To understand the evolution of the gasdermin family we reconstructed a phylogeny of gasdermin proteins. We screened a total of 3443 genomes covering major classes of animal orders from public databases using gasdermin-specific hidden Markov models to identify all putative gasdermin genetic loci using trained gene predictors. We uncovered 106 previously unannotated GSDM proteins. Our final catalog included 1256 gasdermin protein sequences. Gasdermin-like proteins have been reported in bacteria in phage defense islands alongside activating proteases (Johnson et al., 2022; Wein and Sorek, 2022), and in fungi (Daskalov et al., 2020). We limited our searches to animals because sequences from other kingdoms are distantly similar (Yuan et al., 2022; Angosto-Bazarra et al., 2022). Our approach extends previous studies by using a more comprehensive dataset, which allows us to more accurately retrace gasdermin gene family evolution withing the context of vertebrate radiation.

In comparison to previously published studies, we used different methods to construct our gasdermin phylogenetic tree, with the result that our tree has a different topology. The topology of our tree is likely to be affected by our increased sampling of gasdermin sequences; we included 1256 gasdermin sequences in comparison to 300 or 97 sequences used in prior studies. Prior studies used maximum likelihood tree building techniques, whereas we used a more comprehensive estimation of gasdermin divergence time using Bayesian method implemented in BEAST with strict molecular clocks using fossil calibrations with soft bounds. We think that this substantially increased sampling paired with time calibration allow us to produce a more accurate phylogeny of the gasdermin protein family. We constructed a well-supported Bayesian tree with 71% of the nodes with posterior probability >0.8 and a minimum age of protogasdermin emergence at 1100 million years with 95% higher-posterior density intervals (Figure 1).

Similar to other trees, GSDME and PJVK cluster together in a clade separate from the clade containing GSDMA–D and GSDMEc. However, the internal order within this second clade was divergent from other trees, described in more detail below.

Near the midpoint of our tree is a gasdermin that is preserved in mollusks, coral, hydras, and other invertebrates referred to as GSDM_in. For such species, caspase-3 cleaves this gasdermin to execute pyroptosis, much like GSDME (Jiang et al., 2020; Qin et al., 2023; Chen et al., 2023).

The first gene duplication of this gasdermin occurred after the divergence of lancelets and chordates resulting in GSDME and PJVK. Bony fish encode both caspase-1 and -3, and both cleave GSDME to execute pyroptosis (Jiang et al., 2019; Chen et al., 2021b; Xu et al., 2022; Chen et al., 2021a; Wang et al., 2020a; Li et al., 2020). Similarly, lancelets cleave GSDME using caspase-1- and caspase-3-like (Wang et al., 2023b). In contrast, PJVK developed a new function of promoting peroxisome proliferation in response to excess reactive oxygen species, utilizing a shorter C-terminus that detects ROS (Delmaghani et al., 2015; Defourny et al., 2019). Although it has not been shown yet, this function is likely related to a pore-forming ability. We also found additional ancient gasdermin sequences in bony fish with a GSDM C-terminal domain, but sometimes without the GSDM N-terminal domain, that appeared in a clade that branches off before the divergence of GSDME and PJVK (Figure 1, purple, GSDMEc-like).

An ancient ancestor of GSDME was duplicated and gave rise to the second gasdermin clade (sometimes called the GSDMA family). This gasdermin is not universally present in bony fish, but in some it is preserved as GSDMEc (Yuan et al., 2022). This gasdermin was recently identified as a potential ortholog of GSDMA. It was called GSDMEc, following the nomenclature of other duplications of GSDME in bony fish which have been named GSDMEa and GSDMEb. Our method brought to light this gasdermin is present broadly in sharks, which appeared in a separate clade near the root of the GSDMA family proteins (Figure 1, purple, GSDMEc Shark). This discovery places the emergence of GSDMEc at the common ancestor of jawed vertebrates. The activating protease for the GSDMEc proteins remains undetermined, but they do not contain apparent caspase cleavage motifs.

