPYoshizawaMCarlsonBMaldonadoEGrossJBCartwrightRARohnerNWarrenWCMcGaughSE2018The role of gene flow in rapid and repeated evolution of cave-related traits in Mexican tetra, Astyanax mexicanusMolecular Ecology274397441610.1111/mec.1487730252986HillJInderTNeilJDierkerDHarwellJVan EssenD2010Similar patterns of cortical expansion during human development and evolutionPNAS107131351314010.1073/pnas.100122910720624964HofmanMAFalkD2012Evolution of the Primate Brain: From Neuron to BehaviorSpringerHollandLZCarvalhoJEEscrivaHLaudetVSchubertMShimeldSMYuJK2013Evolution of bilaterian central nervous systems: a single origin?EvoDevo42710.1186/2041-9139-4-2724098981HollandLZHollandND2021Cephalochordates: A window into vertebrate originsCurrent Topics in Developmental Biology14111914710.1016/bs.ctdb.2020.07.00133602486HoopsDVidal-GarcíaMUllmannJFPJankeALStait-GardnerTDuchêneDAPriceWSWhitingMJKeoghJS2017Evidence for concerted and mosaic brain evolution in dragon lizardsBrain, Behavior and Evolution9021122310.1159/00047873828869944HuntPWhitingJNonchevSShamMHMarshallHGrahamACookMAllemannRRigbyPWGulisanoM1991The branchial Hox code and its implications for gene regulation, patterning of the nervous system and head evolutionDevelopmentSuppl 263771688181JaggardJBStahlBALloydEProberDADuboueERKeeneAC2018Hypocretin underlies the evolution of sleep loss in the Mexican cavefisheLife7e3263710.7554/eLife.3263729405117JaggardJBLloydEYuiskaAPatchAFilyYKowalkoJEAppelbaumLDuboueERKeeneAC2020Cavefish brain atlases reveal functional and anatomical convergence across independently evolved populationsScience Advances63126314210.1126/sciadv.aba3126JefferyWRMartasianDP1998Evolution of Eye Regression in the Cavefish Astyanax: Apoptosis and the Pax-6 Gene1American Zoologist3868569610.1093/icb/38.4.685JefferyWR2001Cavefish as a model system in evolutionary developmental biologyDevelopmental Biology23111210.1006/dbio.2000.012111180948JefferyWR2008Emerging model systems in evo-devo: cavefish and microevolution of developmentEvolution & Development1026527210.1111/j.1525-142X.2008.00235.x18460088KaracaEHarelTPehlivanDJhangianiSNGambinTCoban AkdemirZGonzaga-JaureguiCErdinSBayramYCampbellIMHunterJVAtikMMVan EschHYuanBWiszniewskiWIsikaySYesilGYuregirOOTug BozdoganSAslanHAydinHTosTAksoyADe VivoDCJainPGeckinliBBSezerOGulDDurmazBCoguluOOzkinayFTopcuVCandanSCebiAHIkbalMYilmaz GulecEGezdiriciAKoparirEEkiciFCoskunSCicekSKaraerKKoparirADuzMBKiratEFenerciogluEUlucanHSevenMGuranTElciogluNYildirimMSAktasDAlikaşifoğluMTureMYakutTOvertonJDYukselAOzenMMuznyDMAdamsDRBoerwinkleEChungWKGibbsRALupskiJR2015Genes that affect brain structure and function identified by rare variant analyses of mendelian neurologic diseaseNeuron8849951310.1016/j.neuron.2015.09.04826539891KidwellCUSuCYHibiMMoensCB2018Multiple zebrafish atoh1 genes specify a diversity of neuronal types in the zebrafish cerebellumDevelopmental Biology438445610.1016/j.ydbio.2018.03.00429548943KoYAmentSAEddyJACaballeroJEarlsJCHoodLPriceND2013Cell type-specific genes show striking and distinct patterns of spatial expression in the mouse brainPNAS1103095310010.1073/pnas.122289711023386717KozolRAJamesDMVarelaISumathipalaSHZüchnerSDallmanJE2021Restoring Shank3 in the rostral brainstem of shank3ab-/- zebrafish autism models rescues sensory deficitsCommunications Biology4141110.1038/s42003-021-02920-634921227KozolRAConithAJYuiskaACree-NewmanATolentinoBBaneshKPazALloydEKowalkoJEKeeneACAlbertsonRCDuboueER2022A Brain-Wide Analysis Maps Structural Evolution to Distinct Anatomical ModulesbioRxiv10.1101/2022.03.17.484801KrumlaufRMarshallHStuderMNonchevSShamMHLumsdenA1993Hox homeobox genes and regionalisation of the nervous systemJournal of Neurobiology241328134010.1002/neu.4802410067901322KunstMLaurellEMokayesNKramerAKuboFFernandesAMFörsterDDal MaschioMBaierH2019A cellular-resolution atlas of the larval zebrafish brainNeuron103213810.1016/j.neuron.2019.04.03431147152LangenbergTBrandM2005Lineage restriction maintains a stable organizer cell population at the zebrafish midbrain-hindbrain boundaryDevelopment1323209321610.1242/dev.0186215958515LevineMDavidsonEH2005Gene regulatory networks for developmentPNAS1024936494210.1073/pnas.0408031102LoomisCPeußRJaggardJBWangYMcKinneySARaftopoulosSCRaftopoulosAWhuDGreenMMcGaughSERohnerNKeeneACDuboueER2019An adult brain atlas reveals broad neuroanatomical changes in independently evolved populations of mexican cavefishFrontiers in Neuroanatomy138810.3389/fnana.2019.0008831636546LustKMaynardAGomesTFleckJSCampJGTanakaEMTreutleinB2022Single-cell analyses of axolotl telencephalon organization, neurogenesis, and regenerationScience377eabp926210.1126/science.abp926236048956MaityTFuseNBeachyPA2005Molecular mechanisms of Sonic hedgehog mutant effects in holoprosencephalyPNAS102170261703110.1073/pnas.050784810216282375MenuetAAlunniAJolyJ-SJefferyWRRétauxS2007Expanded expression of Sonic Hedgehog in Astyanax cavefish: multiple consequences on forebrain development and evolutionDevelopment13484585510.1242/dev.0278017251267MitchellRWW1977Mexican Eyeless Characin Fishes, Genus Astyanax: Environment, Distribution, and Evolution, in Special Publications of the Museum of Texas Tech UniversityTexas Tech PressMolinaMDSalóECebriàF2007The BMP pathway is essential for re-specification and maintenance of the dorsoventral axis in regenerating and intact planariansDevelopmental Biology311799410.1016/j.ydbio.2007.08.01917905225MolnárZKaasJHde CarlosJAHevnerRFLeinENěmecP2014Evolution and development of the mammalian cerebral cortexBrain, Behavior and Evolution8312613910.1159/00035775324776993MontgomerySHMundyNIBartonRA2016Brain evolution and development: adaptation, allometry and constraintProceedings. Biological