ABSTRACT
Type I lissencephaly is a neuronal migration disorder caused by haploinsuffiency of the LIS1 gene and is characterized in humans by agyria, mislamination of brain structures, developmental delays, and epilepsy. Here, we investigate the impact of LIS1 mutation on the cellular migration, morphophysiology, microcircuitry and genomics of mouse hippocampal CA1 parvalbumin-containing inhibitory interneurons (PV+INTs). We find that WT PV+INTs consist of two physiological subtypes (80% fast-spiking (FS), 20% non-fast-spiking (NFS)) and four morphological subtypes (basket, axo-axonic, bistratified, radiatum-targeting). We also discover that cell-autonomous mutations within interneurons disrupts morphological development of PV+INTs and results in the emergence of a non-canonical “intermediate spiking (IS)” subset of PV+INTs. In the GlobalLis mutant, IS/NFS cells become the dominant PV+INT subtypes (56%) and the percentage of FS cells shrinks to 44%. We also find that IS/NFS cells are prone to entering depolarizing block, causing them to temporarily lose the ability to initiate action potentials and control network excitation, potentially promoting seizures. Finally, single-cell nuclear RNAsequencing of PV+INTs revealed several misregulated genes related to morphogenesis, cellular excitability, and synapse formation.
INTRODUCTION
Excitation in neocortical and hippocampal circuits is balanced by a relatively small (10-15%) yet highly heterogenous population of GABAergic inhibitory interneurons (Pelkey et al., 2017). During embryogenesis inhibitory interneurons are generated in the ganglionic eminences, then tangentially migrate to reach their cortical or hippocampal destination (Bartolini et al., 2013). Upon reaching the cortex or hippocampus, interneurons migrate along radial glial cells to their final laminar position and integrate into developing circuits (Lim et al., 2018). The process of neuronal migration requires molecular interactions of motor proteins with networks of microtubules; therefore, genetic mutations disrupting such proteins compromise neuronal migration and promote abnormal brain development (Wynshaw-Boris and Gambello, 2001; Corbo et al., 2002; Tissir and Goffinet, 2003).
LIS1 (PAFAH1B1) encodes a protein (Lis1) that regulates dynein microtubule binding and is essential for neuronal migration (Wynshaw-Boris and Gambello, 2001). Consequently, LIS1 haploinsufficiency results in classical, or Type I, lissencephaly (“smooth brain”), a rare neurodevelopmental disorder characterized in humans by brain malformation, intellectual disability, motor impairment, and drug-resistant epilepsy (Kato and Dobyns, 2003; Di Donato et al., 2017). Full loss of LIS1 is embryonically lethal (Hirotsune et al., 1998).
Classical lissencephaly can be modeled in mouse lines generated through heterozygous removal of LIS1, which results in enlarged ventricles and disorganization of brain structures (Hirotsune et al, 1998). Structural abnormalities are particularly notable in the hippocampus, where the normally tightly compacted layer of pyramidal cells fractures into multiple bands of loosely organized cells (Fleck et al., 2000, D’Amour et al 2020). Mice heterozygous for LIS1 share symptoms with human lissencephaly patients, including learning deficits, motor impairments, increased excitability and decreased seizure threshold (Paylor et al., 1999; Fleck et al., 2000; Greenwood et al., 2009; Menascu et al., 2013; Herbst et al., 2016). Due to the high density of recurrent excitatory connections and the reliance on inhibitory interneurons to control network excitability, the hippocampus and neocortex are prone to generating epileptic seizures (McCormick and Contreras, 2001). Thus, the increased propensity for seizures in LIS1 mutants may be indicative of dysfunctional inhibition. Indeed, specific deficits in inhibitory interneuron wiring with pyramidal cell targets have been identified in LIS1 mutant mice, but the origin of seizures remains unclear (Jones & Baraban, 2009; D’amour et al., 2020).
Inhibitory interneurons are classified based on a combination of their morphological, biochemical, intrinsic electrical, and connectivity properties (Lim et al., 2018). Advances in single-cell RNA sequencing have revealed enormous diversity in interneuron genomics, and current efforts attempt to correlate transcriptomic data sets with previously identified interneuron subtypes (Tasic et al., 2018; Muñoz-Manchado et al., 2018; Gouwens et al., 2019; Que et al., 2020). In CA1 hippocampus alone, inhibitory synaptic transmission is mediated by at least 15 different subtypes of GABAergic inhibitory interneurons (Pelkey et al., 2017). Three canonical interneuron subtypes express the calcium-binding protein parvalbumin (PV): basket-cells, axo-axonic cells and bistratified cells. PV-containing inhibitory interneurons (PV+INTs) are often classified as “fast-spiking” cells due to their ability to sustain high-frequency discharges of action potentials with minimal spike-frequency adaptation/accommodation (Pelkey et al., 2017). Fast-spiking interneurons are essential for proper network oscillations and disrupting the function of PV+INTs can generate spontaneous recurrent seizures (Drexel et al., 2017; Panthi and Leitch, 2019). Recent transcriptomics suggests that there are several genomically distinct subpopulations of PV+INTs (Hodge et al., 2019; Gouwens et al., 2020), some of which may correspond to unique PV+INT subtypes that have remained largely understudied relative to the canonical FS subtypes listed above.
A current model for the formation of neural circuits posits that pyramidal cells (PCs) instruct radial migration and synaptic connectivity of INTs (Pelkey et al., 2017; Wester et al., 2019). In the cortex, INTs are initially dispersed throughout cortical layers, only sorting into their final positions between the 3rd and 7th postnatal day (Miyoshi & Fishell, 2010). Interneurons have programs that enable both cell type-specific and cellular compartment-specific targeting. For example, PV+INTs make connections with PCs and other PV+INTs, but rarely contact other subtypes of INTs (Kohus et al., 2016). Furthermore, different subtypes of PV+INTs target specific regions of PCs such as dendrites (bistratified cells), the axon initial segment (axo-axonic cells), or the perisomatic region (basket cells; Pelkey et al., 2017). Mutations to chemokine receptors can alter this connectivity, and complete loss or reprogramming of cellular identity is possible when proteins are missing in development (Ye et al., 2015; Pelkey et al., 2017; Mahadevan et al., 2020).
Previous lissencephaly studies have demonstrated that migration of inhibitory interneurons is disrupted in LIS1 heterozygous mutants (Fleck et al., 2000; McManus et al., 2004). In particular, PV+INTs adopt atypical positions in the hippocampus, including between heterotopic bands of pyramidal cells and within stratum radiatum, a layer where PV+INTs are rarely found in wildtype (WT) CA1 (Fleck et al., 2000; Jones and Baraban, 2009; D’Amour et al., 2020). Despite ectopic positioning of inhibitory interneurons and layer-specific reorganization of inhibitory inputs, the nature and consequences of PV+INT morphophysiological development and microcircuit organization following LIS1 mutations have remained elusive.
Here we report the impact of LIS1 mutations and resulting neuronal migration deficits on the lamination, morphology, intrinsic physiology, connectivity, synaptic transmission dynamics and genomics of hippocampal parvalbumin-containing inhibitory interneurons. Cell-autonomous loss of LIS1 within interneurons results in the emergence of a novel physiological population of PV+INT, comprising ~50% of the total PV+INT cohort. Compared to canonical FS PV+INTs, these altered cells have lower firing rates, provide less reliable inhibition to pyramidal cells and have a higher propensity to enter depolarization block. Single-cell nuclear RNA sequencing (snRNA-seq) revealed multiple disruptions to the expression of ion channels regulating PV+INT excitability. We propose that disrupted physiological development and deficient inhibitory output of PV+INTs likely contributes to the spontaneous seizures observed in classical lissencephaly.