GSDMEc then gave rise to GSDMA in the common ancestor of amphibians, mammals, reptiles, and birds (De Schutter et al., 2021; Angosto-Bazarra et al., 2022; Yuan et al., 2022). No additional gasdermins are found in amphibians or birds, therefore, these clades encode GSDME, PJVK, and GSDMA. Among these, GSDME is activated by caspase-3 to execute pyroptosis (Li et al., 2022). Our and other molecular phylogeny analysis identified three amphibians in the caecilian order with GSDMA genes (Yuan et al., 2022; Wang et al., 2023b). GSDMA is absent in frogs and salamanders, including amphibian model organisms with well characterized genomes like Xenopus tropicalis (amphibian, Western clawed frog). This observation suggests that GSMDA is present in the common ancestor of tetrapods, but was lost in most amphibians. In support of this, caecilian GSDMA resides in the same locus as mammal, reptile, and bird GSDMA. The proteases that activate GSDMA in these clades had not been investigated.

We and others identified a poorly characterized gasdermin in the GSDMB locus, which is separated by only two genes from GSDMA which we have labeled as GSDMB-like, despite little sequence similarity with mammal GSDMB (Wang et al., 2023b). This gasdermin grouped in a clade alongside fish and shark GSDMEc. The presence of this gasdermin in the same locus as mammal GSDMB in reptiles before and after the divergence of turtles, crocodiles, and birds from lizards and snakes suggests that this gasdermin was present in the common ancestor reptiles and birds.

Previous trees suggest that mammal GSDMA and GSDMB evolved in a separate clade from GSDMC and GSDMD (De Schutter et al., 2021; Wang et al., 2023b). Surprisingly, in our tree, amphibian, bird, and reptile GSDMA are paralogs to mammal GSDMA rather than orthologs. Both our Bayesian (Figure 1) and maximum likelihood (Figure 1—figure supplement 3) trees produced this result, suggesting this placement is due to increased sampling of gasdermin sequences. Importantly, the GSDMA gene is found in the same locus in all animals in these clades (Figure 1—figure supplement 4). This suggests that the common ancestor of amphibians and mammals duplicated GSDMA first to give rise to mammal GSDMB, then sequentially gave rise to GSDMC, and then GSDMD, leaving the syntenic GSDMA gene at the ancestral locus.

The absence of GSDMB and GSDMC in the ancient monotreme mammals at first glance appears to be in conflict with this sequence of evolution. Our tree reconciles this observation by suggesting that these gasdermins were lost in the monotreme lineage represented by Ornithorhynchus anatinus (monotreme mammal, platypus) in our tree.

Marsupials and placental mammals can encode all of GSDME, PJVK, GSDMA, GSDMB, GSDMC, and GSDMD. Mammals conserved the typical caspase-3 activation of GSDME (Wang et al., 2017; Rogers et al., 2017; Van Rossom et al., 2015). In mammals, GSDMA is cleaved by SpeB (Deng et al., 2022; LaRock et al., 2022). GSDMB is activated by granzyme A delivered by cytotoxic lymphocytes (Wang et al., 2023a; Zhong et al., 2023; Yin et al., 2023). GSDMC is activated by caspase-8 (Hou et al., 2020). Lastly, GSDMD is activated primarily by caspase-1 and related proteases (Kayagaki et al., 2015; Ding et al., 2016; He et al., 2015; Kambara et al., 2018; Burgener et al., 2019; Sarhan et al., 2018; Orning et al., 2018).

Numerous mammals have lost GSDMB and GSDMC. Previous studies reported this occurred in the monophyletic group that comprises mice, rats and rabbits. We identified the remnant GSDMB in rabbits which is interrupted by an ORM-1 like protein, as well as GSDMB in hamsters. This definitively shows the existence of GSDMB in this clade and suggests that there are multiple instances of its loss. This captures a snapshot of the rapid evolution of GSDM family proteins in mammals.

Overall, we discovered that the GSDMA common ancestor of amphibians, birds, and reptiles gave rise to mammal GSDMA-D. This is surprising given that GSDMA from both of these clades appear in the same locus (Figure 1—figure supplement 3). We speculated that this divergence in grouping reflected a divergence in function.

