Document [original]

TYPE Hypothesis and Theory

PUBLISHED 04 May 2023

DOI 10.3389/fncir.2023.1172464

OPEN ACCESS

EDITED BY

Gabrielle Pouchelon,

Cold Spring Harbor Laboratory, United States

REVIEWED BY

Daniel Vogt,

Michigan State University, United States

Maximiliano Jose Nigro,

Norwegian University of Science and

Technology, Norway

*CORRESPONDENCE

Joram Keijser

[email protected]

Henning Sprekeler

[email protected]

RECEIVED 23 February 2023

ACCEPTED 30 March 2023

PUBLISHED 04 May 2023

CITATION

Keijser J and Sprekeler H (2023) Cortical

interneurons: ﬁt for function and ﬁt to

function? Evidence from development and

evolution. Front. Neural Circuits 17:1172464.

doi: 10.3389/fncir.2023.1172464

open-access article distributed under the terms

of the Creative Commons Attribution License

(CC BY). The use, distribution or reproduction

in other forums is permitted, provided the

original author(s) and the copyright owner(s)

are credited and that the original publication in

this journal is cited, in accordance with

accepted academic practice. No use,

distribution or reproduction is permitted which

does not comply with these terms.

Cortical interneurons: ﬁt for

function and ﬁt to function?

Evidence from development and

evolution

Joram Keijser1,2*and Henning Sprekeler1,3*

1Modelling of Cognitive Processes, Technical University of Berlin, Berlin, Germany, 2Einstein Center for

Neurosciences, Charité University Medicine Berlin, Berlin, Germany, 3Bernstein Center for

Computational Neuroscience Berlin, Humboldt University of Berlin, Berlin, Germany

Cortical inhibitory interneurons form a broad spectrum of subtypes. This diversity

suggests a division of labor, in which each cell type supports a distinct function.

In the present era of optimisation-based algorithms, it is tempting to speculate

that these functions were the evolutionary or developmental driving force for

the spectrum of interneurons we see in the mature mammalian brain. In this

study, we evaluated this hypothesis using the two most common interneuron

types, parvalbumin (PV) and somatostatin (SST) expressing cells, as examples.

PV and SST interneurons control the activity in the cell bodies and the apical

dendrites of excitatory pyramidal cells, respectively, due to a combination of

anatomical and synaptic properties. But was this compartment-speciﬁc inhibition

indeed the function for which PV and SST cells originally evolved? Does the

compartmental structure of pyramidal cells shape the diversiﬁcation of PV and

SST interneurons over development? To address these questions, we reviewed

and reanalyzed publicly available data on the development and evolution of PV

and SST interneurons on one hand, and pyramidal cell morphology on the other.

These data speak against the idea that the compartment structure of pyramidal

cells drove the diversiﬁcation into PV and SST interneurons. In particular, pyramidal

cells mature late, while interneurons are likely committed to a particular fate (PV

vs. SST) during early development. Moreover, comparative anatomy and single cell

RNA-sequencing data indicate that PV and SST cells, but not the compartment

structure of pyramidal cells, existed in the last common ancestor of mammals

and reptiles. Speciﬁcally, turtle and songbird SST cells also express the Elfn1 and

Cbln4 genes that are thought to play a role in compartment-speciﬁc inhibition in

mammals. PV and SST cells therefore evolved and developed the properties that

allow them to provide compartment-speciﬁc inhibition before there was selective

pressure for this function. This suggest that interneuron diversity originally resulted

from a different evolutionary driving force and was only later co-opted for the

compartment-speciﬁc inhibition it seems to serve in mammals today. Future

experiments could further test this idea using our computational reconstruction

of ancestral Elfn1 protein sequences.

KEYWORDS

inhibition, interneuron, evolution, development, microcircuits, single cell RNA seq, neural

morphology, pyramidal cell dendrites

Frontiers in Neural Circuits 01 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

1. Introduction

Cortical inhibitory interneurons are a highly diverse

group, differing in their morphology, connectivity, and

electrophysiology (Tremblay et al., 2016). Decades of experimental

and theoretical work have suggested a role for interneurons in

many functions (Kepecs and Fishell, 2014;Tremblay et al., 2016;

Sadeh and Clopath, 2021), including the regulation of neural

activity (Vogels et al., 2011;Wu et al., 2022), control of synaptic

plasticity (Letzkus et al., 2015;Williams and Holtmaat, 2019),

increasing temporal precision (Wehr and Zador, 2003;Bhatia et al.,

2019), predictive coding (Keller and Mrsic-Flogel, 2018;Hertäg

and Clopath, 2022), and gain modulation (Fu et al., 2014;Ferguson

and Cardin, 2020). Many of these functions come down to the

control of excitation.

