We first determined the role of overall local neural activity in maintaining baseline arterial diameter. We refer to the diameter of the vessel during periods lacking locomotion (see Materials and methods) as the ‘basal’ diameter. We suppressed the activity of all neurons via intracranial infusions of muscimol (a GABA_A receptor agonist) and compared the changes in neural activity and arterial diameter dynamics to artificial cerebrospinal fluid (aCSF- a vehicle control) infusions in the same mouse. Infusions of muscimol substantially decrease (by 70–95%) neural activity ~1.5 mm from the cannula (Winder et al., 2017; Zhang et al., 2019), and so will suppress activity in all cortical layers, unlike topical administration of drugs which typically only affect the superficial layers (Ferezou et al., 2006). Control experiments have shown that cannula implantations do not alter hemodynamic responses (Winder et al., 2017). To visualize any change in vessel diameter across all the measured vessels we made a scatter plot of vessel diameters (Figure 1G–H), with the abscissa the diameter measured after the control (vehicle) infusion, and the ordinate the diameter after the drug infusion (muscimol in this case). If there is no effect of the drug, the points will be scattered along the unity line. If the drug causes constriction or dilation, the points will be respectively above or below the line. We found that after the infusion of muscimol, there was a substantial decrease in baseline and locomotion-evoked gamma-band power relative to within animal aCSF controls (Figure 1F, basal −75.2 ± 7.1%, p<3.0×10⁻⁴; locomotion, −77.0 ± 6.6%, p<2.0×10⁻⁴). We found that muscimol infusions led to a substantial decrease in the basal arteriole diameter (Figure 1G–H, −25.4 ± 2.3%, LME p<2.8×10⁻⁴, n = 6 mice, 26 arterioles; Figure 1—figure supplement 2B). The effects of muscimol infusions were not due to direct actions on the cerebral vasculature, as multiple large-scale single-cell sequencing studies have shown that mice lack GABA receptors on endothelial cells (Sabbagh et al., 2018; Vanlandewijck et al., 2018; Zeisel et al., 2018) and physiological experiments using optogenetic stimulation of interneurons with and without GABA-receptor blockade have shown there is no impact of blocking GABAergic signaling on vessel dilation (Anenberg et al., 2015; Vazquez et al., 2018). Rather than being driven by direct action of muscimol on the vasculature, this constriction is likely due to the removal of tonic vasodilatory signals released by neurons.

To determine if this decrease in basal arteriole diameter drove a decrease in blood flow, we then measured red blood cell (RBC) velocity in the capillaries (Kleinfeld et al., 1998; Chaigneau et al., 2003; Drew et al., 2010a). Capillary RBC velocity was markedly decreased in muscimol-infused mice relative to aCSF infusions (Figure 1I and J, −34.5 ± 27%, LME p<1.5×10⁻³, n = 8 mice, 22 capillaries). We observed no significant change in the diameter of the capillaries following the infusion of muscimol (aCSF-muscimol difference: 0.03 ± 0.85 µm, LME p<0.94, Figure 1—figure supplement 3A and B). Suppressing neural activity with muscimol did not abolish spontaneous fluctuations in arteriole diameter (Figure 1—figure supplements 4A, -14.2±67%, LME p<0.06, n = 6 mice, 26 vessels), consistent with previous reports of vessel-autonomous oscillations in arterioles (Winder et al., 2017). Taken together, these results suggest that baseline levels of neural activity have a tonic vasodilatory effect on cerebral arterioles (either directly, or through astrocytes), and decreasing neural activity causes a corresponding decrease in arteriole diameter and blood flow.

We then examined the role of neural activity in locomotion-evoked dilations. We calculated locomotion-triggered averages using only the arterioles histologically identified to be in the FL/HL region (using cytochrome oxidase staining [Adams et al., 2018]), as vessels outside this region are less responsive to locomotion (Gao et al., 2015). We also restricted this locomotion-triggered analysis to arterioles < 25 microns in diameter, as arterial reactivity decreases with increasing arteriole size (Lee et al., 2001; Drew et al., 2011; Gao et al., 2015), so larger vessels tend to be less responsive to changes in neural activity than smaller vessels. We found that the locomotion-evoked dilation of these arterioles was blocked by local silencing of neurons with muscimol (Figure 1E locomotion-evoked dilation: aCSF 15.8 ± 2.7%; muscimol −21.6 ± 1.9%, LME p<5.2×10⁻⁶, n = 6 mice, 12 arterioles in FL/HL representation), suggesting that these arteriole dilations were not due to non-specific cardiovascular responses (Huo et al., 2015a) or modulatory input from other brain regions directly onto the vessels (Drew et al., 2008), but rather controlled by local activity. Similar effects were observed in penetrating arterioles (Figure 1—figure supplement 2B and C; see also Figure 6—figure supplement 2). Note that the locomotion-evoked numbers reported in the text (unless otherwise stated), are relative to the control baseline, and thus include both the effects of the manipulation on the basal diameter and the evoked diameter. We split out the effects of our manipulations on the locomotion-evoked response in the analysis in Figure 6—figure supplement 2. For comparison of the results obtained with these two approaches, during muscimol infusions, there was a small locomotion-evoked increase from the pre-locomotion baseline (15.37 ± 8.11% vs. 5.14±11.09%, LME p<0.001, n = 5 animals, 12 vessels) (Figure 6—figure supplement 2). Infusions of muscimol and other manipulations of neural activity did not cause significant differences in the amount of locomotion relative to controls, indicating that behavior is not an origin of this difference (Figure 1—figure supplement 5), and consistent with previous work showing that pharmacological silencing of motor cortex disrupted complex movements, but not locomotion (Beloozerova and Sirota, 1993). Thus, both basal diameter and sensory-evoked arterial dilations in the FL/HL region of somatosensory cortex are under control of local neural activity.