Unlike other trees (Yuan et al., 2022; De Schutter et al., 2021; Angosto-Bazarra et al., 2022; Wang et al., 2023b), we found that amphibian, bird, and reptile GSDMA are paralogs to mammal GSDMA, not orthologs, meaning we found that they appear in separate clades. Using the wealth of gasdermin sequences we assembled for our phylogenetic analysis, a remarkable trend in the bird GSDMA sequences became clear: nearly a third of all birds encoded a YVAD-like tetrapeptide in the linker region (Figure 2A, B). This sequence is famously cleaved by caspase-1 in mammals and is the peptide sequence utilized as an inhibitor and fluorometric substrate for measuring caspase-1 activity (Garcia-Calvo et al., 1998). The remaining bird GSDMA sequences contained a similar tetrapeptide also predicted to be cleaved by caspase-1, with a bulky P4 residue and aspartic acid at P1, most commonly FVSD or FASD (Figure 2B; Agard et al., 2010; Shen et al., 2010; Timmer and Salvesen, 2007; Thornberry et al., 1997). Amphibians, lizards, and snakes have similar tetrapeptides, albeit with more variation (Figure 2B).

An alignment of the linker regions of GSDMA proteins from selected birds, reptiles, and amphibians revealed conservation of these tetrapeptide sequences, despite differences in the length and overall lack of conservation within the linker (Figure 2C). These tetrapeptides appear similar to those commonly seen in the mammal GSDMD sequences, for example FLTD (human) and LLSD (mouse). Additionally, the P1′ residue in amphibian, bird, and reptile GSDMA was often a glycine, and the P2′ residue was often a proline, especially in birds with FASD/FVSD tetrapeptides (Figure 2B). A small P1′ residue is preferred by all caspases (Stennicke et al., 2000). By using a peptide library, glycine has been determined to be the optimal P1′ residue for caspase-1 and -4 (Talanian et al., 1997). Furthermore, in a review of the natural substrates of caspase-1, glycine was the second most common P1′ residue, and proline was the most common P2′ residue. These preferences were not observed for caspase-9 (Shen et al., 2010).

Amphibians, reptiles, and birds all express caspase-1 and have expanded their gasdermin family beyond GSDME to include GSDMA but notably lack GSDMD. We hypothesized that GSDMA may functionally mimic mammal GSDMD.

We focused our studies on birds, where the abundance of available sequences is high, and tissues are readily available. We first evaluated the expression of GSDMA and GSDME in Gallus gallus (chicken, which encodes an FASD tetrapeptide) because it is one of the best studied bird species and tissues are readily available for study. The gasdermin family proteins were so named because of their expression in the gastrointestinal tract (gas-) and the skin (-derm), with GSDMA being specifically expressed in the skin and upper gastrointestinal tract in mice (Tamura et al., 2007). If bird GSDMA functions similar to mammal GSDMD, then we would expect their expression profiles to be similar. Mammal GSDMD is primarily expressed broadly in immune cells as well as epithelial cells (De Schutter et al., 2021). Quantitative reverse transcription PCR (qRT-PCR) analysis from various chicken tissues revealed that chicken GSDMA was preferentially expressed in the Bursa of Fabricius (bird-specific primary lymphoid organ) and the spleen (Figure 3A). These data suggest that GSDMA is preferentially expressed in immune cells, similar to the expression of GSDMD in mammals. We additionally evaluated these same tissue sites for expression of CASP1 and GSDME and found that they were expressed similarly across most tissues (Figure 3—figure supplement 1). Interestingly, unlike mammal GSDMA, chicken GSDMA was not enriched in the skin. These expression data, together with the caspase-1 cleavage site in the linker region suggest that bird GSDMA may serve as their GSDMD.