Why would the control of excitation require a diversity of

interneurons? A key reason could lie in the complexity of excitatory

cells (Fishell and Kepecs, 2020;Keijser and Sprekeler, 2022).

Pyramidal cells (PCs) consist of several cellular compartments that

have different physiological properties [e.g., sodium vs. calcium

spikes (Larkum et al., 1999)], receive different inputs [e.g., top-

down vs. bottom up (Petreanu et al., 2007;Larkum, 2013), although

see Ledderose et al., 2022] and follow distinct synaptic plasticity

rules (Letzkus et al., 2006;Sjostrom et al., 2008;Udakis et al.,

2020). The control of different pyramidal cell compartments

might therefore require inhibition from designated types of

interneurons. Indeed, the two most common interneuron types—

parvalbumin (PV)- and somatostatin (SST)-expressing cells—are

classically distinguished by their connectivity with pyramidal cells:

whereas PV-expressing basket cells mainly target the somata of

PCs, SST-expressing Martinotti cells mainly target their apical

dendrites (Tremblay et al., 2016). The cellular and synaptic

properties of these interneurons also seem adapted to this purpose.

SST interneurons receive facilitating synapses from PCs (Reyes

et al., 1998;Silberberg and Markram, 2007), rendering them

sensitive to bursts of action potentials (Goldberg et al., 2004;

Murayama et al., 2009;Berger et al., 2010) triggered by plateau

potentials in the apical dendrite of PCs (Larkum et al., 1999;

Williams and Stuart, 1999). Indeed, SST interneurons control

dendritic excitability and bursting of PCs (Murayama et al., 2009;

Gentet et al., 2012;Lovett-Barron et al., 2012). PV interneurons,

on the other hand, receive depressing synapses (Reyes et al.,

1998;Caillard et al., 2000), rendering them less sensitive to these

signals (Pouille and Scanziani, 2004). The presynaptic dynamics

of PV and SST interneurons therefore seem particularly well-

matched to the physiology of pyramidal cells, although both

types also inhibit non-pyramidal cells and other interneurons

(see e.g., Jiang et al., 2015;Campagnola et al., 2022). These

and similar observations have led to the view that interneuron

diversity can be understood from a functional perspective, in which

the morphology and synaptic and cellular properties of different

interneurons are fit to specific functions (Figure 1A) (Kepecs and

Fishell, 2014;Fishell and Kepecs, 2020). Consistent with this idea

that interneurons are adapted to control different pyramidal cell

compartments, we recently showed that properties (connectivity

and short-term plasticity) of PV and SST interneurons emerge

when optimizing a network model for compartment-specific

inhibition (Figure 1B) (Keijser and Sprekeler, 2022).

The specialization of PV and SST interneurons to pyramidal

soma and dendrites, respectively, makes it tempting to speculate

that the diversification of these interneuron subtypes was

driven by pyramidal cell properties, either during evolution

or during development (Figure 1C). This hypothesis predicts

a specific temporal order: during evolution or development,

the compartmentalization of pyramidal cells should predate

interneuron diversification (Figure 1D).

Here, we evaluate this idea, with a focus on PC and interneuron

properties that seem particularly well-adapted to each other: the

active dendrites of pyramidal cells, and the connectivity and

short-term plasticity of interneurons. Reviewing and reanalyzing

recent evolutionary and developmental data, we reconstruct the

developmental and evolutionary history of these three properties.

We find no support for the idea that interneurons develop or

evolved to control preexisting compartments of pyramidal cells.

Instead, the central properties of PV and SST interneurons that led

to this idea emerge before the PC properties they seem adapted

to, in both development and evolution. Rather than pyramidal

physiology driving interneuron diversification, this suggests a

model in which existing interneuron properties enabled new

pyramidal cell functions.

2. Developmental trajectory of

compartment-speciﬁc inhibition

We first discuss the developmental trajectory of pyramidal cells

and PV and SST interneurons in the mammalian cortex, to assess

whether the diversification of PV and SST interneurons during

development is driven by pyramidal cell properties. We mostly

consider data from rodents, but many of the findings seem to be

conserved among mammals (Hansen et al., 2013;Ma et al., 2013;

Shi et al., 2021;Schmitz et al., 2022).