We then asked if we could bi-directionally manipulate arterial diameters using chemogenetic techniques to drive changes in overall neural activity. We expressed either hM4D-G(i) or hM3D-G(q)-coupled DREADDs (Urban and Roth, 2015) pan-neuronally under a human synapsin promoter (hSyn) in C57BL/6J mice. G(i)- and G(q)-coupled DREADDs will respectively decrease and increase the excitability of the cells expressing them after injection of the DREADDs agonist CNO. As a control, mice infected with an AAV encoding only the fluorescent reporter mCherry did not show significant changes in neural activity, or in basal or evoked arterial diameter after the injection of CNO (Figure 2—figure supplement 1B–D), showing CNO has minimal off-target effects on neural activity and arterial diameters.

We imaged arterioles in the DREADD-expressing region of cortex (identified by mCherry expression seen in vivo, Figure 2A, and histologically, Figure 2B). Injections of CNO in mice with pan-neuronal expression of hM4D-G(i) DREADDs resulted in a non-significant decrease in gamma-band power relative to vehicle injections (Figure 2C, basal −1.9 ± 27.2%, p<0.09; locomotion, +3.6 ± 10.3%, p<0.07), likely because of the parallel changes in excitability of inhibitory and excitatory cells cancelled out. Still, activation of hM4D-G(i) DREADDs resulted in significant decreases in basal arterial diameter (Figure 2D, −7.1 ± 7.0%, LME p<0.02 n = 7 mice, 45 vessels), which was particularly noticeable in the largest arterioles. We note that the hM4D-G(i) DREADDs did not uniformly lower neural activity, as we observed bursting activity in our electrophysiological recordings every few seconds (Figure 2—figure supplement 2A), and corresponding arterial dilations at similar frequencies (Figure 2—figure supplement 2B). While we saw clear constrictions in arteriole diameters when activating hSyn G(i)-DREADDs relative to baseline, there were also large spontaneous dilations, which on average cancelled out. The net effect of hM4D-G(i) DREADDs on the locomotion-evoked response was minimal (Figure 2E, locomotion-evoked response: vehicle 20.5 ± 2.0%, CNO 18.3 ± 4.1%, LME p<1, n = 7 mice, 25 arterioles in FL/HL representation).

When we pan-neuronally expressed hM3D-G(q) DREADDs, which will increase excitability and intracellular calcium levels (Alexander et al., 2009; Pati et al., 2019), activation of the DREADDs with CNO injections caused increases in gamma-band power relative to vehicle (Figure 2F, basal 18.6 ± 11.2%, p<0.02; locomotion, 43.6 ± 15.0%, p<7.1×10⁻³). Injections of CNO caused substantial increases in the basal arterial diameter over controls (Figure 2G, +18.8 ± 5.6%, LME p<0.03, n = 7 mice, 37 vessels). The increase in basal arteriole diameter by hM3D-G(q) DREADD activation was comparable to the locomotion-evoked response (Figure 2H, locomotion-evoked response: vehicle 18.6 ± 1.7%, CNO 26.0 ± 4.2%, LME p<1, n = 7 mice, 23 arterioles in FL/HL representation). These results suggest that neurons, or a subset of neurons, tonically release vasodilator(s) due to ongoing background neural activity, and elevation of this activity can drive arterial dilations.

The importance of local neural activity on the regulation of arterial dilation raises the question of which sub-population(s) of neurons are involved. We manipulated the excitability of pyramidal neurons by injecting AAVs encoding hM3D(q) or hM4D(i) DREADDs under a CaMKIIa promoter, which will restrict the expression to pyramidal neurons (López et al., 2016; Yuan and Grutzendler, 2016). CaMKIIa-DREADDs labeled cells colocalized with CaMKIIa expression and were primarily localized to neurons with cell bodies in layer 5 (Figure 3A). Previous work has shown that optogenetic stimulation of layer 5 pyramidal neurons drives increases in neural activity in all layers via recurrent connections, and these increases in neural activity are very similar to what is seen with sensory stimulation (Vazquez et al., 2014; Vazquez et al., 2018), so chemogenetic stimulation of these cells will likely produce neural activity patters similar to those normally found in the cortex. The dilatory signal in the deeper cortical layers is electrically conducted through the vasculature rapidly by gap junctions (Bagher and Segal, 2011; Welsh et al., 2018) to the pial vessels (Rungta et al., 2018). Activating hM4D(i) DREADDs in excitatory neurons with CNO caused large decreases in basal gamma-band power (Figure 3B and D, basal −36.0 ± 16.4%, p<0.01; locomotion, −26.8 ± 15.8%, p<0.02). When hM3D(q) DREADD receptors in excitatory neurons were activated, a corresponding large increase in basal gamma-band power was observed (Figure 3C and G, basal +52.4 ± 18.9%, p<5.0×10⁻³; locomotion, +50.9 ± 31.6%, p<0.02). Despite the large changes in neural activity, we did not see significant changes in the basal arterial diameters with manipulation of pyramidal neuron activity (−2.2 ± 5.7%, LME p<1, n = 7 mice, 38 arterioles for inhibition and +7.2 ± 8.8%, LME p<0.66, n = 6 mice, 41 arterioles for excitation) (Figure 3E and Figure 3H, respectively). The locomotion-evoked dilations were also not affected by manipulations of neural excitability (Figure 3F G(i) DREADDS: vehicle 25.3 ± 1.9%, CNO 21.6 ± 4.3%,LME p<1, n = 7 mice, 29 arterioles in FL/HL representation; Figure 3I G(q) DREADDS: vehicle 19.1 ± 2.3%, CNO 17.3 ± 4.7% LME p<1, n = 6 mice, 13 arterioles in FL/HL representation).