In the absence of GSDMD, we hypothesized that caspase-1 may recognize and cleave the linker region of GSDMA in amphibians, reptiles, and birds, the gasdermin that gave rise to GSDMD. We next tested whether caspase-1 can cleave bird GSDMA. Caspase-1 is activated upon homodimerization through its CARD domain (Thornberry et al., 1992). To simulate this in vitro, we replaced the chicken caspase-1 CARD domain with 2xDmrB domains that can be homodimerized by the addition of the small molecule AP20187 (Figure 3B; Burnett et al., 2004; Freeman et al., 2003). We transfected 293T/17 cells with chicken 2xDmrB-caspase-1 and 3xFLAG-GSDMA (contained an FASD tetrapeptide). When we treated these cells with the AP20187 dimerizer, we observed strong intensification of the N-terminal GSDMA band (Figure 3C) that increased in a time-dependent manner (Figure 3D). The weak cleavage observed in the absence of dimerizer likely resulted from overexpression-dependent autoactivation. Importantly, mutating the predicted cleavage site residue of chicken GSDMA, D244A, prevented caspase-1-dependent cleavage (Figure 3E), confirming that chicken GSDMA is cleaved at 240FASD244 by caspase-1.

We then observed that crow GSDMA (containing a YVAD tetrapeptide) was similarly cleaved by chicken caspase-1 (Figure 3F). This cleavage was abolished when the crow YVAD is mutated to YVAA. Thus, this cleavage appears to be conserved across bird species. Interestingly, we observed two N-terminal GSDMA fragments, both of which were ablated by the YVAA mutation (Figure 3F). This is similar to how cnidarian and mollusk GSDM release two N-terminal fragments when cleaved by caspase-3. Interestingly, in those species, both fragments are competent to form pores (Jiang et al., 2020; Qin et al., 2023). Alternatively, the second cleavage may be an inactivation site (Xu et al., 2022; Taabazuing et al., 2017).

We next tested the ability of human and mouse caspase-1 to cleave chicken GSDMA, and observed that both of these mammal caspase-1 proteins cleaved chicken GSDMA (Figure 3F). This cleavage was again abolished by mutation of the aspartate cleavage site (Figure 3F). This cleavage was also seen with purified human caspase-1 (Figure 3—figure supplement 2). Human and mouse caspase-1 also cleaved crow GSDMA, again dependent upon the YVAD tetrapeptide (Figure 3E). Chicken caspase-1 is also able to cleave both human and mouse GSDMD at the expected site (Figure 3G, H). Thus, caspase-1 activity well is conserved in diverse species. None of the tested caspase-1 proteases could cleave human GSDMA (Figure 3I).

We hypothesized that activating pyroptosis using separate gasdermins for caspase-1 and -3 is a useful adaptation and allows for fine-tuning of these separate pathways. In mammals, this separation depends on the activation of GSDMD by caspase-1 and the activation of GSDME by caspase-3. Birds, too, encode caspase-1 and -3. In species with as distant common ancestors as coral, GDSME is cleaved by caspase-3 (Qin et al., 2023; Chen et al., 2023; Jiang et al., 2020). In mammals, caspase-3 inactivates GSDMD by cleaving within the N-terminal pore-forming domain to prevent pore formation during apoptosis. Bird GSDMA proteins appear to have a caspase-3 tetrapeptide sequence at a similar location in the N-terminus. Indeed, when we activated human caspase-3 via caspase-9 (Straathof et al., 2005) we observed the inactivating cleavage event (Figure 3J, Figure 3—figure supplement 2), demonstrating this crosstalk exists in species that use a gasdermin besides GSDMD for caspase-1-driven pyroptosis. Also conserved in chickens is the activation of GSDME via caspase-3, which was abolished by mutating the P1 aspartic acid (Figure 3J, Figure 3—figure supplement 2).

To confirm that the N-terminal fragment of chicken GSDMA forms a pore and causes pyroptosis, we transfected 293T/17 cells with a plasmid encoding the N-terminal domain of chicken GSDMA or the full-length protein. The N-terminal domain resulted in 85% reduction in host cell viability (Figure 4A). To directly demonstrate that caspase-1 cleavage of GSDMA results in cell death, we co-expressed these proteins in 293T/17 cells. Dimerization with AP20187 caused cell death that was abrogated when the FASD cleavage site was mutated to FASA (Figure 4B). HeLa cells displayed pyroptotic morphology with ballooning membrane formations when transfected with the chicken GSDMA N-terminal domain (Figure 4C). In these experiments, 293T/17 cells were used for their high transfection efficiency, and HeLa cells were used in microscopy studies for their larger size and improved adherence. Overall, our data suggest that the activation of chicken caspase-1 can lead to the cleavage of GSDMA to trigger pyroptosis.