In contrast to pyramidal cells, interneurons are not born in the

developing cortex, but subcortically (Anderson et al., 1997;Flames

et al., 2007) (Box 1). It is only upon migrating to the cortex that

interneurons acquire their mature morphology and physiology.

The long period between interneuron birth and maturation has

led to different models of interneuron development (Kepecs and

Fishell, 2014;Wamsley and Fishell, 2017). One model attributes

the late maturation of interneurons to a late specification of their

cellular identity, possibly based on external cues within the circuit

they embed themselves in Kepecs and Fishell (2014),Wamsley and

Fishell (2017). Alternatively, the late emergence of characteristic

features could be due to the slow unfolding of a predetermined

genetic program that happens independently of the surrounding

circuit (Lim et al., 2018a).

The malleability of interneuron properties during development

is therefore currently an open question: Which properties are

adapted to the surrounding circuit, and which are predetermined?

Whatever properties are adapted, cellular identity (e.g., PV vs. SST)

is probably not one of them (Wamsley and Fishell, 2017;Lim

et al., 2018a). Interneuron types—at least on a coarse level—are

Frontiers in Neural Circuits 02 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

FIGURE 1

Do development and evolution ﬁt interneurons to function? (A) The connectivity and short-term plasticity (STP) of PV and SST positive interneurons

seem adapted to the morphology and electrophysiology of pyramidal cells, as highlighted by an optimization-based model (B). In this model,

optimizing interneuron parameters to provide compartment- (soma/dendrite) speciﬁc inhibition causes interneurons to diversify into two groups

that resemble PV and SST interneurons in their connectivity and short-term plasticity (Keijser and Sprekeler, 2022). (C) The existence of PV and SST

subtypes might therefore result from an developmental or evolutionary tuning of interneurons based on pyramidal properties. (D) This predicts that

immature (or ancestral) circuits contain bursting pyramidal neurons and undiversiﬁed interneurons. Development (or evolution) then diversiﬁes the

interneurons into PV and SST subtypes. PC activity and short-term plasticity simulated with models from Tsodyks et al. (1998), Naud et al. (2013), and

Naud and Sprekeler (2018), respectively. Animal silhouettes from https://beta.phylopic.org/.

determined by their time and place of birth. Future PV and

SST interneurons, for example, are preferentially generated within

different parts of the same embryonic structure (Box 1) (Wonders

and Anderson, 2006;Lim et al., 2018a).

Recent data suggests that not just interneuron types (e.g., PV

vs. SST), but also interneuron subtypes (e.g., SST Martinotti vs.

SST non-Martinotti) are specified early in development. Lim et al.

(2018a) showed that Martinotti and non-Martinotti cells migrate

to the developing cortex via different routes (Box 1). In addition,

a developing interneuron’s transcriptional profile can be used to

predict its future fate (Mayer et al., 2018;Mi et al., 2018;Bandler

et al., 2021;Shi et al., 2021).

Although interneurons are therefore likely hardwired to

become a certain subtype, it is still possible that interneuron

properties such as short-term plasticity or connectivity are

subject to activity-dependent fine-tuning. For example, the

development of short-term facilitation or a layer 1 axon of SST

Martinotti cells might emerge in dependence on pyramidal neuron

bursting. In this case, bursting should develop ahead of these

SST features.

When do developing pyramidal cells first show dendrite-

dependent bursting? Their electrophysiology matures relatively

late: dendritic plateau potentials emerge only in the third postnatal

week (Franceschetti et al., 1998;Zhu, 2000). This is consistent with

the late maturation of their dendritic morphology. PCs develop

their intricate apical arborization and tuft dendrites after the second

postnatal week (Zhu, 2000;Romand et al., 2011). For example,

the tuft length increases almost twofold during the third postnatal

week (Romand et al., 2011), and dendritic spikes fail to reach the

soma on postnatal day 14 and 28 (Zhu, 2000).

When does short-term facilitation (STF) of PC→SST synapses

arise during development? Could its development be driven by

bursting in pyramidal cells? Some of the early experiments showed

such STF in rat cortex during the third postnatal week (Reyes et al.,

1998;Beierlein and Connors, 2002;Silberberg and Markram, 2007).