The lack of change in the arterial diameters could not be due to a lack of an effect of the chemogenetic manipulations as the changes in neural activity were very large. The decreases in basal neural activity generated by activation of G(i)-coupled DREADDs in pyramidal neurons were nearly as large as those elicited by muscimol (Figure 3B), and the increases in basal neural activity generated by activation of G(q)-coupled DREADDs in pyramidal neurons were as large as those seen during locomotion (Figure 3C), yet in neither of these cases did we observe arterial diameter changes of similar magnitude. Thus, pyramidal neuron activity is a strong driver of gamma band LFP power, but only has a moderate effect on basal arterial diameter, and little effect on the locomotion-evoked dilations.

Previous work has implicated neuronally-generated nitric oxide (NO) (Hosford and Gourine, 2019), generated by interneurons (Krawchuk et al., 2020; Lee et al., 2020) in the dilation of cerebral arteries. To test the role of neuronal nitric oxide synthase (nNOS) expressing neurons, we expressed hM3D(q) and hM4D(i) DREADDs along with the reporter protein mCherry in nNOS-expressing neurons using flexed AAVs in nNOS-cre- mice (B6.129-Nos1^tm1(cre)Mgmj/J, Jackson Laboratory #017526) (Leshan et al., 2012; Augustine et al., 2018). Expression of the reporter protein mCherry was primarily in the deeper and more superficial layers of the cortex (Figure 4A,B), consistent with the observed densities of nNOS-expressing neurons (Perrenoud et al., 2012). Co-labeling with antibodies for nNOS (Figure 4A,B) showed co-expression (white) of mCherry (magenta) and nNOS (green) in both cell bodies and processes of neurons. Neurons showing strong anti-nNOS labeling were taken to be type 1 nNOS neurons, with-type 2 nNOS neurons showing less intense anti-nNOS labeling, consistent with previous anatomical studies (Perrenoud et al., 2012). Immunolabeling (n = 2 mice) revealed that 28.6% (2/7 cells) type 1% and 35.2% (19/54) of type 2 nNOS neurons expressed DREADDs, and that 44.6% (21/47) of DREADD-expressing cells were nNOS positive. Neurons expressing mCherry also expressed GAD65, indicating they were GABAergic interneurons (Figure 4C).

CNO injections in mice expressing G(i)-DREADD in nNOS-expressing neurons resulted in no significant change in gamma-band power (Figure 4D, basal −1.2 ± 11.6%, p<1; locomotion, −5.7 ± 2.2%, p<0.09). Increasing the excitability of nNOS-expressing neurons resulted in a significant decrease in basal gamma-band power (Figure 4G, basal −19.8 ± 2.6%, p<4.0×10⁻³; locomotion, −12.8 ± 4.7%, p<0.10), consistent with these neurons being inhibitory. However, decreasing the excitability of nNOS neurons drove substantial and significant decreases in both basal arteriole diameter (−10.3 ± 6.4% LME p<1.3×10⁻⁴, n = 9 mice, 49 vessels Figure 4E) and the locomotion-induced dilation (Figure 4F, vehicle 15.3 ± 2.0%, CNO 0.3 ± 4.3%, LME p<6.3×10⁻³, n = 6 mice, 22 arterioles in FL/HL representation). Elevating the excitability of nNOS-expressing neurons significantly increased the basal arterial diameter (+5.8 ± 6.6%, LME p<2.7×10⁻⁴, n = 8 mice, 52 vessels, Figure 4H), but did not increase the locomotion-evoked increase above the control (Figure 4I vehicle 17.4 ± 1.8%, CNO 19.6 ± 3.8%, LME p<1, n = 8 mice, 25 arterioles in FL/HL representation). Interestingly, the diameter changes moved in the opposite direction as neural activity when the excitability of nNOS-expressing neurons was increased, consistent with nNOS neurons being inhibitory.