Surprisingly counter the basic dogma of caspase enzymatic biology, it was recently discovered that caspase-1 cleaves GSDMD after the P1 aspartic acid independent of the tetrapeptide sequence (Wang et al., 2020b). The specificity is instead mediated by extensive interactions between an exosite in GSDMD that interfaces with caspase-1. Thus, the human GSDMD tetrapeptide FTLD can be mutated to AAAD without compromising caspase-1 cleavage (Wang et al., 2020b; Liu et al., 2020).

This interaction uses hydrophobic resides found in the region between alpha helix α7′ and α8 of the GSDMD C-terminus (Liu et al., 2019). We found this region is truncated in the AlphaFold prediction of chicken GSDMA (Figure 5A, B; Table 1; Jumper et al., 2021), preventing the interhelix loop from providing a groove for caspase-1. Furthermore, this region of chicken GSDMA does not contain hydrophobic residues (Figure 5C). This suggests that GSDMD inserted hydrophobic residues between the these α helices to better engage caspase-1. We hypothesized that chicken GSDMA instead fully relied on their tetrapeptide sequence for their activation by caspase-1.

Consistent with previous studies, human caspase-1 cleaves human GSDMD in a tetrapeptide-independent manner (Figure 5D, E). In contrast, activation of chicken GSDMA by chicken caspase-1 was lost when the FASD tetrapeptide was mutated to AAAD (Figure 5D, E). Because nearly a third of birds utilize a YVAD-like tetrapeptide in GSDMA, we modified the native chicken FASD to YVAD to determine if the precise sequence of the tetrapeptide affects cleavage efficiency. Indeed, YVAD was cleaved more efficiently than FASD (Figure 5D, E). We also mutated the human GSDMD tetrapeptide to YVAD but only observed slightly darker cleavage bands (Figure 5D, E). Therefore, the tetrapeptide sequence has a greater effect on cleavage efficiency for bird GSDMA compared to mammal GSDMD. These observations suggest that caspase-1 originally evolved alongside a gasdermin where cleavage was determined primarily by the tetrapeptide sequence. Because the YVAD-like tetrapeptides appear in a monophyletic group (Figure 2B), these data suggest that it was advantageous for some birds to use this sequence for GSDMA.

Together, these data suggest that GSDMA in birds does not use exosite-mediated interactions with caspase-1, and that these exosites developed after the gene duplication that gave rise to GSDMD, resulting in bird GSDMA fully relying on the tetrapeptide for cleavage.

Birds share a close common ancestor with reptiles. To determine if bird GSDMA is representative of other non-mammal GSDMA, we evaluated the ability of GSDMA from Alligator mississippiensis (crocodilian reptile, American alligator) to be cleaved by the same caspase-1 proteins studied above. This reptile was selected as crocodilians share a more recent common ancestor with birds than with lizards and snakes (Figure 2B). Chicken, human, and mouse caspase-1 proteases could all cleave A. mississippiensis GSDMA (Figure 6A), which has a caspase-1 like FAHD tetrapeptide. The GSDMA tetrapeptides from more distantly related reptiles have more diverse tetrapeptide sequences. For example, GSDMA from Anolis carolinensis (reptile, Carolina anole) has a VVSD tetrapeptide. Even so, this GSDMA was also cleaved by all of these caspase-1 proteases (Figure 6B).

To more directly assess the origin of GSDMA, we looked to the clade that has the most ancient GSDMA, the amphibians (Yuan et al., 2022; Wang et al., 2023b). Although many amphibians have lost GSDMA, the caecilian order, which separate from frogs and salamanders early in the speciation of amphibians (Upham et al., 2019), including Microcaecilia unicolor (amphibian, tiny cayenne caecilian) retains it. All three tested caspase-1 proteins were all able to cleave this amphibian GSDMA (Figure 6C).

In further support of our hypothesis that GSDMA in all tetrapods except mammals serves as caspase-1 target gasdermin, in amphibians GSDMA is not enriched in the skin, but rather in the kidney (Figure 6—figure supplement 1; Torres-Sánchez et al., 2019), an organ that caecilians use for hematopoiesis (Bain and Harr, 2022), and also a site where GSDMD is enriched in mammals (Uhlén et al., 2015).