Sciences283183810.1098/rspb.2016.043327629025MontgomerySHRossiMMcMillanWOMerrillRM2021Neural divergence and hybrid disruption between ecologically isolated Heliconius butterfliesPNAS118e201510211810.1073/pnas.201510211833547240MoranDSoftleyRWarrantEJMistlbergerRE2014Eyeless mexican cavefish save energy by eliminating the circadian rhythm in metabolismPLOS ONE9e10787710.1371/journal.pone.0107877NeubauerSGunzPScottNAHublinJ-JMitteroeckerP2020Evolution of brain lateralization: A shared hominid pattern of endocranial asymmetry is much more variable in humans than in great apesScience Advances6eaax993510.1126/sciadv.aax993532110727NorthcuttRG2002Understanding vertebrate brain evolutionIntegrative and Comparative Biology4274375610.1093/icb/42.4.74321708771O’GormanMThakurSImrieGMoranRLChoySSifuentes-RomeroIBilandžijaHRennerKJDubouéERohnerNMcGaughSEKeeneACKowalkoJE2021Pleiotropic function of the oca2 gene underlies the evolution of sleep loss and albinism in cavefishCurrent Biology313694370110.1016/j.cub.2021.06.07734293332OikonomouGAltermattMZhangRWCoughlinGMMontzCGradinaruVProberDA2019The serotonergic raphe promote sleep in zebrafish and miceNeuron10368670110.1016/j.neuron.2019.05.03831248729ParidaenJTMLDanesinCElasATvan de WaterSHouartCZivkovicD2009Apc1 is required for maintenance of local brain organizers and dorsal midbrain survivalDevelopmental Biology33110111210.1016/j.ydbio.2009.04.02219397905ParsonsKJAlbertsonRC2009Roles for Bmp4 and CaM1 in shaping the jaw: evo-devo and beyondAnnual Review of Genetics4336938810.1146/annurev-genet-102808-11491719691427Pereira-PedroASBrunerEGunzPNeubauerS2020A morphometric comparison of the parietal lobe in modern humans and NeanderthalsJournal of Human Evolution14210277010.1016/j.jhevol.2020.10277032247107Ponce de LeonMSBienvenuTMaromAEngelS2021The primitive brain of early HomoScience37216517110.1126/science.aaz003233833119PortellaACCarvalhoFFaustinoLWondisfordFEOrtiga-CarvalhoTMGomesFCA2010Thyroid hormone receptor beta mutation causes severe impairment of cerebellar developmentMolecular and Cellular Neurosciences44687710.1016/j.mcn.2010.02.00420193766PottinKHinauxHRétauxS2011Restoring eye size in Astyanax mexicanus blind cavefish embryos through modulation of the Shh and Fgf8 forebrain organising centresDevelopment1382467247610.1242/dev.05410621610028RandlettOWeeCLNaumannEANnaemekaOSchoppikDFitzgeraldJEPortuguesRLacosteAMBRieglerCEngertFSchierAF2015Whole-brain activity mapping onto a zebrafish brain atlasNature Methods121039104610.1038/nmeth.358126778924ReardonPKSeidlitzJVandekarSLiuSPatelRParkMTMAlexander-BlochAClasenLSBlumenthalJDLalondeFMGieddJNGurRCGurRELerchJPChakravartyMMSatterthwaiteTDShinoharaRTRaznahanA2018Normative brain size variation and brain shape diversity in humansScience3601222122710.1126/science.aar257829853553Rodriguez-MoralesRGonzalez-LermaPYuiskaAHanJHGuerraYCrisostomoLKeeneACDuboueERKowalkoJE2022Convergence on reduced aggression through shared behavioral traits in multiple populations of Astyanax mexicanusBMC Ecology and Evolution2211610.1186/s12862-022-02069-836241984SanekNAGrinblatY2008A novel role for zebrafish zic2a during forebrain developmentDevelopmental Biology31732533510.1016/j.ydbio.2008.02.02918358464SansaloneGAllenKLedogarJALedogarSMitchellDRProficoACastiglioneSMelchionnaMSerioCMondanaroARaiaPWroeS2020Variation in the strength of allometry drives rates of evolution in primate brain shapeProceedings. Biological Sciences2872020080710.1098/rspb.2020.080732635870SasagawaSTakabatakeTTakabatakeYMuramatsuTTakeshimaK2002Axes establishment during eye morphogenesis in Xenopus by coordinate and antagonistic actions of BMP4, Shh, and RAGenesis33869610.1002/gene.1009512112877SchierAFNeuhaussSCHarveyMMalickiJSolnica-KrezelLStainierDYZwartkruisFAbdelilahSStempleDLRanginiZYangHDrieverW1996Mutations affecting the development of the embryonic zebrafish brainDevelopment12316517810.1242/dev.123.1.1659007238SchlagerS2017Morpho and Rvcg -Shape analysis in RZhengGLiSSzekelyGStatistical Shape and Deformation AnalysisAcademic Press217256SchlagerS2018Morpho: Calculations and Visualizations Related to Geometric MorphometricsCRANSchumacherELCarlsonBA2022Convergent mosaic brain evolution is associated with the evolution of novel electrosensory systems in teleost fisheseLife11e7415910.7554/eLife.7415935713403ShimamuraKRubensteinJL1997Inductive interactions direct early regionalization of the mouse forebrainDevelopment1242709271810.1242/dev.124.14.27099226442SittaramaneVSawantAWolmanMAMavesLHalloranMCChandrasekharA2009The cell adhesion molecule Tag1, transmembrane protein Stbm/Vangl2, and Lamininalpha1 exhibit genetic interactions during migration of facial branchiomotor neurons in zebrafishDevelopmental Biology32536337310.1016/j.ydbio.2008.10.03019013446SneveMHGrydelandHRosaMGPPausTChaplinTWalhovdKFjellAM2019High-expanding regions in primate cortical brain evolution support supramodal cognitive flexibilityCerebral Cortex293891390110.1093/cercor/bhy26830357354SoaresDYamamotoYStricklerAGJefferyWR2004The lens has a specific influence on optic nerve and tectum development in the blind cavefish AstyanaxDevelopmental Neuroscience2630831710.1159/00008227215855759SpitzFGonzalezFPeichelCVogtTFDubouleDZákányJ2001Large scale transgenic and cluster deletion analysis of the HoxD complex separate an ancestral regulatory module from evolutionary innovationsGenes & Development152209221410.1101/gad.20570111544178StoykovaATreichelDHallonetMGrussP2000Pax6 modulates the dorsoventral patterning of the mammalian telencephalonThe Journal of Neuroscience208042805010.1523/JNEUROSCI.20-21-08042.200011050125StriedterGF1998Progress in the study of brain evolution: from speculative theories to testable