We then tested whether the effects of nNOS-expressing neuron activity on arteriole diameter was mediated by NO, as these neurons can also release vasoactive peptides (Cauli et al., 2004). Previous studies with NOS inhibitors have shown conflicting effects (Lindauer et al., 1999; Stefanovic et al., 2007). Discrepancy in the literature could be due to differences in experimental methodology: topical application of drugs may not reach lower cortical layers (Greenberg et al., 1997; Ferezou et al., 2006) where most of the highly expressing nNOS neurons reside, and systemic administration impacts the cardiovascular system. To avoid these experimental confounds, we infused the water-soluble NO synthase inhibitor, Nω-nitro-L-arginine methyl ester (L-NAME) through an implanted cannula to decrease NO-mediated signaling. In our hands, pharmacological infusions through cannula affect all cortical layers within ~1.5 mm of the infusion site (Winder et al., 2017). Note that recent studies, using both single-cell genomic methods as well as immunohistochemistry, have shown that some cortical neurons express eNOS (Yousef et al., 2004; Lein et al., 2007; Tasic et al., 2016), and that some endothelial cells of cerebral arterioles express nNOS (Vanlandewijck et al., 2018; Zeisel et al., 2018), so the endothelial/neural specificity of these enzymes is not complete. Additionally, crystal structures and careful pharmacology experiments have revealed that ‘selective’ pharmacological inhibitors of nNOS are much less selective than was originally thought (Bland-Ward and Moore, 1995; Reiner and Zagvazdin, 1998; Engelhardt et al., 2006; Pigott et al., 2013; Poulos and Li, 2017), so it is likely that there are no inhibitors of nNOS or eNOS are specific enough to target one enzyme over another in vivo. Finally, locomotion-evoked dilations are unlikely to be mediated by eNOS expressed in endothelial cells as increases in eNOS activity in response to flow-evoked changes in diameter evolve over minutes, not seconds (Kim et al., 2016) (too slow to account for the locomotion-evoked dilations), and there are no shear-stress changes seen with sensory-evoked dilation (Ngai and Winn, 1996). Blocking the production of NO caused no significant change in the basal gamma-band power (Figure 4—figure supplement 1A, basal −5.0 ± 8.5%, p<0.55; locomotion, −7.5 ± 9.6%, p<0.43), but drove a decrease in basal and locomotion-evoked arterial diameter (−10.6 ± 4.6%, LME p<7.6×10⁻⁶, n = 6 mice, 36 arterioles, and vehicle 13.2 ± 2.4%, L-NAME 4.2 ± 3.8%, LME p<0.04, n = 6 mice, 20 arterioles in FL/HL representation, Figure 4—figure supplement 1B and C). The basal arterial diameter constrictions produced by L-NAME were not significantly different from those produced by G(i) DREADDs in nNOS neurons (−10.6% vs. −10.3%, LME p<0.57 Figure 4—figure supplement 1D). Although there is likely some contribution from the inhibition of eNOS contributing to these constrictions (Atochin and Huang, 2011), these results are consistent with the hypothesis that the NO produced by nNOS-expressing neurons contributes maintaining the diameter of cerebral arterioles.

As we saw expression of DREADDs primarily in type 2 nNOS neurons, we wondered how activation of type 1 nNOS neurons might affect arteriole dynamics. Type 1 nNOS neurons are unique among cortical neurons in that they express the receptor for Substance P (NK1R) (Vruwink et al., 2001; Tomioka et al., 2005; Tomioka and Rockland, 2007; Gerashchenko et al., 2008; Kilduff et al., 2011; Kubota et al., 2011; Dittrich et al., 2012; Dittrich et al., 2015; Endo et al., 2016; Williams et al., 2019; Williams et al., 2018), whose activation drives depolarization and spiking (Dittrich et al., 2012; Endo et al., 2016). As multiple single-cell genomics studies have shown there is no NK1R expressed in mouse cerebral vascular cells or in glial cells (Vanlandewijck et al., 2018; Zeisel et al., 2018), we can pharmacologically manipulate the activity of Type 1 nNOS neurons independently of other neurons by infusing SP, or an antagonist for the SP receptor, CP-99994 through a chronically implanted cannula. Infusions of CP-99994 caused non-significant changes in gamma-band power (Figure 5A, basal −5.4 ± 4.7%, p<0.28; locomotion, +7.3 ± 8.0%, p<0.37) and a non-significant decrease in basal arteriole diameter (−8.0 ± 7.2%, LME p<1, n = 7 mice, 28 vessels, Figure 5B) and locomotion-evoked dilation response (vehicle 13.7 ± 2.6%, CP-99994 9.7 ± 3.3%, LME p<1, n = 7 mice, 14 arterioles in FL/HL representation, Figure 5C). The lack of a substantial effect suggests that either basal activation of the Substance P receptors in the cortex is low, generating a floor effect and/or the basal activity of the type 1 nNOS neurons is low. When we activated the type 1 nNOS neurons via infusions of SP, we observed non-significant decreases in gamma-band power (Figure 5D, basal −20.0 ± 32.4%, p<0.53; locomotion, +5.2 ± 16.0%, p<0.73), but a substantial increase in basal diameter (+13.4 ± 4.4%, LME p<3.6×10⁻³, n = 7 mice, 30 arterioles, Figure 5E), although the effect was not significant for locomotion-evoked responses (vehicle 13.0 ± 2.2%, SP 24.8 ± 4.7%, LME p<0.75, n = 7 mice, 14 arterioles in FL/HL representation, Figure 5F). To test the hypothesis that the Substance P facilitated dilation was mediated by increases in neural activity, we co-infused SP with muscimol to block local neural activity, and observed a reduction of the arteriole diameter (−20.0 ± 14.9%, LME p<3.4×10⁻⁷, n = 5 mice, 27 arterioles, Figure 5G) and no locomotion-evoked dilation (vehicle 14.2 ± 3.0%, SP+muscimol −14.5 ± 3.7%, LME p<2.8×10⁻³, n = 5 mice, seven arterioles in FL/HL representation, Figure 5H), consistent with the action of SP being mediated through local neural activity. These results suggest that type 1 nNOS neurons play an important role in NO-mediated changes in arteriole diameter.

In addition to nitric oxide, there are multiple vasodilatory pathways that have been implicated in signaling the vasculature (Attwell et al., 2010; Otsu et al., 2015; Bazargani and Attwell, 2016; Drew, 2019). We first examined how chemogenetic modulation of astrocytes impacted basal and evoked arterial dilations (Figure 5—figure supplement 1). We found no significant modulation of arteriole diameter when either G(i)- or G(q)-coupled DREADDs on astrocytes were activated, consistent with previous reports that did not see coupling between astrocytes and evoked arterial responses (Nizar et al., 2013; Bonder and McCarthy, 2014; Tran et al., 2018). This does not rule out an astrocytic role for the control of the cerebral vasculature, as the signaling could work via mechanisms that are not coupled to G-protein signaling.