RESULTS AND DISCUSSION
Generation and characterization of LIS1 mutant lines
To investigate the cell-autonomous and non-autonomous effects of LIS1 heterozygous mutations on PV+INT migration and development, we crossed LIS1floxedl/+ breeders to three separate Cre lines: Sox2-Cre to generate heterozygous LIS1 mutations in all cells (“GlobalLis”); Nkx2.1-Cre to generate heterozygous LIS1 mutations specifically in medial ganglionic eminence-derived interneurons (“NkxLis”); and Emx1-Cre to generate heterozygous mutations specifically in pyramidal cells (“EmxLis”). These lines were further crossed to PV-TdTomato (TdT) reporter lines to enable selective targeting of PV+INTs during physiological recordings.
As previously reported (Hirotsune et al., 1998; Fleck et al., 2000; D’Amour et al., 2020) hippocampal lamination is disrupted in GlobalLis mice. The normally compact layer of pyramidal cells (PCs) (stratum pyramidale; s.p.) fractures into heterotopic bands, typically with a normotopic layer resembling the WT band, and an ectopic layer often fragmented into stratum oriens (s.o.; Figure 1A). This general pattern of disrupted hippocampal lamination is also observed in the EmxLis mouse line, but not in the NkxLis mouse, indicating that LIS1 expression in pyramidal cells, but not MGE-derived interneurons, is essential for proper hippocampal pyramidal cell layer formation (Figure 1A).
Radial migration of inhibitory interneurons is disrupted in LIS1 mutants
Previous studies have demonstrated severe cellular disorganization in rodent models of lissencephaly, including impaired of parvalbumin-containing inhibitory interneurons (PV+INTs; Fleck et al., 2000; Jones and Baraban, 2009; D’Amour et al., 2020). To investigate the nature of the aberrant migration of PV+INTs we first quantified their relative densities using immunohistochemistry. GlobalLis mutants exhibited no overall change in PV density in the CA1 subfield. However, as we reported previously (D’Amour et al 2020) PV+INT density was reduced in both s.o. and s.p. and increased in stratum radiatum (s.r.) and stratum lacunosum-moleculare (s.l.m.; Figure 1B-C). In WT CA1, an overwhelming majority (>95%) of PV+INTs are found in s.o. and s.p., while less than 5% reside in s.r. or s.l.m., indicating a strong preference for PV+INTs to typically inhabit deeper regions of the hippocampus. In contrast, in GlobalLis 77% of PV+INTs were found in s.o/s.p, with the proportion of cells in s.r./s.l.m. expanded to 23% of the total PV population (Figure 1D).
We next used the NkxLis and EmxLis lines, where LIS1 is eliminated only in MGE-derived interneurons and pyramidal neurons respectively, to assay the impact of cell-autonomous and non-autonomous mutations on PV+INT migration. Interestingly, both genotypes had similar patterns of PV+INT somatic distribution to that observed in the GlobalLis CA1. In these mutants there was a significantly decreased density of PV+INTs in s.o./s.p. and increased density in s.r./s.l.m., with ~80% of PV+INTs found in deep regions (s.o/s.p.) and ~20% found in superficial regions (s.r./s.l.m.; Figure 1B-D). Thus, proper migration and lamination of PV+INTs requires both Lis1-dependent cell-intrinsic mechanisms (as revealed by the disruption in NkxLis PV cells) and non-cell-autonomous cues from pyramidal neurons (demonstrated by the disruption in EmxLis PV cells).
LIS1 heterozygous mutation alters PV+INT morphophysiological development
The vast majority of parvalbumin-containing inhibitory interneurons are classified as “fast-spiking” (FS) cells, due to their ability to sustain high-frequency discharges of action potentials (Pelkey et al., 2017). In cortical circuits, FS cells contribute to both feedforward and feedback inhibition and are essential in generating network oscillations; as such, disrupted FS cell function can lead to uncontrolled excitation and seizures (Hu et al., 2014).
To functionally characterize hippocampal PV+INTs we first examined the intrinsic electrophysiological properties of TdTomato+ (TdT) WT CA1 PV cells (Figure 2A). Accuracy of TdT labeling of PV+INTs was confirmed by quantifying the percentage of TdT and PV-immunostaining overlap: 92% of WT and 93% of GlobalLis hippocampal TdT+ cells were immunopositive for PV, and 97% of the immunopositive PV cells were labeled by TdT, enabling reliable targeting of PV+INTs (Figure S1).
WT fast-spiking PV+INTs have stereotypical intrinsic properties that include a low input resistance (75-90 MΩ), high rheobase (330-400 pA), high firing frequency at 2x (130-150 Hz) and 3x threshold (160-180 Hz), narrow action potential half-width (0.40-0.44 ms) and minimal spike-frequency-adaptation (0.75-0.85; Table 1). Surprisingly, we routinely observed an additional population (~15-20%) of WT TdT-labeled PV+INTs that did not possess stereotypical fast-spiking firing properties. This NFS subpopulation also displayed numerous intrinsic properties distinct from canonical FS cells, including a low firing frequency, high input resistance and low rheobase (i.e. Figure 2A top right). Clusters of atypical PV+INTs have been previously reported in subiculum (“quasi fast-spiking interneurons”; Nassar et al., 2015) and striatum (“fast-spiking-like cells”; Munoz-Manchado et al., 2018).
To independently verify whether WT hippocampal PV+INTs could be functionally segregated into distinct clusters in an unbiased fashion we performed principal component analysis (PCA) using several key intrinsic physiological features (action potential half-width, firing frequency at 2x threshold, firing frequency at 3x threshold, adaptation ratio at 2x threshold, input resistance, rheobase, sag index). WT PV+INTs neatly parsed into two subtypes comprised of a large majority FS cohort, and a small minority atypical subset (Figure 3A-D). Due to their intrinsic physiological differences from FS cells and inability to sustain high frequencies of action potentials, for ease of discussion we designate this unique subpopulation of PV+INTs as “non-fast-spiking” (NFS) cells. PV+NFS cells have lower firing frequencies at 2x (55-75 Hz) and 3x (85-105 Hz) threshold, lower adaptation ratios (0.60-0.75) broader action potential half-widths (0.55-0.65 ms), larger input resistances (120-170 MΩ), and lower rheobases (120-180 pA) than standard PV+FS cells (Figure 3E; Table 1).
With respect to their morphology, hippocampal PV+INTs are routinely parsed into three primary subtypes based on axonal arborization: basket cells (BCs; which target cell pyramidal cell (PC) bodies and proximal dendrites), axo-axonic cells (AACs; which target PC axon initial segments), and bistratified cells (BiCs; which target PC apical and basal dendrites; Pelkey et al., 2017). Post-hoc anatomical recoveries of recorded cells regularly revealed these three standard morphologies, and additionally a unique hippocampal PV+INT, which we designate “radiatum-targeting cells” (RTC) as this subtype confines its axon to s.r. and is presumably a subtype of dendrite-targeting cell. Figure 2A shows typical morphologies and firing patterns at threshold (with expanded action potentials; red) and at 2x threshold current injection (phase plots in blue) of WT CA1 PV+INTs (from left to right: BC, AAC, BiC, PV+RTC).