hypothesesThe Anatomical Record25310511210.1002/(SICI)1097-0185(199808)253:4<105::AID-AR5>3.0.CO;2-S9740033SylvesterJBPottinKStreelmanJT2011Integrated brain diversification along the early neuraxesBrain, Behavior and Evolution7823724710.1159/00032984021860219TallinenTChungJYBigginsJSMahadevanL2014Gyrification from constrained cortical expansionPNAS111126671267210.1073/pnas.140601511125136099UlhaqZSKishidaM2018Brain aromatase modulates serotonergic neuron by regulating serotonin levels in zebrafish embryos and larvaeFrontiers in Endocrinology923010.3389/fendo.2018.0023029867763WarrenWCBoggsTEBorowskyRCarlsonBMFerrufinoEGrossJBHillierLHuZKeeneACKenziorAKowalkoJETomlinsonCKremitzkiMLemieuxMEGraves-LindsayTMcGaughSEMillerJTMommersteegMTMMoranRLPeußRRiceESRiddleMRSifuentes-RomeroIStanhopeBATabinCJThakurSYamamotoYRohnerN2021A chromosome-level genome of Astyanax mexicanus surface fish for comparing population-specific genetic differences contributing to trait evolutionNature Communications12144710.1038/s41467-021-21733-z33664263WartelALindenforsPLindJ2019Whatever you want: Inconsistent results are the rule, not the exception, in the study of primate brain evolutionPLOS ONE14e021865510.1371/journal.pone.021865531329603WatanabeABalanoffAMGignacPMGoldMELNorellMA2021Novel neuroanatomical integration and scaling define avian brain shape evolution and developmenteLife10e6880910.7554/eLife.6880934227464WileDFA2005Evolutionary MorphingProceedings of the IEEE Visualization (VIS ’05431438WilsonLMadenM2005The mechanisms of dorsoventral patterning in the vertebrate neural tubeDevelopmental Biology28211310.1016/j.ydbio.2005.02.02715936325WolffGHThoenHHMarshallJSayreMEStrausfeldNJ2017An insect-like mushroom body in a crustacean braineLife6e2988910.7554/eLife.2988928949916WoychJOrtega GurrolaADeryckereAJaegerECBGumnitEMerelloGGuJJoven ArausALeighNDYunMSimonAToschesMA2022Cell-type profiling in salamanders identifies innovations in vertebrate forebrain evolutionScience377eabp918610.1126/science.abp918636048957YamamotoYJefferyWR2000Central role for the lens in cave fish eye degenerationScience28963163310.1126/science.289.5479.6311091562810.7554/eLife.80777.sa0Editor's evaluationBronnerMarianne EReviewing Editorhttps://ror.org/05dxps055California Institute of TechnologyUnited States
The authors ask if brain regions change based on the functional constraints or developmental constraints. To address this, the authors introduce an automated method for brain segmentation based on the zebrafish tool to study brain evolution in Astyanax.
10.7554/eLife.80777.sa1Decision letterBronnerMarianne EReviewing Editorhttps://ror.org/05dxps055California Institute of TechnologyUnited StatesYoshizawaMasatoReviewerhttps://ror.org/01wspgy28University of Hawaii at ManoaUnited StatesSoaresDaphneReviewerhttps://ror.org/05e74xb87New Jersey Institute of TechnologyUnited States
Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.
Decision letter after peer review:
[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]
Thank you for submitting the paper "Automated anatomical mapping finds brainwide evolution occurring in distinct developmental modules" for consideration by eLife. Your article has been reviewed by 3 peer reviewers and the evaluation has been overseen by a Senior Editor. The following reviewers have agreed to share their identity (Reviewer 2: Masato Yoshizawa; Reviewer 3: Daphne Soares).
We are sorry to say that, after consultation with the reviewers, we have decided that this work will not be considered further for publication by eLife.
Comments to the Authors:
The authors tackle the role of functional and developmental constraints in brain evolution, an important and long-standing question. The approach nicely tests the developmental constraint hypothesis, which was highlighted in their correlation studies of volumetric and shape change, but is less successful with functional constraints. The reviewers appreciate the advance of automated brain characterization including shape changes with the large quantity data on brain shape-genetics but raised questions about the ability to reliably detect smaller subregions. They noted the need to add additional data on dark-raised individuals and also raised questions about novelty. Given these issues and the amount of additional work it requires, the paper seems more appropriate for a specialized journal. Please see the detailed review comments below.
Reviewer #1 (Recommendations for the authors):
To meet criteria for eLife, I would expect that the issues detailed should be addressed, and the following.
Additional external validations of the atlas should be provided. Molecular markers would be ideal for this process.
To identify whether the adaptation to cave environments proceeds by generalizable paths, an atlas for an unrelated Astyanax lineage should be compared.
L153 and Figure 4d,g. What is the statistical support for labeling Cluster 3 as an independent group? Is the node that branches cluster 3 from 2 statistically significantly supported?
Figure 4c. Does mapping the 6 clusters onto a brain reveal regional similarities in co-evolving region sizes?
Figure S5 legend is too sparsely worded to be clear. What the red circles represent should be clearly stated. What do the A and B and X and Y variables represent? These are not defined. "PC1 describes preoptic width, while PC2 describes length." Correct me if I'm wrong, but here I think you mean something more subtle than what's stated. Shouldn't this be something about the loading of these PCs having an over representation of width/length distances. If the simple statement is true, why not report the absolute width and length in addition to the PCA?