Evidence for an even earlier presence of STF in these synapses

comes from molecular studies. The Gosh laboratory has shown that

the short-term facilitation in hippocampal (Sylwestrak and Ghosh,

2012) and cortical (Stachniak et al., 2019) PC→SST synapses is

due to the transmembrane protein Elfn1, which is expressed by

SST mouse and human interneurons (Box 2,Figure 2). In these

experiments, STF was measured in the second postnatal week, and

the expression of Elfn1 was detected already one week after birth

(Tomioka et al., 2014;Favuzzi et al., 2019), providing an early

molecular signature of short-term facilitation in SST cells. Short-

term facilitation in PC→SST synapses is therefore present before

dendrite-dependent bursting in PCs.

What about the second difference between PV and SST

neurons, their compartment-specific output synapses? SST and

PV cells form compartment-specific synapses in visual cortical

organotypic cultures that lack external inputs (Cristo et al.,

2004). This strongly suggests a role for genetic encoding rather

than experience-dependent activity. Indeed, recent work identified

important molecular players in the establishment of compartment-

specific synapses (Favuzzi et al., 2019). Several genes are involved

in the formation of compartment-specific synapses. For example,

suppressing Cbln4 in SST interneurons decreased inhibition onto

PC dendrites. An over-expression of the same gene in PV

interneurons, on the other hand, increased inhibition onto PC

dendrites (Favuzzi et al., 2019). Other genes contribute to somatic

Frontiers in Neural Circuits 03 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

BOX 1 Birth and migration of cortical interneurons.

Cortical GABAergic interneurons are born in a transient region of the developing brain known as the ganglionic eminence (Anderson et al., 1997;Wamsley and

Fishell, 2017;Lim et al., 2018a), from where they tangentially migrate to the cortex (Marín and Rubenstein, 2001). The ganglionic eminence can be divided into

multiple subregions patterned by unique combinations of transcription factors (Flames et al., 2007;Hansen et al., 2013;Ma et al., 2013;Hu et al., 2017), that

activate distinct genetic programs. Since each genetic program corresponds to a different cell type, the majority of the cells born in the medial ganglionic eminence

(MGE) will become PV and SST interneurons, whereas the caudal ganglionic eminence (CGE) generates, among others, vasoactive intestinal peptide

(VIP)-expressing interneurons (Wichterle et al., 2001;Nery et al., 2002;Xu et al., 2004;Butt et al., 2005;Miyoshi et al., 2010). A key example for a patterning

transcription factor that shapes interneuron identity is Nkx2-1, which is expressed within the MGE but not CGE (Sussel et al., 1999;Butt et al., 2005). Nkx2-1

knockout leads MGE-derived interneurons to adopt the fate of CGE-derived interneurons (Butt et al., 2008). Molecular gradients have also been shown to

contribute to interneuron diversity within the same eminence: the dorsal-caudal and rostral-ventral MGE preferentially generate SST and PV neurons, respectively

0maturation, franceschetti1(Fogarty et al., 2007;Wonders et al., 2008;Inan et al., 2012;He et al., 2016;Hu et al., 2017;McKenzie et al., 2019).

After birth, interneurons migrate to the developing cortex via two different routes: The superficial marginal zone (the MZ, which will develop into cortical layer 1)

and the deeper subventricular zone (SVZ). These different migration routes are used by distinct layer 2–3 (L2–3) SST subtypes (Lim et al., 2018a). Whereas L2-3

SST Martinotti cells migrate via the marginal zone from where they descend to their future location, non-Martinotti cells migrate via the subventricular zone (Lim

et al., 2018b). This indicates that SST subtypes (at least in L2–3) are predetermined before their arrival in cortex, possibly even before they “choose” one migratory

route over the other. Future L2–3 Martinotti cells forced to migrate via the wrong route (the SVZ) still become Martinotti cells in terms of their transcriptional

profile and electrophysiology, but they lack a fully developed layer 1 axon (Lim et al., 2018b). This suggests that developing L2/3 Martinotti cells cannot send their

developing axon from deeper to upper layers, but have to leave it there while their cell body descends. Translaminar axons of other neurons such as a less studied

PV subtype (Lim et al., 2018b), and cerebellar granule cells (Rakic, 1971) are established via a similar mechanism, suggesting it might be the only reliable way for

neurons to develop translaminar projections.