Recent reports have implicated inwardly-rectifying potassium channels in the capillary bed in mediating functional hyperemia (Longden et al., 2017). To test the role of these channels in basal and evoked arterial diameter control, we infused Ba²⁺ or ML-133 (Longden and Nelson, 2015), blockers of inwardly rectifying potassium channels. We found that both of these inwardly rectifying K⁺ channel blockers drove very large increases in neural activity (Ba²⁺, Figure 5—figure supplement 2: basal +88.3 ± 44.5%, p<0.01; locomotion, +145.6 ± 86.1%, p<0.02; and ML-133 S12E: basal +147.2 ± 77.7%, p<0.01; locomotion, +157.7 ± 86.4%, p<0.01). These observations of large increase in basal and evoked neural activity are consistent with previous work that has shown that barium increases the activity of neurons in the awake cortex (Harnett et al., 2013). Barium is known to block inwardly rectifying potassium channels found on neurons, such as KIR2.1 (Alagem et al., 2001), Kir2.2 (Amarillo et al., 2014), and Kir2.3 (Schram et al., 2003). Blocking these channels will increase neural excitability, so it is not surprising that barium infusions increase overall neural activity. Blocking inwardly rectifying K⁺ channels drove basal arteriole dilation (Figure 5—figure supplement 2: +12.4 ± 17.2%, LME p<2.0×10⁻³, n = 5 mice, 24 arterioles; S12F: +11.8 ± 17.0% LME p<3.7×10⁻⁴, n = 5 mice, 36 arterioles), and this basal dilation partially occluded the locomotion-induced response (Figure 5—figure supplement 2: vehicle 15.1 ± 2.7%, BaCl₂ +20.2 ± 5.3%, LME p<1, n = 5 mice, 16 arterioles in FL/HL representation and Figure 5—figure supplement 2: vehicle 14.01 ± 3.0%, ML-133 +21.5 ± 5.8%, LME p<1, n = 5 mice, 26 arterioles in FL/HL representation). As with chemogenetic increases in pan-neuronal activity (Figure 2), the increase in neural activity by barium drove a baseline dilation that nearly occluded the locomotion-induced response.

Because of the large increase in neural activity caused by blockade of inwardly rectifying potassium channels in the awake brain, and the likely release of vasodilator resulting from this increase in activity, it is hard to parse out any effects mediated by blockade of inward rectifying channels on the vasculature without transgenic models where the inwardly rectified channels are selectively removed from the vasculature (Longden et al., 2017). These results, along with the lack of effects of manipulation of G-protein-coupled pathways in astrocytes on arterial diameter highlight the need for the development of innovative manipulations of signaling pathways in astrocytes and endothelial cells as has been so successfully used for neurons.

It is informative to compare the responses of pial and penetrating arterioles to the manipulations performed here. We saw some differences between pial and penetrating arterioles, both in changes to their basal diameters, and in locomotion-evoked dilations. When we looked at basal diameter changes, we saw a significantly smaller constriction of penetrating arterioles than pial arterioles with muscimol infusions (Figure 6A). With pan-neuronal excitatory DREADDs, we saw a significantly smaller dilation in penetrating arterioles than in the pial arteries (Figure 6D). For all other manipulations, which in general produced smaller changes than muscimol and pan-neuronal excitatory DREADDs, there was no significant difference in the induced basal diameter changes between penetrating and pial arteries.

We then examined the percentage change in diameter during locomotion in pial and penetrating arterioles, here defined as the change from the respective pre-locomotion baseline normalized by the pre-locomotion baseline, for each of the manipulations. For this metric, a manipulation that causes a 10% constriction of baseline and shows a dilation of 5% above the control baseline will have a 15% locomotion-induced dilation (Figure 6—figure supplement 1). If the vessel has a 10% baseline dilation and during locomotion dilates to 15% above the control baseline, the percentage change will be 5%. This metric only captures the locomotion-induced dilation component and differs from the measures of dilation reported in the sections above, which included the effects of the baseline shift in diameter. When we looked at all the control condition (vehicle injections and aCSF infusions Figure 6—figure supplement 2), we saw a large and substantial difference in the pial arterial dilation versus the penetrating arteriole dilation (17.57 ± 9.35% pial vs. 6.79 ± 6.54% penetrating, LME p<8×10⁻³⁷). This difference is consistent with previous measurements of pial and penetrating arteriole dynamics during locomotion (Gao et al., 2015), so we treated penetrating and pial arteries as separate groups for further statistical analysis. We then compared the percentage diameter change during locomotion for pial arterioles and penetrating arterioles (Figure 6—figure supplement 2). With this metric, we saw significant differences in the locomotion-evoked responses of pial arterioles with muscimol, pan-neuronal excitation (hsyn-Gq), excitation of pyramidal neurons (CaMKIIa-Gq), inhibition of nNOS neurons (nNOS-Gi) and L-NAME. Interestingly, the expression of excitatory DREADDs in pyramidal neurons and pan-neuronally had the paradoxical effect of decreasing the amplitude of the locomotion-evoked dilation when quantified relative to its within treatment baseline. This result can be explained by the increased baseline neuronal activity increasing baseline arteriole diameters while not extending the maximum amplitude of arterial dilations. However, none of our manipulations produced a significant change in the locomotion-evoked diameter change of penetrating arterioles from the control condition (Figure 6—figure supplement 2).