Hippocampal layers differ in their composition of PV+INT morphophysiological subtypes. In WT deeper regions (s.o./s.p.) are populated primarily by PV+FS cells (~90%), consisting of all of the morphological forms (BC, AAC, BiC, RTC). In contrast, all PV+NFS cells residing in s.o./s.p. had BC morphologies (Figure 3F-G). The small number of PV+FS and NFS cells found in superficial layers (s.r.) all had RTC morphology (Figure 3F-G). In summary, WT hippocampal CA1 PV+INTs consist of two physiological (FS and NFS) and four morphological subtypes (BC, AAC, BiC, RTC), and the overwhelming majority of PV+INTs are found in deep hippocampal layers, consistent with previous reports (Pelkey et al 2017).
Using the same strategy, we next targeted PV+INTs in the GlobalLis mouse line and utilized PCA to cluster GlobalLis PV+INTs by their intrinsic physiological properties. Unlike WT PV+INTs, GlobalLis cells segregated into three clusters. (Figure 4A-D). In addition to the FS and NFS cell clusters, a third entirely new cluster of PV+INTs emerged (Figure 4A-D). GlobalLis FS cells had identical intrinsic properties as WT FS cells with the exception of shorter action potential half-widths (0.32-037 ms vs 0.41-0.44 ms; Table 1). GlobalLis NFS cells were indistinguishable from WT NFS cells. A number of intrinsic physiological properties of the third group of cells fell between those of FS and NFS cells, including firing frequency at 2x (90-110 Hz) and 3x threshold (120-140 Hz), adaptation ratio (0.65-0.85), AP half-width (0.45-0.50 ms) and input resistance (80-100 MΩ; Figure 4E; Table 1). Consequently, we refer to this emergent physiological PV+INT subtype as “intermediate spiking” (IS) cells. Of particular interest 47% of all GlobalLis PV+INTs consisted of the IS subtype and only 44% consisted of FS cells (Figure 4F) compared the 80% observed in WT CA1 PV+INTs. Within CA1, superficial layers had proportionally more IS cells (~60%) than deeper layers (~25%; Figure 4F).
In addition to somatic misplacement and physiological disruption, the morphological development of GlobalLis PV+INTs is radically disrupted. Many cells did not resemble stereotypical morphologies of any WT PV+INT subtype and often had ectopic axonal branching that extended in all directions, rather than forming the tight plexus seen in WT. Other mutant PV+INTs took on combinations of hippocampal PV cell features including bistratified-like cells with baskets, and radiatum-targeting cells with axons that also extended into the s.l.m. (Figure 2B).
Sholl analyses of digitally reconstructed axonal and dendritic arbors or recorded PV+INTs revealed that WT and GlobalLis FS cells and GlobalLis IS cells had larger and more complex axonal and dendritic trees than NFS cells of both genotypes (Figure S2A-D). The smaller axonal arbors of NFS cells (WT: 131 Sholl intersections, GlobalLis: 133 intersections) relative to FS (WT: 222 Sholl intersections, GlobalLis: 271 intersections) and IS cells (224 intersections) may imply lower synaptic connectivity of this subtype (Figure S2D; Table 1). In conclusion, global mutations to LIS1 disrupt morphophysiological identity in a large number of PV+INTs, however overall axonal and dendritic growth is not inhibited.
Cell-autonomous LIS1 mutation within interneurons disrupts PV+INT development
We next recorded from PV+INTs in both the EmxLis and NkxLis mutants to determine if the disruption of PV+INT morphophysiological development emerges from interactions in a malformed hippocampus (EmxLis, Figure 1B-D) or from selective disruption of inhibitory interneuron intrinsic developmental programs (NkxLis, Figure 1B-D). PCA of intrinsic physiological properties of PV+INTs from the EmxLis hippocampus identified 2 clusters (corresponding to FS and NFS), consistent with WT (Figure S3A). To increase the power of the PCA and confirm accurate clustering, we combined and analyzed the WT and EmxLis datasets together. Combining WT and EmxLis cells resulted in identical clusters of FS and NFS cells, indicating the ease of identifying normally developed FS and NFS cell types (Figure S3B). The total ratio of FS/NFS cells in CA1 as well as in each individual layer remains relatively unchanged in the EmxLis hippocampus (Figure S3C). Furthermore, morphological development of PV+INTs was relatively unaffected by non-autonomous LIS1 mutation: PV+ cells developed into identifiable BCs, BiCs, AACs, and RTCs (Figure S3D). In conclusion, in EmxLis mutants PV+INTs develop into canonical morphophysiological subtypes despite disorganized pyramidal cell layers.
We next examined PV+INTs in the NkxLis (interneuron-specific mutation) hippocampus and used PCA to analyze the physiological properties. Similar to the GlobalLis mutant, we detected 3 clusters of PV+INTs, however clustering accuracy was not optimal (i.e. occasional IS cells were classified as NFS cells etc.; Figure S4A). To increase analytical power and improve clustering accuracy, we combined the NkxLis and GlobalLis datasets and again found three clusters: FS, IS and NFS cells (Figure S4B). The novel IS cell subtype accounts for ~50 percent of total PV+INTs in both the NkxLis and GlobalLis genotypes, with more IS cells found in superficial hippocampal layers (Figure S4C). Thus, with cell-autonomous LIS1 mutations within interneurons, the overall distribution of PV+INT physiological subtypes shifts away from the 80% FS, 20% NFS distribution in WT (and EmxLis) to 44% FS, 56% IS/NFS in the GlobalLis mutant and 29% FS, 71% IS/NFS in the NkxLis mutant. Despite the normal lamination of pyramidal cell layers in the NkxLis hippocampus, the morphology of NkxLis PV+INTs appear similar to the GlobalLis mutants; cells take on hybrid combinations of morphological features and distinguishing meaningful subtypes becomes nontrivial, if not impossible (Figure S4D).
Finally, we tested whether subtypes of PV+INTs (FS, IS, NFS) shared intrinsic physiological properties across genotypes. Similar to our observation of shorter action potential half-widths in GlobalLis FS cells, half-width was also shorter than WT in NkxLis (0.32-0.38 ms) and EmxLis (0.34-0.39 ms) genotypes (Table 1). All other intrinsic properties of PV+INT physiological subtypes were preserved across all genotypes (Figure S5; Table 1). Taken together, these results indicate that while both cell-autonomous and non-autonomous LIS1 mutations can disrupt somatic positioning, only cell-autonomous mutations within interneurons perturb the morphophysiological identity of PV+INTs.
PV+INT microcircuit rearrangements in the GlobalLis hippocampus
We next investigated the impact of LIS1 haploinsufficiency on microcircuit formation by examining unitary inhibitory post synaptic currents (uIPSCs) using dual whole-cell recordings between synaptically coupled pairs of PV+INTs and CA1 pyramidal cells (Figure 5A-B).
In general, the connection probability of WT FS cells onto PCs was higher than that observed between NFS cells and PCs (32% vs 14% connected). Connectivity rates of GlobalLis FS cells (27%) were similar to WT FS cells. IS cells had a 27% connection probability onto PCs and 0/3 NFS cells were connected to PCs (Figure 5C).
WT FSBCs connected to PCs with high potency synapses (170±75 pA) which had an extremely high neurotransmitter release probability (0.99±0.01) and quick latency to release (1.2±0.1 ms). Compared to FSBCs, WT FSBiC-PC connections had weaker synapses (25±5 pA), slightly lower release probability (0.89±0.05) and a nearly identical latency (1.1±0.1 ms), while NFSBCs also had weaker connections (50±15 pA), they also had lower release probability (0.83±0.12) but longer transmission latency (2.4±0.3 ms). In the GlobalLis mutant, FS-PC connections were weaker (70±15 pA) than WT FSBCs, but not significantly different when we pooled the datasets of WT FSBC and FSBiC-PC connections (110±45 pA). GlobalLis FS cell release probability (0.97±0.02) and latency (0.9±0.1 ms) were indistinguishable from WT FS cells. Similar to NFS cells, IS cells had lower release probability (0.83±0.06) and longer latency (1.7±0.2 ms), contrasting FS-PC connections (Figure 5D; Table 2). Unfortunately, no GlobalLis NFS-PC connections were obtained.