BOX 2 Genetic basis of short-term facilitation.

Pyramidal cells form short-term depressing synapses onto PV neurons, but short-term facilitating synapses onto SST neurons. This difference is partly attributed

to the postsynaptic expression of Elfn1 by SST neurons (Sylwestrak and Ghosh, 2012;Tomioka et al., 2014;Stachniak et al., 2019). Elfn1 is a synaptic protein that

contacts the presynaptic boutons of pyramidal cells and controls their release properties. Specifically, Elfn1 induces presynaptic localization of metabotropic

glutamate receptor 7 (mGluR7) (Tomioka et al., 2014). Grm7, the gene coding for mGluR7, is near-ubiquitously expressed in mouse (and human) neurons (data

from Tasic et al., 2018;Bakken et al., 2021). mGluR7 has a low affinity for glutamate: only high glutamate levels caused by repeated presynaptic stimulation will

lead mGluR7 to activate calcium channels, which increase synaptic release and thereby mediate synaptic facilitation. Elfn1 causes facilitation of PC→SST synapses

in the hippocampus and different cortical layers (Stachniak et al., 2019). As expected from their expression of Elfn1, human SST (and VIP) interneurons receive

facilitating inputs (Campagnola et al., 2022). However, in the mouse brain the correlation between the short-term facilitation and the expression of Elfn1 is very

high, but not perfect (Stachniak et al., 2021).

inhibition in a seemingly analogous way (Favuzzi et al., 2019).

Both loss and gain of function were shown around P14. Similarly,

somatic inhibition in CA1 abruptly emerges at the end of the

second postnatal week (Dard et al., 2022). It is therefore by

the second postnatal week that PV and SST interneurons are

committed as to where to direct their output synapses.

Intriguingly, Cbln4 is only expressed in a subset of neurons

(Figure 3). Clustering revealed that these Cbln4+ neurons

correspond to previously identified subtypes. The Tac1 cluster

labels non-Martinotti cells that target the dendrites of L4

cells (Nigro et al., 2018;Scala et al., 2019;Gouwens et al., 2020),

and the Calb2 and Etv1 clusters correspond to fanning-out

Martinotti cells (Gouwens et al., 2020;Wu et al., 2022), consistent

with a role of Cbln4 in establishing dendritic synapses. But only a

subset of the Myh8 cluster—corresponding to T-shaped Martinotti

cells (Wu et al., 2022)— expressed Cbln4, suggesting diverse

mechanisms for dendritic targeting.

An interneuron’s cell type, the plasticity of their input synapses

from PCs, and the PC compartments they target are therefore

determined before interneurons are fully embedded within cortical

circuits, and before pyramidal neurons develop their characteristic

morphology and electrophysiology. This suggest that while PV

and SST interneurons are fit for the function of compartment-

specific inhibition of PCs, some of their characteristic properties

are probably not developmentally driven by PC activity.

3. Evolutionary trajectory of

compartment-speciﬁc inhibition

On a much longer timescale than development, evolution

also changes the properties of cell types. This raises the question

whether the differentiation of PV and SST interneurons preceded

the evolution of the compartmental complexity of pyramidal

neurons.

If natural selection tuned PV and SST neurons to pyramidal cell

properties, the brains of mammalian ancestors must have contained

pyramidal cells with elaborate dendrites, while interneurons were

still undifferentiated (Figures 1C,D). This hypothesis cannot be

tested directly since our mammalian ancestors are no longer alive,

and their fossils provide no information regarding cell types. We

therefore have to infer the evolutionary history of cell types by

comparing data from modern-day species (Figure 4A) (Arendt

et al., 2016;Tosches, 2021b). Although many cell type-specific

properties such as short-term plasticity have not been measured

in non-standard model organisms (see refs. Gidon et al., 2020;

Beaulieu-Laroche et al., 2021;Campagnola et al., 2022 for recent

exceptions), transcriptomic correlates can be studied using single

cell RNA sequencing (scRNA-seq) (Tang et al., 2009), offering a

means for defining and comparing cell types across species (Tanay

and Sebé-Pedrós, 2021;Tosches, 2021b).

Frontiers in Neural Circuits 04 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

FIGURE 2

Elfn1 expression correlates with short-term facilitation in mammals. (A) UMAP (McInnes et al., 2018) plot of mouse interneurons colored by subclass

(left) and Elfn1 expression (right). The two interneuron types—SST and VIP interneurons—known to receive facilitating synapses both express Elfn1.