Figure 5. As commented above, there is a conflation of PC with a simple length measurement. The optic tectum varies on PC1, and you show it is longer. How much of PC1 is explained by length, for example. Why not provide the length measurement?
Line 198-200. The conclusions from this paragraph are too strong. One issue is why one must invoke developmental constraints in the patterning of the brain. It may be true that expansion of one brain region comes at the expense of another, but these data do not demonstrate this.
Figure 5c,d. The differences between these two graphs are trivial, as I understand them, and there seems to be no new information provided by using two graphs. Why not just provide 5d, with regions labeled on the Y-axis?
Line 219-220. This conclusion is not supported by a direct analysis. Volume covariation and shape covariation are not directly compared. Rather, in the volumetric analysis, only 4 regions are analyzed, rather than 13 for the shape analysis. These should be analysed in the same way in order to make this statement. But even this would be simply a correlation.
But further, could a similar covariation (if found) be simply a result of the fact that volume is a function of the parameters (e.g., length, width) analysed during shape analysis? As a result, shape and volume are not orthogonal variables for which a finding of distinct developmental mechanisms would be compelling.
Line 260 and elsewhere. "Our data show that the dorsal-caudal areas of the brain evolve together…" This work shows that in the Pachon line has evolved in this way. The sample size is derived from F2 animals from a outcross of this line, and as such the separate animals represent unique genetic combinations of Pachon alleles, but it is from a single evolutionary trajectory. As such, additional examples of evolution (for example from another cave) may operate by changing other brain region sizes. I see that this caveat is then raised (L 266), but the conclusions remain overstated in this paper elsewhere.
Reviewer #2 (Recommendations for the authors):
This is not the authors' fault, but there is confusion between these two hypotheses in general – both (functional and developmental) can be regulated by genes. Developmental constraints can act brain-wide and local (like islet1/2 expressing domain). Functional constraints can also act on local-nearby regions but also long distances such as between the cerebrum and cerebellum (co-evolution of cerebrum and cerebellum were presented as an example of the functional constraints). Thus, brain-wide or not is not such a good criterion to argue these two hypotheses although Montgomery et al., 2016 mentioned. To me, these two hypotheses are proxies of the question of whether the brain regions are regulated by developmental genes (Pax6, Hoxs, Otx etc) vs. functional axonal/synaptic genes (Netrin, semaphorin, GluN receptors etc). The authors briefly mentioned that cluster3 (subpallium and dorsal diencephalon) could be an example of the functional constraints but it was with not so positive reasons; that is, this region was not constrained by other regions (L258-259). If the authors would like to mention the functional constraints, I feel they need to show positive data, such as neural activities are not tightly associated with other brain regions compared with between the clusters 1 and 2.
My suggestion is that the authors may need to restate their research question to align what they have in the data.
here are the other points I am concerned about:
L34-39: "Two central hypotheses are thought to drive anatomical brain evolution;…"
I am afraid that the sentence for the developmental constraint hypothesis is a little misleading by stating "..change together in a concerted matter." many studies showed the brain develops in a mosaic fashion. I guess the authors would like to state as "the developmental constraints hypothesis suggests that most of the individual brain components tend to evolve together" as Montgomery et al., 2016 stated.
Also the authors mentioned 'mosaic' relating to the functional constraints hypothesis, yet, in developmental biology, 'mosaic' is also used for developmental processes. I suggest the authors clarify whenever they use 'mosaic' for functional or developmental.
L101: "…brain regions (Figure 2 – Supplement 1a and b)"
need reference
Figure 4e and 4f, the figures were swapped according to the Main body text and Figure 5 figure legend.
Figure 5—figure supplement 1, difficult to understand where the landmarks were from. Please superimpose the preoptic area's image on the landmarks.
I do not see if the original image data and other related data will be shared or not.
Reviewer #3 (Recommendations for the authors):
I thoroughly enjoyed this manuscript, I think it can eventually be a fantastic paper! I would be very happy to look at it again and I have no problems with the data. Please understand that my criticism is basically saying that I don't think you completely know HOW interesting/impactful your results are because you do not address to the long arguments and disagreements in the literature. I think you should spend a while really dissecting what people have been thinking about the evolution of vertebrate brain. Read the vast body of literature on brain evolution, the works of William Hodos, Ann Buttler, Georg Strieder and Glenn Northcutt especially come to mind right away. Your results based on their original work can be that much more insightful. These established authors have raised many questions which you can chime in with your new techniques. It will be a wasted opportunity to just argue that astyanax is a good model, show the reader WHY it is a good model, what do the results mean. I think then it will elevate the paper to an eLife level.
I am including one specific set of points on the abstract that illustrate the same issues again and again in the entire text (I'll be blunt because it is easier, keep in mind that I am a fan of the work):
"Brain anatomy is highly variable and it is widely accepted that anatomical variation impacts brain function and ultimately behavior."
I think the initial statement of the abstract is already confounding. You are already not bringing evolutionary neuroscientists to your side. It is too simplistic. I think anyone who knows anything about the issue will either dismiss you, thinking that you have not done your homework in terms the literature, or that your knowledge of the subject is not sophisticated. There is an enormous body of evidence on the literature that show how conserved brains actually are. Diversity and homology are actually one of the main tenants of MANY brain evolutionary studies. Also revise the sentence to avoid repetition of "anatomy/anatomical" for stylistic reasons.
"The structural complexity of the brain, including differences in volume and shape, presents an enormous barrier to define how variability underlies differences in function."
I respectfully disagree, the challenge is no longer in this level of analysis. Neurons themselves, circuits and modulation are the next frontier that will inform our questions. I think the first 2 sentences are weak and do not buttress the argument of why this study is important and novel. I would recommend to rewrite it.
"In this study, we sought to investigate the evolution of brain anatomy in relation to brain region volume and shape across the brain of a single species with variable genetic and anatomical morphs."
Again this has been done for years, at this point of reading, I am not convinced why this is novel enough to give insights to the audience of eLife. I think here you really have the chance to drive your point that your results are really important, but you don't yet have the knowledgeable reader on your side.
[Editors’ note: further revisions were suggested prior to acceptance, as described below.]
Thank you for resubmitting your work entitled "Automated anatomical mapping finds brainwide evolution occurring in distinct developmental modules" for further consideration by eLife. Your revised article has been evaluated by Marianne Bronner (Senior Editor) and a Reviewing Editor.