We next investigated short-term transmission dynamics using high-frequency trains of presynaptic action potentials (Figure 5B). All PV-PC pairs showed characteristic synchronous neurotransmitter release and marked synaptic depression across a 50 Hz, 25 pulse train. In WT FSBCs, by the end of the train, uIPSC amplitude at the end of the 50 Hz train was reduced by 70% and the failure rate increased to 29%. Transmission at WT FSBiC connections fell by 61% and possessed a higher failure rate (58%). Unitary connections at WT NFS cells fell to 82% of the first pulse and the failure rate was 73%. In the GlobalLis mutant, FS-PC connection strength dropped by 70% and the failure rate increased to 32% by the end of the train; values close to those seen in WT FS cells. Similarly, GlobalLis IS connections dropped by 71% but the failure rate increased to 62%, similar to NFS cells (Figure 5E). Collectively, these results confirm that at the monosynaptic level, both the unitary synaptic amplitudes and the short-term dynamics of transmission are unchanged in GlobalLis FS-PC connections, however IS-PC synapses resemble those of NFS cells.
Depolarization block of action potential firing is a common feature of IS/NFS but not FS cells
PV+INTs are critical regulators of network excitability, such that their rapid action potential and transmission kinetics act to generate both feedforward and feedback inhibition and local oscillations, as well as preventing cortical network activity from disintegrating into electrographic events that are the underpinnings of seizure activity. Recent evidence suggests that in epileptic human patients, PV+INT function becomes compromised, and overactivation of PV+INTs can drive action potentials into depolarization block, causing a temporary loss of their action potential initiation, consequently resulting in a net decrease in inhibitory tone, and precipitation of seizure propagation (Ahmed et al., 2014; Sudhakar et al., 2019)
We were struck by our observation that the majority of PV+INTs in the GlobalLis mouse are IS/NFS cells (56%) compared to FS, which represent only 44% of the total PV+INT population (cf. 80% in WT). Both IS and NFS cells possess lower maximal firing frequencies and longer duration action potentials compared to FS cells. This relative shift in the overall PV+ cell population and their network dynamics may compromise PV+INT network control over network excitability in the LIS1 mutant and promote lower seizure thresholds observed in previous studies (Fleck et al., 2000).
FS PV+INTs are endowed with both voltage-gated Na+ and K+ channels that enable rapid action potential repolarization and conductance deinactivation, permitting repetitive high frequency firing with little firing accommodation and a resistance to depolarizing block (Rudy & McBain, 2001; and Hu et al., 2014; Pelkey et al., 2017). However, it is unclear whether either NFS or IS PV+INTs share the same resistance to firing accommodation or depolarization block of AP firing during sustained excitation.
Depolarization block susceptibility of WT and GlobalLis PV+INTs was first assayed by injecting PV+INTs with increasing suprathreshold currents (500 ms duration; Figure 6A). Due to their low input resistance, FS cells in both WT and GlobalLis can sustain large current injections before action potentials enter into a depolarization block, with the vast majority of cells (70-77%) fully retaining the ability to fire at high frequencies (>250Hz) even after a maximal current injection of 1500 pA (Figure 6B). In contrast, 100% of NFS in both WT and GlobalLis mice and 92% of GlobalLis IS cells were driven into depolarizing block, and required less current to do so (respectively 690±110 pA, 700±200 pA, 990±80 pA) than the 23-30% of FS cells that we did observe to enter depolarizing block (WT: 1230±180 pA, GlobalLis: 1280±110 pA; Figure 6B).
In our previous study (Fleck et al., 2000), we demonstrated that the CA1 hippocampus in LIS1 mutant mice had a lowered threshold for electrographic events in the 8.5mM K+ mouse model of epilepsy (Traynelis & Dingledine, 1988). In this model a modest 5mM elevation of extracellular K+ promotes cellular depolarization coupled to a change in the EK of +23mV. Previously we did not explore the cellular underpinnings for this change in seizure threshold but now consider that the change in extracellular K+ may drive the expanded noncanonical PV+INT network into a more depolarized state that reduces their propensity for sustained action potential activity. Therefore, we revisited this mouse model of epilepsy to determine whether either hippocampal pyramidal cells or PV+INTs enter depolarization block under conditions of elevated extracellular K known to drive electrographic activity.
We recorded evoked and spontaneous action potentials in WT and GlobalLis PV+INTs and pyramidal neurons during a 15-minute exposure period to elevated extracellular [K+] (8.5 mM; Figure 6C). We found that no recorded WT FS cells (0/10) and only 1/7 GlobalLis FS cells (14%) lost the ability to sustain repetitive action potential firing in the face of elevated [K+]. In contrast, 25% of IS cells (2/8), and 33% of both WT (1/3) and GlobalLis NFS cells (1/3) entered depolarization block following exposure to elevated [K+]. No WT (0/5) or GlobalLis (0/4) hippocampal pyramidal neurons lost the capability to repetitively fire action potentials in 8.5 mM [K+] (Figure 6D). Taken together, these data demonstrate that relative to FS cells, IS and NFS PV+INTs have a strong propensity to enter depolarization block in response to elevated [K+] and suprathreshold depolarizing current injection.
Single-cell nuclear RNAsequencing (snRNA-seq) reveals molecular changes in response to LIS1 loss
Finally, we used snRNA-seq sequencing to understand how LIS1 haploinsufficiency impacts gene expression in the subtypes of PV+INTs. GlobalLis mice were crossed to PVCre;Sun1-GFP mice to enable targeting of single PV+INT nuclei. We microdissected hippocampi of p20 WT and GlobalLis mice, lysed the cells, and performed snRNA-seq on fluorescence-sorted GFP+ PV+INT nuclei (Figure 7A). As a first pass, to establish the identities of the PV+ subtypes, we integrated and aligned this dataset with other established single-cell RNAseq profiles of (i) a publicly available pan-GABAergic Allen Brain Institute mouse dataset (Tasic et al., 2018), and (ii) Nkx2.1-cre, MGE-derived cortical and hippocampal interneurons (Mahadevan et al., 2020) using Seurat v3 (Butler et al., 2018, Stuart et al., 2019). We assigned the putative identities of Gad1+ Pvalb+ subtypes as fast-spiking basket cells (BC), axo-axonic cells (AAC) or bistratified (BiC) subtypes, using the marker expressions Tac1, Pthlh or Sst respectively, and these marker expressions mapped well with the reference datasets indicating a high confidence in the quality of snRNAseq (Figure S6-7). Focusing on the PV+INTs from WT and GlobalLis datasets (1781 and 1623 single-nuclei respectively) for subsequent analysis, we first observed that the entirety of the UMAP space aligns well between WT and GlobalLis (Figure 7Bi), and the cell recovery numbers of the BC, AAC and BiC subtypes matches well between the genotypes (Figure 7Bii). This indicates that LIS1 haploinsufficiency does not lead to gross differences in the fundamental transcriptomic identities or the overall cell recoveries of PV+INT subtypes (Figure 7B).