Overall, the locomotion-evoked changes in penetrating arteriole diameters were smaller than those in the pial arteries, and the size of the penetrating arteriole dilations were not affected by our manipulations. The difference observed between penetrating and pial arterioles cannot be due to imaging artifacts, as sub-micron cellular structures like spines and microglial processes can be imaged at this depth with a thinned skull window (Drew et al., 2010b; Shih et al., 2012a), and our diameter measurement algorithm (see Materials and methods) has been validated to work robustly in much lower signal-to-noise regimes than present here (Gao and Drew, 2014). There are several non-exclusive possibilities for the differences between the responses of pial and penetrating arteries to behavioral and experimenter-evoked changes in neural activity. One possible reason for this difference could be that penetrating arterioles are less sensitive to vasodilators, with a less steep voltage-diameter relationship that pial arterioles. As there are differences in mural cell morphology at different levels of the arteriole tree (Hartmann et al., 2015), it is possible that there are some previously unobserved differences in the smooth muscle or endothelial cells between penetrating and pial arteries. When mice are anesthetized with the powerful vasodilator isoflurane, pial and penetrating arteries dilate to the same extent, suggesting that the vessels have the same maximal dilation (Gao et al., 2015). The diameter of arterioles is linearly related to the membrane potential of smooth muscle cells (Wölfle et al., 2011), and any electrical signals at one point are conducted via gap junctions up to several hundred microns away (Segal, 2005; Bagher and Segal, 2011; Hald et al., 2012). As the pial arteries may be further up the arterial tree than penetrating arteries, they will be also integrating signals from many more smaller vessels (Segal, 2000; Welsh et al., 2018), and this may account for the differences in the amplitudes of the responses of the two vessel types. Another possibility could be the spatial location of the generation of the vasodilator(s) relative to the arterioles, particularly in the case of nitric oxide. Simulation of NO diffusion and degradation dynamics in the tissue have shown that NO production needs to be very close (of order microns to tens of microns) to the arteriole in order to exert an appreciable dilatory effect (Haselden et al., 2020). It could be that NO is produced closer to pial vessel (potentially in the processes of more superficial type two neurons), than near the penetrating arterioles in the superficial layers. Finally, the different mechanical environments of the pial and penetrating arterioles may account for their differing responses. The pial arterioles are surrounded by CSF (Abbott et al., 2018), whereas the penetrating arterioles are surrounded by a thin paravascular space (Iliff et al., 2012) filled with collagen fibers tethering the artery to the brain (Roggendorf and Cervós-Navarro, 1977). When the smooth muscle cells surround pial arteries relax, the vessel’s dilation need only displace CSF, which has viscosity near that of water, while dilation (or constriction) of penetrating arterioles must displace (or pull upon) the viscoelastic brain tissue (Goriely et al., 2015; Weickenmeier et al., 2018a; Weickenmeier et al., 2018b; Kedarasetti et al., 2020). The viscoelasticity of brain tissue will function as an ultra-low-pass mechanical filter on diameter changes of penetrating arterioles, reducing the amplitude of ‘rapid’ dilations (on the order of seconds), although allowing slower dynamics (on the order of minutes), which could explain why we saw similar changes in both pial and penetrating arterioles baseline (which take place over minutes), but no effects on the ‘rapid’ locomotion-evoked dilations. Studies in peripheral vascular structures have shown that when blood vessels are imbedded in tissue, the mechanics of the surround tissues dominates and reduces their dilation (Fung et al., 1966; Liu et al., 2007). Brain tissue is stiffer when the skull is intact than after a craniotomy (Hatashita and Hoff, 1987). Brain tissue exhibits strain-stiffening (becoming stiffer under pressure) (Marmarou et al., 1975; Marmarou et al., 1978), and locomotion is accompanied by an elevation of intracranial pressure (Gao and Drew, 2016; Norwood et al., 2019; Eftekhari et al., 2020). All these factors may exacerbate mechanical restriction of dilation by stiffening the brain tissue, preventing full expression of the arterial dilation to locomotion in penetrating arterioles. These results show that pial and penetrating arterioles display similar basal diameter changes for manipulations that produced moderate changes in diameter, although the locomotion-evoked changes in diameter of penetrating arterioles were not affected by any of the manipulations of neural activity. Elucidating the vascular, neural, and/or mechanical origins of the differences in the responses of pial and penetrating arterioles will be informative for understanding their roles in regulating cerebral blood flow (Lorthois et al., 2011; Gould et al., 2017; Schmid et al., 2017).