The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:
1. The authors must check whether the eye-size is a confounding factor of the brain region volume changes. This could be done by a correlation analysis of eye-size vs brain regions/shape (PCA1 or PCA2 axis value).
2. The discussion and references on brain evolution need to be significantly expanded.
3. Please attend to the more minor revisions raised by the reviewers.
Reviewer #1 (Recommendations for the authors):
Thank you to the authors for their thoughtful responses to my concerns. I find the manuscript to be significantly improved. I feel that the arguments they make are now well-supported. I have a few relatively minor comments on places where the writing was not clear.
Is the segmentation by tERK with labels for region-specific markers done with ONLY the tERK information? Is there any chance that signal from the regional marker could influence the segmentation?
Lines 126-132. The figure references do not refer to the correct panels (wrong genes). Also, anti-TH is not present in any figure; please add this.
Reviewer #2 (Recommendations for the authors):
This is a significant and excellent update to the original submission by Kozol et al. Most of the reviewers' concerns are resolved with a much improved clearer logical flow and newly added Molino cavefish and F2 hybrid data, further adding detailed shape and volume analyses. There are one major concern and several minor points in this resubmitted version of their manuscript.
Although the volume change analyses were now more elaborated, it is unclear if the eye size is a correlated factor to explain the antagonistic volume-changes between the dorso-caudal and ventro-rostral brain regions. The authors explained that this pattern of volume change is the result of developmental/genetic constrain with new F2 data, which is convincing. Yet, the major sensory difference at 6 days old – functional vs non-functional eyes (surface fish and cavefish, respectively) – could affect the brain anatomical changes. I assume that authors took whole body pictures before dissection (or, eyes were scanned together during brain imaging) and the eye-size parameter is available to test whether the whole brain landscape is basically correlated with the eye size. None of the other sensors including the olfactory epithelium, lateral line, and inner ear, was reported with significant differences between surface fish and cavefish, as far as I know. I believe the eye size is the important factor in arguing the brain architecture while I was reading through this manuscript, I assumed the brain developmental genes (otx, islet, emx) might regulate these changes but eye developmental genes (pax 6, lens genes) could be the major driver of these brain morphological changes, instead. To answer their research question, whether the whole brain regions or individual brain regions evolved, I believe the eye is the important brain region the authors should consider. F2 hybrids would be a great resource to check it.
10.7554/eLife.80777.sa2Author response
[Editors’ note: The authors appealed the original decision. What follows is the authors’ response to the first round of review.]
Reviewer #1 (Recommendations for the authors):
To meet criteria for eLife, I would expect that the issues detailed should be addressed, and the following.
Additional external validations of the atlas should be provided. Molecular markers would be ideal for this process.
We agree that an addition of additional markers would make the manuscript stronger for publication. We have now added 5 additional markers. This involved not only adding additional antibody labels, but also devising an HCR protocol that was compatible with brain registration. This classically has been challenging as the harsh molecular washes for in situ hybridization often warp and distort brain tissue such that it cannot be correctly registered. HCR alleviated those concerns, and made for a high throughput strategy for labeling different molecular targets with known expression patterns. The addition of these markers show that the tested markers were bound to the segmented regions which were automatically computed by the program ANTs. Methods and results can be found in Figure 1, Figure 1 —figure supplement 1, and lines 105-118 and 422-440. We believe the addition of these experiments strengthens the validity of our atlas and will be incorporated into the downloadable atlas package.
To identify whether the adaptation to cave environments proceeds by generalizable paths, an atlas for an unrelated Astyanax lineage should be compared.
We agree that assessing another population of fish is a valuable resource, and is needed for claims about ‘generalizability.’ To address this concern, we have repeated our methods that were previously presented for surface fish and Pachón cavefish on another, independently evolved cavefish population, Molino. This is a unique cave, in that it evolved completely independent of the Pachón population in a second migration sweep of surface fish invading caves. Our findings in Molino with brain anatomy in pure populations as well as F2 hybrids parallels our findings in Pachón, suggesting that a ventral expansion and a dorsal contraction of the brain is found in multiple cavefish. These findings and discussions can be found in Figure 2, Figure 3, Figure 4 and lines 132-148, 151-165, 169-188, 189-210, and 289-294.
While claims about evolutionary generalization cannot be made without. an exhaustive assessment of various animals that span the kingdom, similar findings in independently evolved Astyanax morphs suggests that these patterns of brain evolution are likely a generalized feature in these Mexican cavefish.
L153 and Figure 4d,g. What is the statistical support for labeling Cluster 3 as an independent group? Is the node that branches cluster 3 from 2 statistically significantly supported?
The clustering in these studies were performed using hierarchical clustering, and while this approach is a widely used approach for clustering, it does not set the number of clusters, per se. The number of clusters from a hierarchical clustering approach can be as few as one (everything in one cluster) or as many as the number of individuals represented. For instance, the cluster maximum for the surface to Pachón hybrids was 6, while the surface to Molino hybrids was 12. Our cutoff was based on that branch point that gave the most distinct grouped clusters. In order to established statistical basis for calling these groups separate, we tabulated those R-values in each cluster and examined whether there was statistical significance suggesting that they were indeed different. Unsurprisingly, the t [13.82] and p-value [>0.0001] for cluster 2 and 3 revealed that the two were significantly different, suggesting the two groups are indeed different. We have added statistical tables for all comparisons to the supplemental material, Figure 4 —figure supplement 5-10.
Figure 4c. Does mapping the 6 clusters onto a brain reveal regional similarities in co-evolving region sizes?
We thank the reviewer for the question and have gone back to map our 6 clusters onto the Astyanax atlas. We originally only mapped back the large subdivisions of our atlas due to labor constraints, but have now written additional code that maps back brain regions for all 180 areas into their respective clusters. This not only allowed us to address this reviewers concern, but we hope that it serves as a resource for the community.
After mapping all 180 regions that fell into 6 clusters back onto the brain, we found the same trend as we did with our 13 brain region clusters, specifically that all 6 clusters are arranged in a dorsal ventral fashion and fell into groups of closely adjacent regions. We also found that dorsal to ventral comparisons of the 6 clusters also showed strong negative associations similar to those found across cluster 1 and 2 of the 13 brain region atlas. We then repeated this analysis in surface to Molino F2 hybrids and found that brain regions are grouped in 12 distinct dorsal to ventral clusters, with similar positive and negative relationships found in surface to Pachon F2 hybrids. We have now replaced Figure 4 with a new figure depicting the un-clustered and clustered arrays of the 180 brain regions, along with projection maps and coronal optical sections providing examples of the clustered regions. The previous Figure 4 has now been added to the supplemental data as Figure 4 – Supplement 1. Results and discussion text can be found in lines 169-188, 189-210, and 284-294.