O. anatinus (monotreme mammal, platypus) GSDMD groups at the root node of mammal GSDMA proteins (Figure 1). The linker region of this gasdermin contains an FVKD tetrapeptide, but whether this divergent GSDMD was a target of caspase-1 had not been investigated. Again, all the caspase-1 proteases tested were able to cleave platypus GSDMD (Figure 6D), but not platypus GSDMA (Figure 6E). These data suggest that the ancestral GSDMA that gave rise to mammal GSDMA was indeed cleaved by caspase-1, and that this function was shifted into mammal GSDMD early in the evolution of mammals. In other words, caspase-1 cleaved GSDMA until GSDMD arose, allowing caspase-1 to free this ancient GSDMA to acquire a new function.

Human and mouse GSDMA are cleaved by a bacterial cysteine protease called SpeB that is secreted by group A Streptococci during invasive skin infection. SpeB cleaves GSDMA and triggers pyroptosis in keratinocytes, where GSDMA is specifically expressed (Deng et al., 2022; LaRock et al., 2022; Tamura et al., 2007). Previously, this was the only known activator of any GSDMA. To determine if this cleavage is conserved in chicken GSDMA, we incubated lysates containing GSDMA with active SpeB and observed an N-terminal fragment that was consistent with cleavage in the GSDMA linker (Figure 6F). In humans, SpeB cleaves GSDMA in two nearby sites the linker region after VIL/IQ/ASD (LaRock et al., 2022; Deng et al., 2022). This protease preferentially cleaves peptides with a hydrophobic residue at P2 with negatively charged amino acids from P3′ to P6′. SpeB demonstrates a preference for aromatic residues at P2 for some substrates (Blöchl et al., 2021), and indeed chicken GSDMA has a phenylalanine in the tetrapeptide.

We found that SpeB cleavage of chicken GSDMA was abrogated when using a AAAD tetrapeptide mutant (Figure 6F). Conversely, the negative charge from the P1 aspartic acid was dispensable (Figure 6F). These data reveal that the caspase-1 tetrapeptide of GSDMA in birds allows for cleavage by both caspase-1 and SpeB. We observed similar SpeB cleavage of GSDMA from M. unicolor and A. carolinensis, but not A. mississippiensis (Figure 6G). These data suggest that the ancestral GSDMA possessed the capacity to detect SpeB-like proteases within the cytosol of cells.

Putative gasdermins proteins were obtained using three complementary approaches: keyword search on the NCBI website, protein similarity searches in NCBI nr (https://www.ncbi.nlm.nih.gov/; last accessed 2021), and UniProt (Bateman et al., 2021) (last accessed 2021) using human gasdermins sequences as seed using BLASTp (Altschul et al., 1997) (e-value of 0.01 as cut-off). Retrieved sequenced were combined with gasdermins sequences used by De Schutter et al., 2021. We partitioned sequences according to their type (e.g., A, B, C, and D). Redundant proteins were removed by clustering using CD-HIT (Li and Godzik, 2006). We keep only one sequence per sequence (1-to-1 orthologs) using a reciprocal best BLASTp approach (e-value <10⁻⁵ as threshold). For each set, multiple protein alignments were generated using Clustal Omega (Sievers et al., 2011). Partial sequences were removed using trimAl (Capella-Gutiérrez et al., 2009) (criteria <50% of alignment length, >50% Gap) and/or visual inspection in Jalview (Waterhouse et al., 2009). Curated alignments were used to produce for Hidden Markov models and consensus sequences using HMMER (http://hmmer.org/), and protein profiles for AUGUSTUS-PPX (Keller et al., 2011). Abbreviations for animal genera and species are listed in Table 2.

Genome sequences for selected species were downloaded from NCBI. Corresponding raw reads (Illumina fastq files) were downloaded from NCBI SRA when available. For species with annotated protein genes, we searched all gasdermins using BLASTp and if absent we used TBLASTn with an e-value cut-off of 0.1 against the genome (Altschul et al., 1997). If no hit found, a local de novo assembly from raw reads using Spades (Bankevich et al., 2012) was attempted. This has had a limited success because many genomes were constructed using older versions of Illumina reads (50 nucleotides).