Because we observed robust disruption in the morphophysiological development and laminar positioning of GlobalLis PV+INTs, we examined the full range of transcriptional impairments triggered by LIS1 haploinsufficiency in PV+INT subtypes, by performing differential gene expression testing. At a stringent false-discovery rate (FDR) <0.01, 376 genes passed the 10%-foldchange (FC) threshold across the PV+INT subtypes of which, 126 genes were commonly differentially expressed (DE) between the subtypes and the remaining 250 genes were uniquely DE across the subtypes (Figure 7C; Table 3). To assess the broad biological impact of the DE genes (DEG), we applied the Ingenuity Pathway Analysis (IPA) framework. These analyses revealed that the DEGs primarily serve to regulate synaptogenesis signaling pathways, glutamatergic/GABAergic neurotransmission, different distinct guidance cues, cell-cell adhesion, and maintenance of extracellular matrix (ECM; Figure 7D; Table 4).
First, we observed misregulated expressions of cell-adhesion molecules (CAMs) belonging to cadherin family (Cdh8-13), contactin and related family (Cntn3-6, Cntnap3,5b), IgCAM family (Alcam, Dscam, Dscaml1, Ncam2, Kirrel1,3, Igsf11), including multiple CAM-modifiers (St3gal4, St6gal1) (Figure 7E). It is notable that multiple members of the CAM family are established regulators of interneuron synapse assembly, axonal and dendritic arborization (Brennaman and Maness 2008, Brennaman et al., 2013, Guan and Maness 2010, Gomez-Climent et al., 2011, Krocher et al., 2014, Gao et al., 2018). Next, we also observed several members of signaling pathways belonging to netrin family (Ntng1, Unc5b, Dcc, Ntn4), ephrin family (Efna5, Epha3, Epha4), robo family (Slit1, Slit3) and semaphorin family (Sema5a, Sema6a, Neto1, Neto2), are robustly downregulated subsequent to LIS1 haploinsufficiency (Figure 7F), many of which have well-defined functions in MGE-derived interneuron migration and morphological development (Andrews et al., 2006, 2008, Tran et al, 2007, van den Berghe et al., 2013, Steinecke et al., 2014). This collectively indicates that multiple cellular mechanisms can converge towards regulating PV+INT morphology in a LIS1-dependant cell-autonomous manner.
Based on the LIS1 haploinsufficiency associated changes in PV+INT firing properties and propensity for depolarization block we further scrutinized our transcriptome datasets for cell-autonomous changes that could impact membrane excitability. Several critical regulators of PV+INT excitability (Pelkey et al., 2017) including members of potassium channel family (Kcnj3, Kcnq5, Kcnh7, Kcnb2, Kcnk2, Kcnip1, Kcnip2) were misregulated in GlobalLis PV+INTs (Figure 7G; Table 3). In addition, two GlobalLis PV+INT clusters displayed altered levels of the key neurotransmitter release regulator Syt2 (Table 3), potentially reflecting the increased contribution of IS PV+INTs which display reduced unitary amplitudes and release probability in combination with longer latencies. We also observe that the chloride loader NKCC1 (Slc12a2) appears to be decreased in subset of basket cells which might also influence somatic chloride extrusion during high-frequency firing in PV+INTs, thereby influencing PV+INT excitability (Otsu et al., 2020). Finally, we observed altered expression patterns of glutamate receptors and associated auxiliary subunits (Grik1-4, Neto1, Neto2) (Figure 7H) that are previously established to regulate excitatory recruitment of PV+INTs (Pelkey et al., 2017, Christensen et al., 2004, Wyeth et al., 2017, Mulle et al., 2000).
CONCLUSIONS
LIS1 haploinsuffiency disrupts numerous genetic programs controlling PV+INT migration, morphogenesis, synapse formation and cellular excitability. We report that global heterozygous LIS1 loss (GlobalLis) and heterozygous loss within interneurons (NkxLis) but not within pyramidal cells (EmxLis) results in the emergence of the non-canonical “intermediate-spiking” (IS) subtype of PV+INTs, despite the disruption of radial migration observed in all three genotypes.
While WT (and EmxLis) FS cells consist of four morphological subtypes (BC, AAC, BiC, RTC) and NFS cells consisted of two subtypes (BC, RTC), clear morphology of GlobalLis and NkxLis PV+INTs could not be resolved. Interestingly, the emergence of IS cells and alteration of morphology in the GlobalLis mutant did not correlate with any additional PV+INT clusters in the snRNA-seq analysis. However, the snRNA-seq presents only a snapshot of the transcriptomic aberrations in PV+INTs due to Global LIS1 haploinsufficiency. Therefore, the lack of a distinct IS population in the snRNA-seq analysis could imply that alterations of a very small number of genes are responsible for changes, which would not result in any additional clustering. It is possible that the cellular changes observed may arise from epigenetic changes, changes in protein expression levels, or changes in modifications of proteins that would not be detectible in this dataset. A third possibility is that changes during development alter the circuit integration and maturation of PV+INTs and these changes might not be evident in their mature transcriptome. Finally, it is possible that the morphological aberrations in PV+INTs are further shaped by the transcriptomic aberrations in other cell types. Because LIS1-disruption in pyramidal cells does not seems to affect PV+INT morphological maturation, future studies could examine LIS1-disruption exclusive to glial cells to examine such non-autonomous mechanisms of regulation of interneuron functions. For example, glial cells that are also impacted by LIS1 haploinsufficiency, also secrete several guidance cues that shape the assembly, lamination and morphology of interneurons (Vallee and Tsai 2006, Yokota et al., 2007). Therefore, such cell-autonomous and non-autonomous mechanisms could converge to regulate PV+INT morphological maturation.
We observed that both WT and EmxLis CA1 are composed of 79-80% FS cells (preferentially found in deep layers) and 19-20% NFS cells (preferentially in superficial layers). In contrast, FS cells make up only 44% of GlobalLis and 29% of NkxLis PV+INTs, with the majority (56-71%) consisting of IS/NFS (which are also preferentially found in superficial layers). Compared to FS cells, IS and NFS cells have lower firing rates and provide less potent, less reliable inhibitory output to pyramidal cells. Thus, the expansion of these non-canonical cell types will erode the precise inhibition usually provided by FS cells, placing network excitability on the precipice for electrographic seizure activity. Finally, due to their propensity to enter depolarization block, IS and NFS cells will likely lose the ability to initiate action potentials and control runaway excitation during seizure episodes, allowing epileptic activity to spread from the hippocampus to other brain regions. Until now, the disrupted neuronal migration in classical lissencephaly has been paradigmatically linked to the generation of epileptiform activity, however our EmxLis and NkxLis experiments challenge this assumption. We propose that it is not disrupted neuronal migration perse that underlies the origin of epilepsy, but rather disruptions to intrinsic developmental programs that result in the emergence of a less efficient PV+INT cell type.
METHODS
Animals
All experiments were conducted in accordance with animal protocols approved by the National Institutes of Health. LIS1+/fl male mice (provided by Anthony Wynshaw-Boris, Case Western Reserve University) were crossed with Sox2-Cre female mice (provided by National Human Genome Research Institute transgenic core, Tg(Sox2-Cre)1Amc/J). Sox2-Cre females exhibit Cre-recombinase activity in gamete tissues, which allow for genotyping and selection of non-conditional LIS1+/− mutants without the Cre allele in a single cross. These mice were bred to wild-type C57BL/6J mice (Jackson Labs stock no. 00064) to maintain global LIS1+/− colonies. To obtain cell-type-specific LIS1 mutations, we crossed LIS1+/fl mice to Nkx2.1-Cre (Jackson Labs stock no. 008661, C57BL/6J-Tg(Nkx2-1-cre)2Sand/J) and Emx1-Cre (Jackson Labs stock no. 005628, B6.129S2-Emx1tm1(cre)Krj/J) lines.
