Unspecific expression in limited excitatory cell populations in interneuron-targeting Cre-driver lines can have large functional effects

Transgenic Cre-recombinase expressing mouse lines are widely used to express fluorescent proteins and opto-/chemogenetic actuators, making them a cornerstone of modern neuroscience. Particularly, the investigation of interneurons has benefitted from the ability to target genetic constructs to defined cell types. However, the cell type specificity of some mouse lines has been called into questions. Here we show for the first time the functional consequences of unspecific expression in a somatostatin-Cre (SST-Cre) mouse line. We find large optogenetically evoked excitatory currents originating from unspecifically targeted CA3 pyramidal cells. We also used public Allen Brain Institute data to estimate expression specificity in other Cre lines. Another SST-Cre mouse lines shows comparable unspecificity, whereas a Parvalbumin-Cre mouse line shows much less unspecific expression. Finally, we make suggestions to ensure that the results from in-vivo use of Cre mouse lines are interpretable.

. Particularly opto-and chemogenetic studies often depend on highly specific expression 44 patterns to disseminate the function of neuronal subtypes. Even though these findings are worrisome, 45 one defense of such mouse lines is that the absolute number of unspecifically targeted cells is small. 46 One could therefore assume that the observed in-vitro and in-vivo effects are dominated by the 47 interneuron type in question. 48 Here we show that in SST-Cre mice (Savanthrapadian et al., 2014), recombination is not only induced 49 in GABAergic interneuron types. In addition, recombination also occurs in a small subset of excitatory 50 neurons largely confined to the CA3 pyramidal cell layer. Moreover, we find powerful functional 51 effects of optogenetic activation that are not only contaminated by unspecifically expressing 52 glutamatergic cells but are completely lacking any interneuron contribution. Finally, we were also 53 unable to find anatomical or functional differences between unspecifically targeted cells and canonical 54 CA3 pyramidal cells. This suggests that these cells are not a specific subtype of CA3 pyramidal cell. 55 Further control experiments should be carried out in a region-specific manner, prior to the use of Cre-56 lines for the investigation of circuit function in behavior. 57 3 Methods 58

Transgenic Animals 59
All animal experiments were carried out according to the guidelines stated in Directive 2010/63/EU of  60  the European Parliament on the protection of animals used for scientific purposes and were approved  61 by authorities in Nordrhein-Westfalen (Landesamt für Natur, Umwelt und Verbraucherschutz 62 Nordrhein Westfalen (LANUV), AZ 84-02.04.2014.A254). 63 The SST-Cre mouse line was kindly provided to us by Marlene Bartos and was described previously 64 (Savanthrapadian et al., 2014). It was also used in a more recent study (Eyre & Bartos, 2019). Animals 65 were bred hemizygous and were genotyped for Cre recombinase using the forward primer 66 CCATCTGCCACCAGCCAG and the reverse primer TCGCCATCTTCCAGCAGG. Animals with an 67 amplified fragment at 281bp were classified as transgenic. For the cross-breeding experiments ( Figure  68 1C) we used the Ai14 reporter line (Jackson Laboratories Stock No 007914). 69

Stereotaxic intracranial viral injections 70
Animals were anesthetized with a ketamine/rompun or a fentanyl/midazolam/medetomidin mixture i.p. 71 Animals also received ketoprofen analgesia (5 mg/kg, 0.1 ml/10 g body weight) before the surgery and 72 daily 2 days after the surgery. Omicron Luxx 473nm laser attached to a light fiber submerged in the ACSF. Light stimuli were 5ms 119 long unless otherwise stated. 120 For the conductance analysis we assumed a chloride reversal potential of -80mV (-78.9mV calculated 121 with Nernst equation) and a cation reversal potential of 0mV. The excitatory conductance was 122 calculated from a current trace measured at a holding voltage near the chloride reversal with gabazine 123 washed-in, to ensure pure excitatory response. To isolate the inhibitory conductance, we subtracted the 124 pure excitatory response at a depolarized holding voltage from the mixed response in normal ACSF. 125 In Figure 2C we only included cells that showed complete block by TTX wash-in. We excluded one 126 cell that did not show complete block, which is likely due to a wash-in failure. 127