If nNOS neurons contribute to the locomotion-induced dilation of arteries, then we would expect their activity to increase during locomotion. Electrophysiological recordings have shown that voluntary locomotion drives large increases in the activity of pyramidal neurons and interneurons in FL/HL representation (Figure 1—figure supplement 1; Zhang et al., 2019), but it is not known if nNOS-positive interneurons exhibit increased activity, as their activity cannot be separated out electrophysiologically. To determine how nNOS-positive interneuron activity is affected by locomotion, we used fiber photometry (Schulz et al., 2012) to simultaneously monitor the activity of nNOS neurons expressing GCaMP6s (Chen et al., 2013) and local cerebral blood volume deep in the parenchyma, which was labeled with an intravenous injection of TRITC dextran (Cerri et al., 2019). Calcium signals have been shown to be strong quantitative predictors of hemodynamic signals (Aydin et al., 2020). Fiber photometry will record bulk calcium signals from the processes and cell bodies of a specific neuronal cell type, and these calcium signals should correlate with nitric oxide production (whose synthesis is known to be calcium dependent) in nNOS-expressing neurons. Fiber photometry is not imaging forming, and will report the summed fluorescence of neurons within ~200 µm from the tip of the fiber (Pisanello et al., 2019), so the contributions of type 1 and type 2 nNOS neurons cannot be differentiated. However, given the low density of type 1 nNOS neurons, the signal is likely to be primarily from type 2 nNOS neurons. We corrected for the absorbance of fluorescence by hemoglobin (Gunaydin et al., 2014; Cerri et al., 2019) in the blood using the TRITC signal, and validated this method on mice that expressed GFP ubiquitously (Okabe et al., 1997; Figure 7—figure supplement 1). The slow (on the timescale of tens to hundreds of seconds) rise and fall of the CBV signals is due to the venous component. The venous contribution to the CBV signal is known to be larger than the arterial component (Kim et al., 2007; Huo et al., 2015b; Huo et al., 2015a) and have slower dynamics than the arterial component (Kim et al., 2007; Drew et al., 2011; Moon et al., 2013; Gao et al., 2015; Huo et al., 2015b; Huo et al., 2015a; Gao and Drew, 2016), particularly in the parenchyma (Kim and Kim, 2010). We found that calcium signals from GCaMP expressed pan-neuronally, in pyramidal neurons, and in nNOS-positive neurons greatly increased during locomotion (Figure 7A–C,E Population GCaMP intensity change in z-units: hSyn: 2.51 ± 0.15, LME difference from CaMKII: p<0.44; CaMKII: 2.59 ± 0.38; nNOS: 1.94 ± 0.37, LME difference from CaMKII: p<1e-3). Activity in all these neuron types was correlated with vasodilation (Figure 7—figure supplement 2, CaMKII: 0.60 ± 0.17; hSyn: 0.54 ± 0.14, LME difference from CaMKII: p<0.7; nNOS: 0.43 ± 0.11, LME difference from CaMKII: p<0.02). We also asked if the activity of nNOS neurons were sustained during locomotion, as there is a sustained dilation of arteries during locomotion (Huo et al., 2015b; Gao and Drew, 2016), and it has been suggested that the response of nNOS neurons is transient (Iadecola, 2017). We found that the responses of nNOS neurons were sustained throughout the locomotion bouts (Figure 7—figure supplement 3 nNOS population GCaMP intensity change in z-units, 10 s: 1.90 ± 0.46, LME: p<0.01, 15 s: 1.84 ± 0.52, LME: p<0.97, 30 s: 1.56 ± 0.40, LME: p<0.72).

When we compared the change in the TRITC fluorescence (a measure of blood volume change) evoked by locomotion in GCaMP-nNOS mice to those in mice expressing GCaMP in pyramidal neurons we found a significantly smaller responses than in mice with calcium indicators in pyramidal neurons (Figure 7D Population TRITC intensity change (z-units), hSyn: 0.42 ± 0.16, LME difference from CaMKII: p<0.08, CaMKII: 0.75 ± 0.38, nNOS: 0.46 ± 0.18, LME difference from CaMKII: p<0.04). Pan-neuronal expression of GCaMP also decreased the CBV change, though not significantly. We do not know the origin of the difference between hemodynamic responses of nNOS and CaMKII-GCaMP mice, but one plausible cause could be that smaller locomotion-evoked response might be caused by the expression of a calcium indicator in the nNOS neurons. Calcium indicators will buffer calcium levels (Higley and Sabatini, 2008; McMahon and Jackson, 2018), reducing the amplitude of activity-dependent calcium transients by competing with endogenous calcium-binding proteins. This calcium buffering could reduce the production of NO in cells expressing the indicator because the activity of nNOS is calcium-dependent (Garthwaite, 2008).

There are several caveats to these experiments. The fluorescence signals from nNOS neurons, whether measured as absolute fluorescence or in the absolute changes in fluorescence, was substantially smaller than the signals from pan-neuronally expressed GCaMP or GCaMP expressed pyramidal neuron localized GCaMP signals. This is a consequence of the much lower number of GCaMP-nNOS neurons than the with the pan-neuronal or pyramidal neuron specific promoters (Figure 7A–C). This large difference in the absolute signals for these different conditions means that it would be inappropriate to draw any strong conclusions in differences in the calcium signal amplitudes or in correlation coefficients between blood volumes and neural activity. Calcium indicators are qualitative indicators of neural activity, as simultaneous calcium imaging and electrophysiology has found that the correlation between calcium signals and spiking activity is moderate (Harris et al., 2016; Theis et al., 2016; Pachitariu et al., 2018), and there are likely cell-type specific differences in calcium levels, indicator expression, and coupling of calcium changes to spiking that makes it untenable to make quantitative comparisons across calcium measurements of activity across different cell types. Different neuronal subtypes are known to have differing calcium buffering (Lee et al., 2000; Egger and Stroh, 2009; Matthews et al., 2013) and nitric oxide synthase expressing neurons show low levels of endogenous calcium binding proteins (Dun et al., 1994). These results show that nNOS neurons, like pyramidal neurons and other interneurons (Zhang et al., 2019), increase their activity during locomotion.

One of the key utilities of hemodynamic imaging is that it allows us to make inferences about neural activity. As we have measured both neural activity and arterial diameter for multiple different perturbations, we sought to understand the relationship between them for different manipulations in pial arteries (Figure 8—figure supplement 1). We plotted the arterial diameter and gamma band power changes for each of the manipulations from the vehicle basal condition in Figure 8, with the basal and locomotion conditions connected by a line. A horizontal (left-right) shift means that neural activity is changed without a corresponding change in arterial diameter, while any vertical shift of the points means that there is change in arterial diameter without a corresponding change in neural activity. If there was a perfect, one-to-one ratio between gamma band power and arterial diameter, all of these points would lie along a line or a curve. If this is the case, by knowing the change in arterial diameter, one could infer the change in neural activity accurately from knowing the changes in arterial diameter. At the other extreme, if there were no relationship between neural activity and diameter, the points would be a shapeless blob. In this situation, it is not possible to infer neural activity from a change in arterial dilation.