Figure S5 legend is too sparsely worded to be clear. What the red circles represent should be clearly stated. What do the A and B and X and Y variables represent? These are not defined. "PC1 describes preoptic width, while PC2 describes length." Correct me if I'm wrong, but here I think you mean something more subtle than what's stated. Shouldn't this be something about the loading of these PCs having an over representation of width/length distances. If the simple statement is true, why not report the absolute width and length in addition to the PCA?
We apologize for the sparse wording and mislabeling of the figure axes. We have added additional descriptive information into the figure legend to assist a reader in understanding our PCA plots. We have also expanded this legend to more clearly state the data presented. We apologize for this oversight, and thank the reviewer for calling our attention to it.
Figure 5. As commented above, there is a conflation of PC with a simple length measurement. The optic tectum varies on PC1, and you show it is longer. How much of PC1 is explained by length, for example. Why not provide the length measurement?
The reviewer is correct that our presentation of these data was misleading. We meant to use PCA to determine statistical relationships between the 3d volumetric shape, and to show, broadly, how these varied in both x and y axis dimensions. However, our description in the text and generalized arrows along physical axes of the brain regions was a mistake. Our intent was not to conflate, but rather to make it easy for a reader to interpret. We have since edited Panels a and b in the figure, and added supplementary figures, Figure 5 —figure supplement 3-6, that provide wire diagrams and landmarks over the projected brain regions to elucidate the 3d complexity being analyzed.
In regards to the PCA, the PCA was generated by placing reference points at pre-defined areas on each neuroanatomical locus, applying 3-dimentional morphometrics, and then using principal component analysis of the high dimensionality data set to determine which principle components explained the largest amount of the difference between the two. The PCA showed the statistical relationship between morphs, and revealed that, as would be expected for a genetically determined trait, that F2 hybrids showed intermediary values that spanned the range of surface cave. Often times, we found that a given principal component aligned with an expansion/contraction along either the x- or y-axis. Other times, the PCA explained a curvy-linear axis, as was the case for the cerebellum (in Figure 5b), and we highlighted that. However, the reviewer is correct: because the PCA was performed on 3d morphometrics and not linear data sets, true relationships between 3d data and x- and y-axis cannot be made.
But further, could a similar covariation (if found) be simply a result of the fact that volume is a function of the parameters (e.g., length, width) analysed during shape analysis? As a result, shape and volume are not orthogonal variables for which a finding of distinct developmental mechanisms would be compelling.
3D shape analysis is certainly a different parameter than volume, yet the two are both continuous variables. We show that the two continuous values correlate beyond statistical significance and believe this suggests that the two are likely related to each other. We do believe, however, as I think the reviewer is stating, that we could be capturing some measure of volume within our shape parameter, and we now make this point in the discussion (lines 386-387).
Line 198-200. The conclusions from this paragraph are too strong. One issue is why one must invoke developmental constraints in the patterning of the brain. It may be true that expansion of one brain region comes at the expense of another, but these data do not demonstrate this.
We agree that the statement was too strong, in lieu of cluster mapping to demonstrate a concomitant tradeoff between dorsal-ventral poles of the brain. However, with the addition of the Molino cave analysis and subsequent mapping of hybrid brain clusters, we believe the data and analysis now supports this conclusion. We have edited the language in the text to now state “an anatomical tradeoff” that appears convergent across cave populations. These changes are now found in lines 167 and 187.
Figure 5c,d. The differences between these two graphs are trivial, as I understand them, and there seems to be no new information provided by using two graphs. Why not just provide 5d, with regions labeled on the Y-axis?
While these two graphs are very similar and we are happy to drop panel 5c we believe that for the sake of transparency, showing clustered and unclustered graphs are useful for the reader, and does not take away from the study. Again, we are happy to remove if needed, but have left unchanged as of now for transparency sake.
Line 219-220. This conclusion is not supported by a direct analysis. Volume covariation and shape covariation are not directly compared. Rather, in the volumetric analysis, only 4 regions are analyzed, rather than 13 for the shape analysis. These should be analysed in the same way in order to make this statement. But even this would be simply a correlation.
We agree with the reviewer that a direct comparison was not performed and have since softened the language throughout the paragraph (line 369-372). This includes pointing out that this is a first effort and will need to analyze the entire atlas in future work. Shape is significantly more laborious to assess, and we were not able to look at shape for every region. Moreover, we believe that the volume is a resource to the community, whereas shape was assessed to try to determine whether there was a relationship. We have also highlighted this discrepancy in the discussion (lines 418 to 420). We did however directly compare dimension reduced 3D morphometric values to volume in 6 regions of the 13 region atlas (Figure 5 —figure supplement 9 and lines 254-264). We hope these changes and the transparency of analytical terms satisfies the valid concerns of the reviewer.
Line 260 and elsewhere. "Our data show that the dorsal-caudal areas of the brain evolve together…" This work shows that in the Pachon line has evolved in this way. The sample size is derived from F2 animals from a outcross of this line, and as such the separate animals represent unique genetic combinations of Pachon alleles, but it is from a single evolutionary trajectory. As such, additional examples of evolution (for example from another cave) may operate by changing other brain region sizes. I see that this caveat is then raised (L 266), but the conclusions remain overstated in this paper elsewhere.
We agree that these statements were overstated without analyzing additional cave populations. We have since analyzed both pure Molino cavefish and surface to Molino cave F2 hybrids. Molino pure and hybrid data was nearly identical to that of Pachón cavefish and suggests this dorsal-ventral shift in brain mass is likely a generalizable trait for Astyanax mexicanus.