Electron microscopy with miniSOG photooxidation 128
SST-Cre animals were virally transduced with EF1a-DIO-Syp-miniSOG-T2A-mCherry. After illumination, slices were stored in cacodylate-buffer for further processing. 140 After photoconversion the converted region containing DAB reaction product in the hippocampus was 141 documented and images were taken at a Zeiss Axiophot light microscope. Thereafter the sections were 142 rinsed three times in 0.1 M sodium cacodylate buffer (pH 7.2-7.4) (Sigma-Aldrich, Germany) and 143 incubated with 1 % osmium tetroxide (Science Services, Germany) in cacodylate buffer for 20 minutes 144 on ice. The osmication of sections was followed by dehydration through ascending ethyl alcohol 145 concentration steps and rinsing twice in propylene oxide (Carl Roth, Germany). Infiltration of the 146 embedding medium was performed by immersing the sections first in a mixture of 2:1 of propylene 147 oxide and Epon (Carl Roth, Germany) then in a 1:1 mixture and finally in neat Epon and polymerised 148 at 60°C for 48 hours. The region of interest was dissected and Ultrathin sections (60 nm) were prepared 149 with a Leica Ultracut UC7. Images were taken using an EM902 transmission electron microscope 150 (Zeiss, Germany) equipped with a CCD in lens 2K digital camera and running the ImageSP software 151 (Tröndle, Moorenweis, Germany). 152

Quantification and Statistical Analysis 153
We used python with matplotlib (Hunter, 2007) Table 1: Experiments and images from the Allen Brain Institute used for the quantification in Figure  166 5. All images can be found here: http://connectivity.brain-map.org/transgenic 167 168 4 Results 169

The SST-Cre line is not specific for SST + interneurons in CA3 170
SST positive interneurons in CA3 are located predominantly in stratum oriens (SO) and stratum 171 radiatum (SR). SST positive cells have a characteristic dendrite morphology, with most of the dendritic 172 arbor confined to the same sublayer as the soma (Freund & Buzsáki, 1996). We expressed a construct 173 that leads to Cre-dependent expression of EYFP in the CA3 region of heterozygous SST-Cre mice 174 using rAAV-dependent gene transfer. We found EYFP expression in cells of the pyramidal cell layer 175 (PCL; Figure 1A). In stratum oriens and stratum radiatum, cells also expressed EYFP but the signal 176 there was almost dominated by the neuropil. EYFP + cells in the PCL showed features typical for CA3 177 pyramidal cells ( Figure 1B) such as thorny excrescences on apical dendrites. 178 To determine if these EYFP + cells are also SST + , we immunostained for SST. This revealed that EYFP 179 expression was highly specific for SST + interneurons in SO, where 50/53 EYFP + cells expressed SST. 180 Similarly, in SR 9/10 EYFP + cells expressed SST. In marked contrast we found that a minority of 181 EYFP + cells in the pyramidal cell layer of CA3 coexpressed SST (21/147 cells, Figure 1D,E). In 182 addition to viral gene transfer of a reporter construct, we also crossed mice of our SST-Cre line with a 183 tdTomato reporter mouse (line Ai14, see methods). As in the previous experiment, we found that most 184 reporter positive cells were in the PCL and showed pyramidal like dendritic morphology in CA3, CA2 185 and CA1. Interestingly, we also found a very small number of granule cell-like neurons in the granule 186 cell layer of the dentate gyrus ( Figure 1C) that were not observed in virally transduced animals. 187 These results show that Cre recombinase is not only targeted to SST + interneurons in the hippocampus. 188 It is also expressed in pyramidal-like neurons within the pyramidal cell layer that are devoid of 189 detectable somatostatin levels. In marked contrast, the SST-Cre mouse line showed high local 190 specificity in CA3 SO, SR and the hilus of the dentate gyrus. ( Figure 2C,D,E). To ascertain which of these components are monosynaptic in nature, we applied the 215 Na + channel blocker tetrodotoxin (TTX, 1µM), which invariably blocked synaptic transmission 216 completely. Coapplying TTX with 4-aminopyridine (4-AP, 200µM) enables direct light-based 217 transmitter release from terminals expressing ChR2, and thus indicates monosynaptic connections. 218 Coapplication of 4-AP recovered EPSCs, but not IPSCs (Figure 2C; EPSCs 217%, IPSCs 1% of 219 baseline). The recovery of EPSCs but not IPSCs indicates that contralateral projections in SST-Cre 220 mice are excitatory. Additionally, these results indicate that the light-evoked IPSCs are due to 221 polysynaptic recruitment of interneurons. This idea is supported by the temporal delay between 222 excitatory and inhibitory conductances ( Figure 3G). Consistent with polysynaptic recruitment of 223 inhibitory interneurons, light-evoked IPSCs were abrogated by blocking glutamatergic transmission 224 with CNQX (50µM) and D-AP5 (200µM; Figure 2D; EPSCs 29%, IPSCs 8% of baseline). Finally, 225 we show thatas expectedlight-evoked IPSCs were sensitive to the GABA-A blocker gabazine 226 (10µM; Figure 2E; EPSCs 114%, IPSCs 14% of baseline). 227 Taken together, we found no evidence for direct commissural inhibition from SST + interneurons from 228 CA3 to CA1. Instead, direct excitatory transmission recruited strong polysynaptic inhibition. 229