Within all the manipulations, we found that the increase in neural activity accompanying locomotion was accompanied by an arterial dilation. However, across conditions there was only a weak relationship between neural activity and arterial diameter. For example, the gamma band power was similar for both CaMKII-G(i) mice during locomotion and stationary nNOS-G(i) mice, but there were very large (~30%) differences in the arterial diameter. As another example, the change in arterial diameter was the same for stationary CaMKII-G(i) mice as for nNOS-G(i) during locomotion, even though there was an approximately threefold difference in neural activity A linear regression between change in neural activity and change in diameter (excluding the outlier muscimol) reveals no significant relationship between the two (p<0.12). We also looked at the relationship between power in other bands in the LFP (1–10 Hz and 10–40 Hz) and there was either a very weak relationship (for 10–40 Hz, p<0.04, R² = 0.04) or no significant relationship (1–10 Hz, p<0.05) (Figure 8—figure supplement 2). The lack of a strong relationship between arteriole diameter and LFP, which reports the summed activity of all neurons, though with large pyramidal neurons giving a disproportionally large contribution (Lindén et al., 2011; Einevoll et al., 2013; Pesaran et al., 2018), implies that a small set of nNOS expressing neurons can influence arterial diameter without influencing the LFP and vice versa.

This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All procedures were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) of Pennsylvania State University (protocol # 201042827). All surgeries were performed under isoflurane anesthesia and every effort was made to minimize suffering. Data were acquired from 209 mice (107 mice for 2PLSM imaging, 61 mice for electrophysiology, 21 mice for fiber photometry, and 20 mice for immunohistochemistry - see Tables 1 and 2) C57BL/6J mice, cre-nNOS mice (B6.129-Nos1^tm1(cre)Mgmj/J), C57BL/6-Tg(CAG-EGFP), and Aldh1L1-cre-mice (B6;FVB-Tg(Aldh1l1-cre/ERT2)1Khakh/J Jackson Laboratory) between 3 and 8 months of age were used. Some mice received infusions of multiple drugs. Mice were given food and water ad libitum and maintained on 12 hr light/dark cycles in isolated cages during the period of experiments. All in vivo experiments were performed between 8:00 and 12:00 ZT. Sample size was chosen to be consistent with previous studies in the literature (Zhang et al., 2019; Aydin et al., 2020).

At the end of the experiment, mice were deeply anesthetized with 5% isoflurane, and then transcardially perfused with heparinized saline followed by 4% paraformaldehyde. The brain was removed from the skull and sunk in 30% sucrose/4% PFA. For cytochrome oxidase staining, the cortex was then flattened and 60 µm tangential sections were cut on a freezing microtome. The cytochrome oxidase in the sections was stained, and the location of the FL/HL representation in somatosensory cortex was reconstructed relative to the vasculature visible through the cranial window (Adams et al., 2018).

For immunohistochemistry labeling, the brains were immersed in a 4% PFA/30% sucrose solution for 1 day. Tissues were sectioned coronally to a 90 µm thickness and then washed in PBS twice. For heat-induced antigen retrieval, the slices were boiled in 10 mM sodium citrate for 10 min. The sections were then incubated with primary antibodies on slides at a 1:200 dilution for two days at 4°C (Mouse monoclonal GAD-65 Antibody (A-3): sc-377145 Santa Cruz Biotechnology, Mouse monoclonal CaMKII Antibody (G-1): sc-5306 Santa Cruz Biotechnology, Rabbit polyclonal to GFAP: ab7260 abcam, Goat polyclonal anti-nNOS: ab1376 Abcam). The brain sections were washed twice and incubated for one hour with secondary antibodies (Abcam Alexa Fluor 488 Goat Anti-Mouse IgG H and L and Goat Anti-Rabbit IgG H and L). The brain sections were washed twice, and coverslipped with DAPI fluoroshield mounting media (Abcam ab104139). The sections were imaged on an Olympus FV10i Confocal. Manual quantification of cell specificity were done in 318 µm x 318 µm x 120 µm volumes. Type 1 nNOS neurons were distinguished from type two neurons by their substantially more intense staining for nNOS (Perrenoud et al., 2012).

Statistical analysis was performed using MATLAB (R2018, MathWorks). All summary data were reported as mean ± standard deviation and plotted as mean ± standard error unless otherwise specified. For comparisons of vehicle vs. treatment conditions, where multiple measurements of different vessel diameters were taken from the same mouse, we used the linear mixed effects (LME) model (MATLAB function: fitlme). For each manipulation (e.g. CNO vs. vehicle in DREADD-expressing mice), we fit a linear mixed effects model given by:∆Dn,a=γ+βawhere ∆Dn,a is the diameter differences between the vehicle and manipulation condition, a is the identifier of the animal group, and n is the identifier of the vessel. The γ term is the (fixed) effect of the manipulation, and βa is the within animal (random) effect that accounts for the within animal correlations (Aarts et al., 2014). To determine if the manipulation caused a significant change in the vessel diameter, we calculated if the γ term differed from 0, after a Bonferroni correction for the number of groups of DREADD-expressing mice (7 groups) or the number of infusions (6 groups). p-Values >1 after correction were rounded down to 1. A difference was determined to be statistically significant if p<0.05 after Bonferroni correction. For plotting the relationship of the basal diameter in the control condition versus the manipulation in the figures, we fit a linear mixed effects model given by:∆Dn,a=γ+βa+αnDnwhere Dn is the diameter of vessel n in the control condition, and αn is a vector of (fixed) coefficients.

For comparisons of vehicle vs. treatment conditions of the gamma-band (averaged power from 40 to 100 Hz after removing 60 Hz noise), we used the paired t-test (MATLAB function: ttest).