Reviewer #2 (Recommendations for the authors):
This is not the authors' fault, but there is confusion between these two hypotheses in general – both (functional and developmental) can be regulated by genes. Developmental constraints can act brain-wide and local (like islet1/2 expressing domain). Functional constraints can also act on local-nearby regions but also long distances such as between the cerebrum and cerebellum (co-evolution of cerebrum and cerebellum were presented as an example of the functional constraints). Thus, brain-wide or not is not such a good criterion to argue these two hypotheses although Montgomery et al., 2016 mentioned. To me, these two hypotheses are proxies of the question of whether the brain regions are regulated by developmental genes (Pax6, Hoxs, Otx etc) vs. functional axonal/synaptic genes (Netrin, semaphorin, GluN receptors etc). The authors briefly mentioned that cluster3 (subpallium and dorsal diencephalon) could be an example of the functional constraints but it was with not so positive reasons; that is, this region was not constrained by other regions (L258-259). If the authors would like to mention the functional constraints, I feel they need to show positive data, such as neural activities are not tightly associated with other brain regions compared with between the clusters 1 and 2.
My suggestion is that the authors may need to restate their research question to align what they have in the data.
These hypotheses are not singularly defined in the field, and interpretations of what developmental and functional means often varies from one usage to another. We agree with the reviewer that an argument for a “functional constraint” would be strengthened by functional experiments, and that arguments for developmental would be augmented by developmental genes such as Pax6 or Nkx2.2, while arguments for function would be bolstered by corresponding mechanisms involving ‘functional’ genes such as Netrin and semaphorins. While this is a long-term goal of the project, we are far away from having these results.
Due to these limitations, we have decided to remove the section arguing for functional constraint of cluster 3, and have tried to be more cautious in our arguments for these two hypotheses throughout the manuscript.
here are the other points I am concerned about:
L34-39: "Two central hypotheses are thought to drive anatomical brain evolution;…"
I am afraid that the sentence for the developmental constraint hypothesis is a little misleading by stating "..change together in a concerted matter." many studies showed the brain develops in a mosaic fashion. I guess the authors would like to state as "the developmental constraints hypothesis suggests that most of the individual brain components tend to evolve together" as Montgomery et al., 2016 stated.
The reviewer’s interpretation of our intent is correct, and we have tried to clarify the manuscript to better reflect that. The text now reads, “, that most of the brains regions tend to evolve together” (lines 37-42).
Also the authors mentioned 'mosaic' relating to the functional constraints hypothesis, yet, in developmental biology, 'mosaic' is also used for developmental processes. I suggest the authors clarify whenever they use 'mosaic' for functional or developmental.
We understand the reviewer’s confusion and have edited the text that describes the two hypotheses. We apologize for the vagueness and hope our revised text satisfies this reviewers concern. Additionally, the term mosaic was removed from the manuscript, resulting in the singular usage of functional constraint to describe this hypothesis.
L101: "…brain regions (Figure 2 – Supplement 1aandb)"
need reference
The main body text has been updated to include the correct references. The text now reads, “(Figure 2 – Supplement 1aandb, Figure 2 —figure supplement 2-17) [49,50]." (line 347-348).
Figure 4e and 4f, the figures were swapped according to the Main body text and Figure 5 figure legend.
We have edited the main body text and figure legends to reflect changes in the figures themselves. Figure 4 is now a cluster analysis of the two hybrid populations. The initial Figure 4 is now Figure 4 – Supplement 1, with the addition of the Molino cavefish hybrid population data.
Figure 5—figure supplement 1, difficult to understand where the landmarks were from. Please superimpose the preoptic area's image on the landmarks.
We apologize for leaving out the preoptic area’s image. We have edited the figure to include mesh models throughout the PCA axes and 3d projections of the preoptic area’s image with the landmarks. This was repeated for all shape regions and added as Figure Supplements to Figure 5 (Figure 5 —figure supplement 1-6).
I do not see if the original image data and other related data will be shared or not.
There is a data sharing section that has the dryad DOI address where all scripts, raw data, analyzed data and tables can be found. We have also uploaded all code used throughout for the reviewers and readers.
Reviewer #3 (Recommendations for the authors):
I thoroughly enjoyed this manuscript, I think it can eventually be a fantastic paper! I would be very happy to look at it again and I have no problems with the data. Please understand that my criticism is basically saying that I don't think you completely know HOW interesting/impactful your results are because you do not address to the long arguments and disagreements in the literature. I think you should spend a while really dissecting what people have been thinking about the evolution of vertebrate brain. Read the vast body of literature on brain evolution, the works of William Hodos, Ann Buttler, Georg Strieder and Glenn Northcutt especially come to mind right away. Your results based on their original work can be that much more insightful. These established authors have raised many questions which you can chime in with your new techniques. It will be a wasted opportunity to just argue that astyanax is a good model, show the reader WHY it is a good model, what do the results mean. I think then it will elevate the paper to an eLife level.
I am including one specific set of points on the abstract that illustrate the same issues again and again in the entire text (I'll be blunt because it is easier, keep in mind that I am a fan of the work):
We agree with the reviewer that an important discussion is needed on current unresolved topics underlying evolution of the brain and how the cavefish fills a need to study the mechanisms underlying anatomical and functional evolution of the brain. We have since edited the introduction significantly (especially lines 79-90) that provides background on how the brain is thought to evolve anatomically, along with two paragraphs in the discussion , (1) briefly discussing how unresolved questions can be addressed going forward utilizing non-traditional methods for neurological inquiry, such as the fossil record (e.g. soft tissue is poorly preserved) (line 330-346), and (2) recent studies in other non-model fish species that are providing a way forward for understanding and testing the functional significance of convergent changes in neuroanatomical evolution (lines 348-365).
"Brain anatomy is highly variable and it is widely accepted that anatomical variation impacts brain function and ultimately behavior."
I think the initial statement of the abstract is already confounding. You are already not bringing evolutionary neuroscientists to your side. It is too simplistic. I think anyone who knows anything about the issue will either dismiss you, thinking that you have not done your homework in terms the literature, or that your knowledge of the subject is not sophisticated. There is an enormous body of evidence on the literature that show how conserved brains actually are. Diversity and homology are actually one of the main tenants of MANY brain evolutionary studies. Also revise the sentence to avoid repetition of "anatomy/anatomical" for stylistic reasons.
We agree with the reviewer that the vertebrate brain, or bauplan of the brain, has remained remarkably conserved. However, variation in the size, shape and function of individual brain regions can be highly variable and contributes greatly to the amazing diversity of behaviors we see across taxa. We have edited the text in the abstract and introduction to help clarify this point (lines 12-19).
"The structural complexity of the brain, including differences in volume and shape, presents an enormous barrier to define how variability underlies differences in function."