Unconditionally transduced and SST-Cre fibers are functionally indistinguishable in 230
Contralateral CA1 231 To investigate if this is consistent with the canonical CA3 to CA1 commissural projection, we induced 232 broad expression of ChR2 in all CA3 cell types using viral gene transfer of an unconditional construct 233 leading to expression of EYFP-hChR2. Light based manipulations should be dominated by activity of 234 pyramidal cells, since they vastly outnumber other neuronal subtypes. Virus injection resulted in strong 235 fluorescence signal in CA1, CA3 and DG that was dominated by fiber signal at the injection site 236 (Figure 3A). Contralateral to the injection site, we found prominent labelling of axons in CA1 and C3 237 in both SR and SO as well as the inner molecular layer of the DG. The DG fiber pattern was consistent 238 with the commissural mossy cell projection and the fiber patterns in CA1 and CA3 with the 239 commissural CA3 projection. We again assessed the monosynaptic transmission onto contralateral 240 CA1 pyramidal cells using combined application of TTX and 4-AP (1µM, 200µM) and found that it 241 completely inhibited IPSCs ( Figure 3C; EPSCs 88%, IPSCs 4% of baseline). Next, we asked if there 242 are quantitative differences between the SST-Cre fibers and the unconditionally transduced fibers. We 243 converted the pharmacologically isolated currents ( Figure 2E) to conductances ( Figure 3D) according 244 to holding and reversal potentials (see Methods). Because the density of EYFP-hChR2 positive fibers 245 is much larger in the unconditional case, the absolute conductances cannot be compared meaningfully. 246 However, because the inhibition is polysynaptic, it is expected to scale to some extent with the 247 excitation. Therefore, the ratio between excitation and inhibition can give insights into differential 248 recruitment in the micronetwork. 249 We found that in the SST-Cre line, the inhibitory conductance was stronger than the excitatory one 250 (Figure 3D,E). Comparing the SST-Cre line with the unconditional case, we did not detect a difference 251 between the ratios of maximum inhibition and excitation ( Figure 3F). In both cases, the amplitude of 252 inhibition was larger than that of inhibition for different strengths of light-based stimulation. 253 Furthermore, the latencies between the onset of excitation and inhibition showed no significant 254 difference ( Figure 3H) and were consistent with values found in CA3 to CA1 Schaffer collateral 255 projections (Scanziani, 2001). However, the latencies between the peak of the excitatory conductance 256 and the inhibitory conductance showed a significant interaction between laser output and the type of 257 expression. Main effects were not significant (Figure3G, Greenhouse-Geisser corrected 2-way 258 ANOVA). 259

Commissural CA3 fibers make synaptic contacts on spines and originate primarily from 260 PCL cells 261
To further confirm that contralateral projections are excitatory, we used miniSOG photooxidation to 262 generate electron-dense labelling in contralateral CA1 SO localized to fibers with Cre recombinase 263 activity in the SST-Cre line (Figure 4). Of 70 miniSOG positive structures, 40 clearly were presynaptic 264 boutons making postsynaptic contacts. All 40 structures made contact on a spine, 4 of them made 265 contact on 2 spines ( Figure 4A). Serial imaging sections of 25 boutons showed that 22 of them 266 unambiguously made contact on spines ( Figure 4B). The other three boutons were not entirely 267 sectioned. The types of most synaptic contacts could not be defined clearly because of the electron 268 dense labelling in the pre-synapse. However, the postsynaptic densities that are clearly in the imaging 269 plane appear asymmetric. Together with the fact that they all contact spines, this data suggest that the 270 direct contacts are predominantly excitatory, and we found no evidence for direct inhibitory contacts 271 in CA1 SO. 272 Next, we used retrograde tracing in CA1 with CT-B to determine which cell types project to 273 contralateral CA1 (Figure 4C). We found that virtually all projecting cells were in the CA3 pyramidal 274 cell layer. With the SST staining we identified 81 cells, none of which was CT-B positive. This data 275 suggests that somatostatin interneurons are not part of the commissural projection. 276 Finally, we tried to relate our findings to other commonly used Cre mouse lines. Therefore, we used 277 data from the Allen Brain Institute to estimate specificity in two other interneuron targeting Cre lines. We found that the SST-IRES-Cre mouse line is similarly unspecific in CA3, with only 48/127 (37.8%) 286 tdTomato + cells being SST-mRNA + in the PCL, 82/100 (82%) in SO and 61/74 (82.4%) in SR ( Figure  287 5B,C). The CA1 area also contained some SST-cells in the PCL but appeared overall more specific 288 with 29/51 (56.9%) tdTomato + cells being SST-mRNA positive, 281/299 (94%) in SO and 20/24 289 (83.3%) in SR (Figure 5D,E). The total amount of cells located in PCL is much smaller in the SST-290 IRES-Cre line (Figure 5A) than the SST-Cre line when crossed with the Ai14 reporter line ( Figure  291 1C). 292 We also quantified colocalization of Cre-induced recombination with PV expression in the Pvalb-293 IRES-Cre mouse line (Hippenmeyer et al., 2005). We found that this mouse line was much more 294 specific than both SST-Cre mouse lines in both the CA3 and CA1 regions (Figure 5F-J 29/34, 85.3% SR). In summary, the problems we described are common to at least one other commonly 297 used SST-Cre line, but less severe in another mouse line used to study interneurons. 298