To enable genetic access and targeting of PV+ cells, Lis1 mutant lines were crossed to PV-tdTomato reporters. LIS1+/− mice were crossed to PV-Cre (Jackson Labs stock no. 017320, B6.129P2-Pvalbtm1(cre)Arbr/J) and tdTomato mice (Jackson Labs stock no. 007909, B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J). LIS1fl/+:Nkx2.1Cre and Lis1fl/+::Emx1Cre lines were crossed to Cre-independent PV-TdTom mice (Jackson Labs stock no. 027395, Tg(Pvalb-tdTomato15Gfng). For single-cell nuclear RNAseq experiments, Lis1+/−:PV-Cre mice were crossed to Sun1-GFP mice (Jackson Labs stock no. 030952, B6.129-Gt(ROSA)26Sortm5(CAG-Sun1/sfGFP)Nat /MmbeJ).
Male and female mice from p19-p60 were used. Mice were housed and bred in a conventional vivarium with standard laboratory chow and water in standard animal cages under a 12 h circadian cycle.
Immunohistochemistry on perfused tissue
Mice were deeply anesthetized, and tissue was fixed via transcardial perfusion with 30mL of phosphate-buffered saline (PBS) followed by 50 mL of 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB, pH 7.6). Brains were post-fixed overnight at 4°C when processed for immunostaining for PV and NeuN. Brains were cryopreserved in 30% sucrose and sectioned on a freezing microtome at 50 μm. Sections were rinsed in PB, blocked for 2 hours in 10% normal goat serum with 0.5% Triton X-100, and then incubated in primary antibody for 2 hours at room temperature or overnight at 4°C. Sections were then rinsed with PB and incubated in secondary antibodies (1:1000) and DAPI (1:2000) for 2 hours at room temperature. All antibodies were diluted in carrier solution consisting of PB with 1% BSA, 1% normal goat serum, and 0.5% Triton X-100. Sections were then rinsed, mounted on Superfrost glass slides, and coverslipped using Mowiol mounting medium and 1.5 mm cover glasses.
Image acquisition and analysis
Confocal images were taken using a Zeiss 780 confocal microscope. For all slices with immunostained or genetically reported somatic signal, 50 μm thin sections were imaged using a Nikon spinning disk (Yokogawa CSU-X) confocal microscope. Counting was performed on four hippocampal sections from each animal. Quantitative analysis of PV+ cell density in each CA1 layer was performed using ImageJ software (NIH, Bethesda, MD, USA).
Slice preparation
Mice (p19-p60) were anesthetized with isoflurane and then decapitated. The brain was dissected out in ice-cold sucrose artificial cerebrospinal fluid (aCSF) containing the following (in mM): 130 NaCl, 3.5 KCl, 24 NaHCO3, 1.25 NaH2PO4, 1.5 MgCl2, 2.5 CaCl2, and 10 glucose, saturated with 95% O2 and 5% CO2. Mice older than p30 were dissected in sucrose-substituted artificial cerebrospinal fluid (SSaCSF) containing the following (in mM): 90 sucrose, 80 NaCl, 3.5 KCl, 24 NaHCO3, 1.25 NaH2PO4, 4.5 MgCl, 0.5 CaCl2, and 10 glucose, saturated with 95% O2 and 5% CO2. Coronal hippocampal slices (300 μm) were cut using a VT-1200S vibratome (Leica Microsystems) and incubated submerged in the above solution at 32-34°C for 30 min and then maintained at room temperature until use. Slices were incubated for at least 45 minutes before conducting electrophysiological recordings.
Whole-cell electrophysiology
For patch-clamp recordings following recovery slices were transferred to an upright microscope (Zeiss Axioskop), perfused with aCSF (with or without SCZD as indicated) at 2-3 ml/min at a temperature of 32°C-34°C. Individual cells were visualized using a 40x objective using fluorescence and IR-DIC video microscopy. Electrodes were pulled from borosilicate glass (World Precision Instruments) to a resistance of 3–5 MΩ using a vertical pipette puller (Narishige, PP-830). Whole-cell patch-clamp recordings were made using a Multiclamp 700B amplifier (Molecular Devices), and signals were digitized at 20 kHz (Digidata 1440A, filtered at 3 kHz) for collection on a Windows computer equipped with pClamp 10.4 software (Molecular Devices). Uncompensated series resistance ranged from 10 to 35 MΩ and was monitored continuously throughout recordings with −5 mV voltage steps. For current-clamp and voltage-clamp recordings of PV-Cre-tdTom+ interneurons and CA1 pyramidal cells, two different internal solutions containing (in mM) were used: A) 130 K-gluconate, 5 KCl, 10 HEPES, 3 MgCl2, 2 Na2ATP, 0.3 NaGTP, , 0.6 EGTA, and 0.2% biocytin (calculated chloride reversal potential (ECl-) of −67 mV) or B) 130 K-Gluconate, 10 KCl, 10 HEPES, 3 MgCl2, 2 Na2ATP, 0.3 NaGTP, 0.6 EGTA and .2% biocytin (calculated ECl- = −27 mV).
Input resistance (Rin) was measured using a linear regression of voltage deflections in response to 500 ms-long current steps of four to six different amplitudes (−200 to +50 pA, increments of 50 pA). Membrane time constant (τm) was calculated from the mean responses to 20 successive hyperpolarizing current pulses (−20pA; 400 ms) and was determined by fitting voltage responses with a single exponential function. Action potential (AP) threshold was defined as the voltage at which the slope trajectory reaches 10 mV/ms. AP amplitude was defined as the difference in membrane potential between threshold and the peak. AP half-width was measured at the voltage corresponding to half of the AP amplitude. Afterhyperpolarization (AHP) amplitude was defined as the difference between action potential threshold and the most negative membrane potential attained during the AHP. These properties were measured for the first action potential elicited by a depolarizing 500 ms-long current pulse of amplitude just sufficient to bring the cell to threshold for AP generation. The adaptation ratio was defined as the ratio of the average of the last 3 interspike intervals relative to the first 3 interspike intervals during a 500 ms-long spike train elicited using twice the current injection necessary to obtain a just suprathreshold response. Firing patterns were investigated by a series of 500 to 800 ms-long current injections (step size 50 pA) until 3x threshold current was reached or depolarization-block was induced. Firing frequency was calculated from the number of spikes observed during the first 500 ms of the spike train. Firing frequency at 3x threshold was substituted by maximum firing frequency in cells with depolarization-block. Ihyp Sag index of each cell was determined by a series of 500 ms-long negative current steps to create V-I plots of the peak negative voltage deflection (Vhyp) and the steady-state voltage deflection (average voltage over the last 200 ms of the current step; Vsag) and used the ratio of Vrest−Vsag/Vrest−Vhyp for current injections corresponding closest to Vsag=−80mV.
For PV+INT-CA1PC paired recordings, the presynaptic PV+INT was held in current clamp with membrane potential biased to −70 mV, while the postsynaptic cell was held voltage clamp at −70 mV. Synaptic transmission was monitored by producing action potentials in presynaptic PV+INTs (held in current-clamp around −70 mV) every 10s by giving 2 ms 1-2 nA current steps. Presynaptic trains to probe unitary transmission dynamics consisted of 25 presynaptic action potentials at 50 Hz. Basal unitary event properties for each cell were analyzed using 10 consecutive events obtained 4-5 minutes after establishing the postsynaptic whole-cell configuration. Amplitudes reflect the average peak amplitude of all events including failures, potency is the average peak amplitude excluding failures. Decay kinetics were measured by single exponential fit of uIPSC potency. The latency of synaptic transmission was defined as the time from the peak of the AP to 5% of the uIPSC potency.