The purpose of the initial screening was to flag genomic regions with putative gasdermins for further annotation. All genomes were initially screened with exonerate (Slater and Birney, 2005) using consensus sequences as seed with relaxed parameters (see GitHub, copy archived at Cisse, 2023). Protein to genome alignments were filtered for spurious alignments and converted in transcripts using gffread utility (Pertea and Pertea, 2020), and protein using Funannotate utility ‘gff2prot’ utility (Palmer and Stajich, 2022).

Once validated, all gasdermin putative loci were annotated using AUGUSTUS-PPX (Keller et al., 2011) and/or BRAKER (Brůna et al., 2021). For species with RNA-seq data were available, RNA-seq reads were mapped to the corresponding genomes using hisat2 (Kim et al., 2019). Alignments were converted in BAM formats using SAMTOOLS (Li et al., 2009) to generate BAM alignments for AUGUSTUS prediction (BAM2hints, Stanke et al., 2006).

Predicted gasdermins were aligned using MAFFT (Katoh et al., 2002). The resulting alignment was inspected for inconsistencies in Jalview (Waterhouse et al., 2009). Maximum likelihood phylogenies were inferred using FastTree (Price et al., 2009) and raxml-ng (Kozlov et al., 2019). Time-calibrated trees were constructed using BEAST (Bouckaert et al., 2014) with log-normal distribution and strict clock using soft calibrations from mammal fossil estimated ages. Species age calibrations from https://vertlife.org/ (Upham et al., 2019).

Chicken GSDMA, GSDME, and CASP1 were isolated from cDNA from chicken tissue. All animals used for this purpose were treated in accordance to Chicken caspase-1 was previously reported to express a short 283 amino acid protein that is truncated when compared to human caspase-1 (Johnson et al., 1998). We cloned a full-length, 393 amino acid isoform from chicken spleens based on more recent NCBI sequences. CARD domains were removed as determined by NCBI annotations. Chicken, crow, mouse, and human GSDM were cloned into pcDNA3 with N-3xFLAG tag. Site-based mutagenesis was performed using in vivo assembly (Watson and García-Nafría, 2019; García-Nafría et al., 2016).

Caecilian, alligator, anole, and platypus gasdermins were synthesized by TwistBioscience (South San Francisco, CA, USA) and cloned using Gateway technology (Thermo Fisher, Waltham, MA, USA) into a pCMV-polysite (Agrotis et al., 2019) plasmid modified to include N-terminal 3xFLAG tag using the primers listed.

The tissue samples were stored in RNAlater RNA Stabilization Solution (AM7020, Thermo Fisher, Waltham, MA, USA) at −80°C if the experiments were not processed immediately. Approximately 20 µg tissue samples were harvested for RNA extraction with TRIzol reagent (15596026, Thermo Fisher, Waltham, MA, USA), and 1 µg total RNA was used for reverse transcription through SuperScript IV Reverse Transcriptase (18090200, Thermo Fisher, Waltham, MA, USA). The cDNA samples were diluted 80 times with molecular biology grade water for the PCR experiments. The genes were tested by normal SYBR green method with PowerTrack SYBR Green Master Mix (A46110, Thermo Fisher, Waltham, MA, USA). qRT-PCR results were analyzed by ΔΔCT method. Primers used are listed in Table 3.

HEK293T/17 (CRL-11268, ATCC, VA, USA) and HeLa (CCL-2, ATCC, VA, USA) cell lines were maintained in Dulbecco’s modified Eagle medium (DMEM; 11995-073, Thermo Fisher, Waltham, MA, USA) supplemented with 1% minimal essential medium non-essential amino acids solution (11140-050, Thermo Fisher, Waltham, MA, USA) and 10% fetal bovine serum (SH30396.03, Cytiva, Utah, USA). Cell lines were purchased from the Duke Cell Culture facility where they underwent STR authentication. Cell lines were tested regularly for mycoplasma infection by PCR (Dreolini et al., 2020). Cell lines were kept under 25 passages after thawing.