(B) Violin plot of Elfn1 expression by subclass. CP10K: counts per 10 thousand. (C, D) As (A, B), but for human interneurons. Data from Tasic et al.

(2018)(A, B), and Bakken et al. (2021)(C, D).

FIGURE 3

Cbln4 is expressed in a subset of mammalian SST interneurons. (A) UMAP plot of mouse and human interneurons, colored by their expression of

Cbln4, a gene that instructs synapse formation onto pyramidal dendrites in mice (Favuzzi et al., 2019). Cbln4 is expressed in certain mouse and

human interneuron subtypes, including SST cells. (B) UMAP of SST cells, clustered into subgroups. (C) Cbln4 is expressed in clusters 0, 3, and 12,

which also express marker genes Tac1,Calb2, and Etv1 (D), respectively. (E) A subset of human Sst cells also express Cbln4. Data from Tasic et al.

(2018)(A–D) and Bakken et al. (2021)(E).

Frontiers in Neural Circuits 05 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

3.1. Interneuron conservation and principal

neuron divergence

The first applications of scRNA-seq in neuroscience profiled

cell types in mice (Zeisel et al., 2015;Tasic et al., 2016). More

recently, scRNA-seq was used to classify neuron types also in

reptiles (Tosches et al., 2018) and songbirds (Colquitt et al., 2021).

The evolutionary relationships of reptiles, birds and mammals

suggest that a feature found in all three lineages predates their

divergence, while a feature found exclusively in the mammalian

lineage is, in fact, a mammalian invention. This idea enables

inferring the evolutionary history of interneurons and pyramidal

cells (Figure 4A).

Let us first consider the general evolutionary trajectory of

excitatory and inhibitory cell types. Tosches et al. (2018) used

scRNA-seq to analyse cells from the turtle and lizard forebrain and

compare them with previously published mammalian data (Tasic

et al., 2016). They found that reptilian inhibitory neurons cluster

into groups that roughly correspond to mammalian interneuron

types (Tosches et al., 2018). These results extend earlier findings

that found similarities between turtle and mammalian interneurons

based on marker genes and morphology (Blanton et al., 1987;

Reiner, 1993). Colquitt et al. (2021) recently made analogous

observations regarding the similarity of songbird and mouse

interneurons (Figures 4B,C). The most parsimonious explanation

of this sharing of interneuron types is that similar types already

existed in a common ancestor of the three lineages, rather than

convergent evolution in three lineages. This homology is likely

due to shared developmental origins: the inhibitory interneurons

of birds and reptiles are born within the conserved ganglionic

eminences. The fact that interneurons of different lineages are

homologous does not mean they are identical. For example, the

correlation between mouse PV and SST cells and the best matching

songbird clusters is 0.37 and 0.31, respectively (Figure 4B).

This is higher than the correlation between the best-matching

glutamatergic types (0.19, see below), but lower than between some

of the different cell types within the same species (mouse PV and

SST cells: 0.58). Mammalian and non-mammalian interneurons,

while homologous, therefore have likely undergone lineage-specific

adaptations.

In contrast to inhibitory interneurons, excitatory neurons

are probably not homologous between reptiles, songbirds, and

mammals (Figures 4B,C) (Tosches et al., 2018;Colquitt et al.,

2021). Excitatory cell types in different species are defined by

different combinations of transcription factors. A clear example is

given by the Fezf2 and Satb2 genes that specify subcortical (Lodato

et al., 2014) and callosal (Alcamo et al., 2008) projections,

respectively, of mammalian pyramidal cells. Strikingly, these genes

are mutually repressive in the mammalian neurons, but co-

expressed in reptilian neurons (Nomura et al., 2018;Tosches

et al., 2018). Comparing excitatory neurons in the songbird and

the mammalian brain revealed an analogous pattern: Although

excitatory neurons in the songbird forebrain express similar

genes as their counterparts in mammalian neocortex, these genes

are regulated by different transcription factors (Colquitt et al.,

2021). Instead, the transcription factors expressed by songbird

glutamatergic neurons are similar to those in e.g. the mouse

olfactory bulb and olfactory cortex. Since transcription factors

specify cellular identity (Hobert, 2008;Arendt et al., 2016)

this suggests that excitatory neurons are not conserved across

mammals, birds and reptiles (Tosches et al., 2018;Colquitt et al.,

2021;Tosches, 2021a).