I respectfully disagree, the challenge is no longer in this level of analysis. Neurons themselves, circuits and modulation are the next frontier that will inform our questions. I think the first 2 sentences are weak and do not buttress the argument of why this study is important and novel. I would recommend to rewrite it.
We agree with the reviewer that the function and modulation of circuits is fundamental to understanding behavioral evolution. We have edited the text to state that anatomy and function together create obstacles for determining how evolution of the brain results in novel behaviors. We have since rewritten the abstract and focus on how anatomical variation can lead to functional and ultimately behavioral variation (see lines 12-15 and 35-40)
"In this study, we sought to investigate the evolution of brain anatomy in relation to brain region volume and shape across the brain of a single species with variable genetic and anatomical morphs."
Again this has been done for years, at this point of reading, I am not convinced why this is novel enough to give insights to the audience of eLife. I think here you really have the chance to drive your point that your results are really important, but you don't yet have the knowledgeable reader on your side.
We agree with the reviewer that our arguments and language did not convey to the reader how the unique characteristics of this model, a single species with variable genetic and anatomical morphs, were leveraged through hybridization to functionally test how every brain region anatomically relates to one-another. We have since changethe sentence to read, “In this study, we sought to investigate the evolution of brain anatomy using a single species of fish consisting of divergent surface and cave morphs, that permits functional genetic testing of regional volume and shape across the entire brain.”(line 17-19) In addition, we believe that added paragraphs addressing previous concerns from the introduction and discussion will help the reader grasp the importance of this model in addressing how anatomical change impacts organismal function and behavior.
[Editors’ note: what follows is the authors’ response to the second round of review.]
The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:
1. The authors must check whether the eye-size is a confounding factor of the brain region volume changes. This could be done by a correlation analysis of eye-size vs brain regions/shape (PCA1 or PCA2 axis value).
We address this comment below. This is a thoughtful comment and an experiments worth doing, but we did not capture images of the eye. Moreover, because the eyes were not the focus of our imaging, in many samples the eyes extended beyond the focal plane. In other words, the full eyes were not captured in our confocal images, precluding any correlation. This however has been the focus of a collaborative study with the Kowalko lab. While these details are complex for this rebuttal, it is noteworthy that not all regions of the brain correlate with eyes.
We agree this is an important point and have highlighted this in the discussion.
2. The discussion and references on brain evolution need to be significantly expanded.
Thank you. We have now added paragraphs to the discussion. Specifically, we have commented on brain development and how our work fits into this large field; we comment on the potential relationship with eyes; and we comment on how expanded brain regions evolve.
Reviewer #1 (Recommendations for the authors):
Thank you to the authors for their thoughtful responses to my concerns. I find the manuscript to be significantly improved. I feel that the arguments they make are now well-supported. I have a few relatively minor comments on places where the writing was not clear.
Is the segmentation by tERK with labels for region-specific markers done with ONLY the tERK information? Is there any chance that signal from the regional marker could influence the segmentation?
No, the ANTs registration to provide the reformatting instructions for overlaying a subject brain to the atlas brain is only using the tERK stained image. Once finished, the registration code produces three files that are used to inverse register the segmented atlas file back onto the subject brain. Therefore, the regional marker files are never part of the processes.
Lines 126-132. The figure references do not refer to the correct panels (wrong genes). Also, anti-TH is not present in any figure; please add this.
We apologize to the reviewer and have updated the text and figure legend. We have added nitrous oxide 1 and the missing anti-tyrosine hydroxylase image to the multi-panel figure. The text now reads,
Line 126- 132, “We then tested the accuracy of smaller regions, such as the dorsal subpallium, medial preoptic region, and thalamus, via gastrulation brain homeobox 1 (gbx1), oxytocin (oxt) and nitrous oxide 1 (nos1) RNA labeling, respectively (Figure 1 – Supplement 3b-d, [35,36,42,43]). Finally, we confirmed the accuracy of the smallest sub-regions of the brain that can be defined molecularly, such as the locus coeruleus and dorsal raphe, using tyrosine hydroxylase (TH) and 5-hydroxytryptamine (5-HT) antibody labeling, respectively (Figure 1 – Supplement 3eandf, [33,35,44-47]).”
Reviewer #2 (Recommendations for the authors):
This is a significant and excellent update to the original submission by Kozol et al. Most of the reviewers' concerns are resolved with a much improved clearer logical flow and newly added Molino cavefish and F2 hybrid data, further adding detailed shape and volume analyses. There are one major concern and several minor points in this resubmitted version of their manuscript.
Although the volume change analyses were now more elaborated, it is unclear if the eye size is a correlated factor to explain the antagonistic volume-changes between the dorso-caudal and ventro-rostral brain regions. The authors explained that this pattern of volume change is the result of developmental/genetic constrain with new F2 data, which is convincing. Yet, the major sensory difference at 6 days old – functional vs non-functional eyes (surface fish and cavefish, respectively) – could affect the brain anatomical changes. I assume that authors took whole body pictures before dissection (or, eyes were scanned together during brain imaging) and the eye-size parameter is available to test whether the whole brain landscape is basically correlated with the eye size. None of the other sensors including the olfactory epithelium, lateral line, and inner ear, was reported with significant differences between surface fish and cavefish, as far as I know. I believe the eye size is the important factor in arguing the brain architecture while I was reading through this manuscript, I assumed the brain developmental genes (otx, islet, emx) might regulate these changes but eye developmental genes (pax 6, lens genes) could be the major driver of these brain morphological changes, instead. To answer their research question, whether the whole brain regions or individual brain regions evolved, I believe the eye is the important brain region the authors should consider. F2 hybrids would be a great resource to check it.
We thank the reviewer for this insightful comment. Unfortunately, we did not take images of the eyes before fixing and imaging. Moreover, analyzing eye size from confocal images presented challenges in itself: Because the eyes were not the focus of the image, it is difficult to say where the eye starts and where it ends. Moreover, because the eyes protrude well beyond the brain, we often did not image the entire eye as they were out of the x-y focal range. We feel that making conclusions from these data would be inconstant.
We have begun to address this in a subsequent collaborative study with the Kowalko lab, which we are hoping to put on a pre-print server soon. The focus of this subsequent study is on the relationship between eyes and different regions of the brain. While this is a large study and the findings are out of the scope of this manuscript, it is noteworthy that not all regions that change correlate with eyes, suggesting more complex underlying mechanisms.