HEK293T/17 cells were transfected using Lipofectamine 3000 (L3000008, Thermo Fisher, Waltham, MA, USA) according to the user manual. All cells received the same amount of DNA with backbone pcDNA3 making up differences in mass. AP20187 (HY-13992, MedChemExpress, Monmouth Junction, NJ, USA) was added to cells by replacing supernatant with OptiMEM (31985-070, Thermo Fisher, Waltham, MA, USA) with 500 nM AP20187 or DMSO for western blot endpoints, or was added to the supernatant for a final concentration of 500 nM in complete DMEM for LDH endpoints. Plasmids used are listed in Table 4.

HEK293T/17 cells were lysed directly in 1xLSB, boiled for 5 min then ran on SDS–PAGE using Bio-Rad TGX gels (Hercules, CA, USA). Total protein was captured after activation with UV for 5 min on Azure Biosystems Azure 500 imager (Dublin, CA, USA). Protein was transferred to membrane using a Bio-Rad Trans-Blot Turbo transfer system. Total protein images used in figures were captured on membranes by UV after transfer unless otherwise specified. For densitometric analysis of bands, ImageJ was used (Schindelin et al., 2012). Antibodies used in western blotting are listed in Table 5.

HEK293T/17 cells were transfected with EGFP and either the full length or N-terminal domain of chicken GSDMA without a 3xFLAG tag. After 48 hr, GFP fluorescence was captured on a BioTek Instruments Synergy H1 microplate reader (Winooski, VT, USA). After all experiments were captured, the GFP full-length intensity was normalized to 100%. N-terminal lysis data were transformed linearly by the same scale as their accompanying full-length data.

HEK293T/17 cells were transfected with chicken caspase-1 and either wild type or D244A mutant GSDMA without a 3xFLAG tag. After 48 hr, AP20187 dimerizer was added in cell culture media to a final concentration of 500 nM for 5 hr before taking the supernatants for LDH assay using CytoTox 96 Non-Radioactive Cytotoxicity Assay (G1780, Promega, Madison, WI). Cell death was measured on a scale with the media without cells set to 0% and cells lysed for 1 hr in 1% Triton X-100 set to 100%.

HeLa cells were plated on Lab-Tek II chambered coverglass (155360, Thermo Fisher, Waltham, MA, USA). Cells were transfected with either chicken N-terminal GSDMA domain or equivalent mass of pmCherry. After 18 hr, dyes were added to a final concentration of 0.5 ng/ml Hoechst (H3570, Thermo Fisher, Waltham, MA, USA) and 160 nM SYTOX Green (S7020, Thermo Fisher, Waltham, MA, USA) to each well. After 30-min incubation cells were imaged at ×20 and ×60 on a Keyence BZ-X810 microscope (Osaka, Japan).

C-terminal domains from the following proteins were visualized using the ColabFold notebook (Mirdita et al., 2022; Mitchell et al., 2019; Mirdita et al., 2017; Berman et al., 2003; van Kempen et al., 2024; Steinegger et al., 2019; Mirdita et al., 2019). C-terminal domains processed included all the protein after the predicted aspartic acid cleavage site in the linker region.

Pymol (Schrödinger, 2015) version 2.6.0a0 (26d795fe92) was compiled from source.

SpeB (20 Units, ACRO biosystems, SPB-S5115, Newark, DE, USA) was incubated with cell lysates of 293T transfected with FLAG-tagged gasdermin A for 30 min at 37°C. For transient transfection for 293T/17, followed manufacturer’s instructions of Lipofectamine 2000 (Thermo Fisher, 11668019, Waltham, MA, USA). Cell lysates were run on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and transferred to nitrocellulose membrane. Immunoblotted using M2 anti-FLAG antibody. For these experiments, total protein was detected by Ponceau S staining (Thermo Fisher, A40000279, Waltham, MA, USA).

R version 4.3.0 was used in these studies. Plots were generated using ggplot2 version 3.4.2.

All animal silhouettes are from PhyloPic and are free of known restrictions or are dedicated to the public domain under the following licenses.