Discussion 299
We show for the first time that CA3 PCs that are unspecifically targeted in an SST-Cre mouse line 300 make functional connections indistinguishable from those of canonical CA3 PCs. While the specificity 301 of SST-Cre lines has been questioned before, the functional relevance of unspecific expression of Cre-302 recombinase was unknown. Estimating the potential effects of unspecific expression is essential for 303 neuronal perturbation studies that seek to isolate the function of specific cell-types. Our data suggest 304 that studies that perturb SST cells in CA3 with the two examined SST-Cre lines would be massively 305 confounded by Cre recombinase expression in CA3 pyramidal cells. 306

How relevant are these findings for other Cre mouse lines? 307
We demonstrate wide-spread physiological effects of unspecific Cre-expression in only one mouse 308 line, but have identified unspecific expression in another, widely used mouse line following 309 quantitative analysis of Allen Brain Atlas data. Indeed, specificity issues with an SST-Cre mouse line 310 were raised previously (Taniguchi et al., 2011). Moreover, a further study has found targeting of a large 311 number (31%) of slow-spiking cells in the CA1 PCL, also consistent with unspecific expression 312 (Mikulovic et al., 2015). It is worth noting that the mouse line we used was generated by the same 313 method as most modern Cre mouse lines, the BAC technology. Specificity can vary widely between 314 Cre lines and brain areas, as our comparison of the SST-IRES-Cre and the Pvalb-IRES-Cre lines shows. 315 Therefore, specificity should not be generalized lightly to other Cre mouse lines or even to other brain 316 areas in the same mouse line. We suggest that pending careful quantitative analysis in all the subregions 317 under investigation in the specific study, caution is warranted in assuming specificity. 318

Do SST-expressing interneurons make contralateral connections? 319
Despite the specificity issues, the SST-Cre mouse line clearly targets SST + INs in CA3. We found no 320 evidence for direct inhibition from those cells onto contralateral CA1 PCs in our patch clamp 321 experiments. Even the unconditionally transduced slices did not reveal monosynaptic inhibition, 322 despite targeting all inhibitory cell types. Furthermore, our anatomical EM data showed no evidence 323 for inhibitory synapses in contralateral CA1 SO. The CT-B data did not reveal cells outside CA3 PCL 324 projecting to contralateral CA1. This leads us to the conclusion that an inhibitory CA3 to contralateral 325 CA1 connection is nonexistent or extremely weak and SST + interneurons do not contribute to it. 326 Although we focused on CA3 and CA1, we noted very sparse fiber signal in the outer molecular layer 327 of DG in the SST-Cre line. Datasets are available on request. The raw data supporting the conclusions of this manuscript will be 371 made available by the authors, without undue reservation, to any qualified researcher. 372 9 Conflict of Interest 373 The authors declare that the research was conducted in the absence of any commercial or financial 374 relationships that could be construed as a potential conflict of interest. Quantification on the right. F) Experiment 111192541, image ID 111192610. Contrast auto adjusted 507 and lookup tables changed. G-J) Example images cropped from F) contrast unadjusted. Quantification 508 on the right. Scalebars: 100 and 20µm. 509