For depolarization block experiments, WT and GlobalLis TdT+PV+INTs were recorded in current-clamp mode. Rheobase, input resistance, AP half-width, firing frequency at 2 threshold x and maximum firing frequency were recorded (and used to identify physiological subtype) using above protocols. Further depolarizing current was injected in 500 ms sweeps with a step size of 50 pA until cells were subject to depolarizing block or 1500 pA of current were injected. Following this, 10 second sweeps of spontaneous activity with 500 ms of evoked firing (with ~1.5x threshold current) were taken. After obtaining a stable baseline (~10 sweeps), the extracellular solution was switched to a solution with 8.5 mM [K+] the same 10 second (500 ms evoked firing) sweeps were taken for up to 15 minutes.
Clustering Analysis
To identify potential subclusters of PV+INTs, we performed principal components analysis (PCA) and hierarchical clustering based on Euclidean distance of normalized (log transformed) intrinsic electrophysiological parameters using R-studio version 0.99.451 and R version 3.4.2.
Anatomical Reconstructions
After biocytin filling during whole-cell recordings, slices were fixed with 4% paraformaldehyde and stored at 4°C then permeabilized with 0.3 % Triton X-100 and incubated with Alexa Fluor 488 or Alexa Fluor 555-conjugated streptavidin. Resectioned slices (75 um) were mounted on gelatin-coated slides using Mowiol mounting medium. Cells were visualized using epifluorescence microscopy and images for representative examples were obtained with confocal microscopy. Cells were reconstructed and analyzed with Sholl analysis using Neurolucida software (MBF Bioscience).
Statistical Analysis
All data were tested for normality and then tested with parametric or nonparametric t-test or ANOVA, as appropriate (Graphpad Prism). Quantification and error bars display standard error of the mean. Intrinsic electrophysiological parameters in the text are values for the upper and lower 95% confidence intervals of the mean. Values shown for unitary synaptic transmission properties consist of mean and standard error of the mean.
Single nucleus isolation
Hippocampus from seven mutant (Lis+/−;PV-Cre+/−;Sun1-GFP+/−) and six WT (Lis−/−;PV-Cre+/−; Sun1-GFP+/−) P21 mice were quickly dissected in ice-cold DPBS, immediately frozen on dry ice and stored at −80°C. We pooled mutant or WT hippocampus into a Dounce Homogenizer containing 1 mL freshly prepared ice-cold lysis buffer (low sucrose buffer with 1 mM DTT, 0.1% NP-40), applying 10 strokes with pestle A followed by 10 strokes with the pestle B. The homogenate was filtered through a 40 μm cell strainer, transferred to a DNA low bind 2 mL microfuge tube and centrifuged at 300 g for 5 min at 4°C. The supernatant was removed, the pellet was gently resuspended in a low sucrose buffer (320 mM sucrose, 10 mM HEPES-pH 8.0, 5 mM CaCl2, 3 mM Mg-acetate, 0.1 mM EDTA) and centrifuged for another 5 min. The nuclei were resuspended in 500 μl 1xPBS with 1% BSA and 0.2 U/μl SUPERaseIn RNase Inhibitor (ThermoFisher, #AM2696) and loaded on top of 900 μl 1.8 M Sucrose Cushion Solution (Sigma, NUC-201). The sucrose gradient was centrifuged at 13,000 g for 45 min at 4°C. The supernatant was discarded, and the nuclei were resuspended in 500ul Pre-FACS buffer (1xPBS with 1% BSA, 0.2 U/μl SUPERaseIn RNase Inhibitor and 0.2 M sucrose). Before sorting, nucleus from the six WT or seven mutant mice were pooled together and 5ul of 5 mM DRAQ5 were added. Samples were processed on a Sony SH800 Cell Sorter with a 100 mm sorting chip. 15,000 GFP+/DRAQ5+ nuclei from mutant and WT samples were collected directly into 1.5 ml centrifuge tubes containing 10ul of the Pre-FACS buffer. PCR cycles were conducted for cDNA amplification, and the subsequent library preparation and sequencing were carried out in accordance with the manufacturer recommendation (Chromium™ Single Cell 3’ Library & Gel Bead Kit 10X v3, 16 reactions). Sequencing of the libraries were performed on the Illumina HiSeq2500 at the NICHD, Molecular Genomics Core facility. The cell number estimates, mean reads per cell (raw), median genes per cell respectively, are as follows LIS1+/+: 8470, 22,289, 2024; LIS1+/−: 8185, 24,652, 2142. Demultiplexed samples were aligned to the mouse reference genome (mm10). The end definitions of genes were extended 4k bp downstream (or halfway to the next feature if closer) and converted to mRNA counts using the Cell Ranger Version 2.1.1, provided by the manufacturer.
RNAseq data processing & analyses, differential expression testing, and visualization
Processing (load, align, merge, cluster, differential expression testing) and visualization of the scRNAseq datasets were performed with the R statistical programming environment (v3.5.1) (R Core Team 2013), and Seurat package (v3.1.5) (Butler et al., 2018, Stuart et al., 2019). The LIS1+/+ and LIS1+/− datasets were first merged with the Allen Institute reference and in-house MGE interneuron reference datasets. To analyze the Allen Institute mouse dataset of the single-cell transcriptomes of ~76,000 cells from >20 areas of mouse cortex and hippocampus, we downloaded the transcriptome/HDF5 file (https://portal.brain-map.org/atlases-and-data/rnaseq) and subsequently converted into Seurat v3-compatible format based on the instructions provided in the Allen Institute Portal (https://portal.brain-map.org/atlases-and-data/rnaseq/protocols-mouse-cortex-and-hippocampus) and custom scripts in R package as previously described (Chittajallu et al., 2020). Single-cell transcriptomes from Nkx2.1-cre:Ai14, MGE-derived cortical and hippocampal interneurons (postnatal day 18-20) were processed as previously described (Mahadevan et al., 2020). To perform integrated analyses, we identified a common set of genes between LIS1+/+, LIS1+/−, Nkx2.1-MGE cortical and hippocampal interneurons and Allen datasets, and utilized these for the initial analyses in Figure S6-7. Data set preprocessing, normalisation, identification of variable genes, canonical correlation analyses were performed according to default Seurat parameters, unless otherwise mentioned. Quality control filtering was performed by only including cells that had between 200-20000 unique genes, and that had <5% of reads from mitochondrial genes. Clustering was performed on the top 25 PCs using the function FindClusters() by applying the shared nearest neighbor modularity optimization with clustering resolution of 0.5. Phylogenetic tree relating the ‘average’ cell from each identity class based on a distance matrix constructed in gene expression space using the BuildClusterTree() function. Overall, we identified 36 clusters using this approach, among which clusters 12, 13, 16, 17 are highly enriched in the LIS1+/+, LIS1+/−, datasets and aligned well in their corresponding UMAP spaces with the reference datasets. The identities of clusters 12, 13, 16, 17 are matched with the top gene markers identified by the FindAllMarkers(). These 4 clusters are Gad1+, Pvalb+ and Vip−, and were putatively annotated as Bistratified.1/2, Fast-spiking and Axo-axonic subsets of PV+INTs, based on marker expression of Sst, Tac1 and Pthlh respectively, as indicated in interneuron literature and previous scRNAseq studies (Fishell and Kepecs, 2020; Hodge et al., 2019; Paul et al., 2017; Pelkey et al., 2017; Saunders et al., 2018; Tasic et al., 2016, 2018; Yao et al., 2020; Harris et al., 2018). Subsequent to dataset validation using references, the LIS1+/+ and LIS1+/− datasets reanalyzed by subsetting the cells that expressed Gad1, Pvalb expression > 0.1 and by excluding the cells containing non-PV+INT genes Slc17a7, Ttr, Scn3a, Gpc5, Slc1a2, Htr2c, Trpm3 expressions < 0.1. Clustering was performed on the top 25 PCs using the function FindClusters() by applying the shared nearest neighbor modularity optimization with clustering resolution of 0.5. Similar to prior analyses, we recovered Bistratified, Fast-spiking and Axo-axonic subsets of PV+INTs. Additionally, a minor population of Nos1+ cells clearly segregated, which represents putative dentate gyrus-expressed PV+INTs (Vaden and Gonzalez et al., 2020, Shen et al., 2019, Jinno and Kosaka 2002).
Differential gene expression testing were performed using the MAST package within the FindMarkers() function to identify the differentially expressed genes between two subclusters (Finak et al., 2015). We applied a stringent false-discovery rate <0.01, and minimum logFC in our DEGs as ±0.1, since MAST has been previously reported to underestimate the magnitude of fold change (Ximerakis et al., 2019, Mahadevan et al., 2020). Moreover, previous studies have demonstrated the MAST approach for DEG testing to be powerful in determining subtle changes in highly transcribed genes, and among abundant populations, additional to underrepresenting changes among weakly transcribed genes (Finak et al., 2015; Ximerakis et al., 2019). Molecular and functional annotation of the DEGs were conducted using Ingenuity Pathway Analyses platform, to identify the biological pathways and disease pathways over-represented. IPA was also used to annotate genes with their known cellular functional classes. Heatmaps for the DEGs were generated using the Morpheus package (https://software.broadinstitute.org/morpheus) within the R framework.
Competing Interests
None.
Figure 8. LIS1-signaling regulates the expressions of members of axon guidance and neuronal excitability pathways. Heatmap of log2 FC of significant DEGs in each PV+INT subtype, showing a subset of A, genes regulating axon guidance; B, postsynaptic glutamate receptor subunits and associated auxiliary subunits; C, other miscellaneous regulators of neuronal excitability.
Figure Supp 1. PV-TdTomato/PV-IHC colocalization. GlobalLis mutant hippocampus. Top image displays PV-IHC, middle displays PV-TdTomato, bottom displays superimposition, in a representative GlobalLis hippocampus. There is high level of colocalization: most TdTomato+ cells were PV-IHC+ and most PV-IHC+ cells were PV-TdTomato+.
Figure Supp 2. Morphological analysis of PV+INT physiological subtypes. A Reconstructions of FS (left), IS (middle) and NFS (right) PV+INTs. Concentric circles (superimposed in black) with a radius interval of 50 um were used for Sholl analysis. B Plots of Sholl intersections in WT FS/NFS cells at 50 um intervals for dendrites (left) and axon (right). C Plots of Sholl intersections in GlobalLis FS/IS/NFS cells at 50 um intervals for dendrites (left) and axon (right). D Quantification of the total number of Sholl intersections for FS/IS/NFS cells independent of genotype (“total”) and by genotype (WT or GlobalLis) for dendrites (left) and axons (right).
Figure Supp 3. Cluster analysis and morphophysiology of EmxLis PV+INTs. A Cluster plot (left) and dendrogram (right) of 38 EmxLis PV+INTs. The dendrogram inset indicates 2 optimal clusters which correspond to FS and NFS cells. B Due to similar sorting, WT and EmxLis datasets were analyzed together, which resulted in identical classification of every FS/NFS cell. C Percentages of FS/NFS cells in CA1 and each sublayer. For reference, the WT distributions (left) are repeated alongside the EmxLis distributions (right). D Examples of EmxLis PV+INTs reconstructed cells (dendrite in black, axon in blue) and firing traces on bottom (as described in Fig. 2). Cell types from left to right: FS basket cell, FS basket cell, FS bistratified cell, NFS radiatum-targeting cell.
Figure Supp 4. Cluster analysis and morphophysiology of NkxLis PV+INTs. A Cluster plot (left) and dendrogram (right) of 41 NkxLis PV+INTs. The dendrogram inset indicates 3 optimal clusters which correspond to FS and NFS cells. B Due to similar sorting, as well as the obvious misclassification some cell types, NkxLis and GlobalLis datasets were analyzed together, which resulted in more reliable classification cell types. C Percentages of FS/IS/NFS cells in CA1 and each sublayer. For reference, the GlobalLis distributions (left) are repeated alongside the NkxLis distributions (right). D Examples of NkxLis PV+INTs reconstructed cells (dendrite in black, axon in blue) and firing traces on bottom (as described in Fig. 2). Physiological cell types from left to right: FS, FS, IS, NFS.
Figure Supp 5. Membrane, firing, and morphological properties of PV+INT physiological subtypes. Frequency distributions of physiological and morphological properties PV+INTs. FS cells are shown in blue, IS cells in purple, NFS cells in pink. The full color circles connecting lines display the average of all genotypes for each physiological subtype. The faded shapes/lines represent distributions for WT (hexagons), GlobalLis (squares), NkxLis (triangles), and EmxLis (diamonds). Note that with the exception of AP half-width, there were no significant differences between genotypes. Almost every property shown here had significant differences between physiological subtype, independent of genotype. See also Table 1.
Figure Supp 6. Integrated analyses of single-cell/nucleus transcriptomes from PV+INTs from Lis1+/+, Lis1+/−, and reference datasets. A UMAP plots of the PV+INTs from Lis1+/+, Lis1+/−, cortical and hippocampal MGE-derived interneurons and pan-GABAergic Allen Institute datasets indicating robust alignments of the PV+INT clusters. UMAP plots representing the expression of B, Gad1 and C, Pvalb across all datasets.
Figure Supp 7. Annotation of PV+INT subtypes based on discrete marker gene expressions. A Single cell heatmap showing scaled expression values for the top 10 transcriptomic markers in each of the PV+INTs clusters across Lis1+/+, Lis1+/−, and reference datasets. B UMAP plots depicting the bistratified, axo-axonic and fast-spiking clusters across all datasets. C Merged UMAP plots representing the expression of top transcriptomic markers across the PV+INT subtypes across all datasets.
Acknowledgements
We thank Dr. Carolina Bengtsson-Gonzales for developing the code for cluster analysis. We thank Drs. Kenneth Pelkey and Ramesh Chittajallu for discussing project ideas and technical details of experiments. We thank Dr. Gülcan Akgül for developing LIS1 primers for genotyping. We thank Steven Hunt for genotyping and cell processing for microscopy. We thank Daniel Abebe for animal support. We thank Dr. Vincent Schram for confocal microscopy support. We thank Drs. Steven L. Coon, Tianwei Li and James R. Iben at the Molecular Genomics Core, NICHD, for RNA sequencing and bioinformatics support. We thank Drs. Apratim Mitra and Ryan Dale (NICHD Bioinformatics and Scientific Programming Core) for assistance with integrated analysis using the Allen Brain scRNAseq dataset. These analyses utilized the computational resources of the NIH HPC Biowulf cluster (http://hpc.nih.gov). We thank Dr. Anthony Wynshaw-Boris for providing the heterozygous floxed LIS1 mouse.