Inhibitory cell types therefore seem more conserved than

excitatory cell types, which appears broadly inconsistent with an

evolutionary adaptation of interneurons to pyramidal cells. This

is further confirmed when considering the evolutionary history

of specific features of excitatory and inhibitory interneurons, in

particular, elaborate dendrites and dendrite-dependent bursting

and short-term plasticity.

3.2. Evolution of cell type-speciﬁc features

We are not aware of direct measurements of short-term

facilitation in non-mammalian species and therefore aimed to

infer its presence from the expression of Elfn1 (Box 2). To this

end, we reanalyzed publicly available gene expression data for

reptilian and songbird interneuron types (Tosches et al., 2018;

Colquitt et al., 2021). We found that Elfn1 is also expressed in the

types corresponding to mammalian SST (and VIP) interneurons

(Figure 5), but not in the type corresponding to PV interneurons.

This suggests that SST-like interneurons expressing Elfn1—and

potentially faciliating glutamatergic input synapses—were already

present in the last common ancestor of reptiles, songbirds and

mammals. In terms of potential transcriptomic correlates of

synaptic specificity, we find that Cbln4 (Figure 3) is expressed in

certain subtypes of turtle SST neurons (Figure 6A). Songbird SST

neurons, on the other hand, do not express Cbln4 (Figure 6B). The

expression of Cbln4 by Sst interneurons therefore correlates with

the presence of apical dendrites in pyramidal cells (Figure 7, see

next). The most parsimonious explanation is that Cbln4 expression

was lost in the songbird lineage. Alternatively, it could have evolved

independently in the mammals and reptiles.

So not just interneuron subtypes, but also some of their

specific properties seem evolutionarily conserved. In contrast,

glutamatergic cell types in reptiles and birds show a very different

dendritic morphology and physiology from their mammalian

pyramidal counterparts. Turtle pyramidal cells have multiple

apical dendrites, but no basal dendrites (Figure 7) (Connors and

Kriegstein, 1986). This clear morphological difference suggests

that turtle pyramidal neurons are also electrophysiogically distinct.

Larkum et al. (Larkum et al., 2008) showed that, in vitro, turtle

pyramidal neurons lack dendritic calcium spikes and dendrite-

dependent bursting. Morphologically similar pyramidal cells in

rodent piriform cortex also lack active dendrites [(Bathellier

et al., 2009;Johenning et al., 2009), but see Kumar et al.,

2018]. Interestingly, this is probably not due to an absence of

calcium channels, but rather to the presence of A-type potassium

channels (Johenning et al., 2009). Songbird excitatory cells have

a stellate morphology, and differ therefore even more from

mammalian pyramidal cells (Figure 7, see e.g., Devoogd and

Nottebohm, 1981;Benezra et al., 2018).

The lack of dendrite-dependent bursting in reptiles and

songbirds is consistent with comparative electrophysiology within

Frontiers in Neural Circuits 06 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

FIGURE 4

Evolutionary conservation of GABAergic cell types. (A) Phylogenetic approach. (B) Pearson correlation between average RNA expression in clusters

of songbird and mouse interneurons. Correlations between GABAergic neurons are typically larger. (C) UMAP plots of integrated gene expression

data for GABAergic and glutamatergic neurons. GABAergic neurons ﬁrst cluster by developmental origin (MGE vs. CGE, see Box 1) and then by

species. Mouse data from Tasic et al. (2018), songbird data and correlation analysis from Colquitt et al. (2021).

FIGURE 5

Evolutionary conservation of Elfn1 expression. (A) UMAP plot showing overexpression of Elfn1 in SST-like and VIP-like interneurons in the turtle

forebrain. Data from Tosches et al. (2018). (B) Violin plots of Elfn1 expression for each of the clusters. (C, D) As (A, B), but for zebra ﬁnch neurons.

Data from Colquitt et al. (2021).

the mammalian brain. Pyramidal neurons in the piriform cortex

are homologous to certain types of glutamatergic turtle and

songbird neurons (Colquitt et al., 2021), and also lack dendritic

plateau potentials (Bathellier et al., 2009). Pyramidal neurons with

mammalian electrophysiological properties therefore evolved after

interneurons differentiated into PV and SST cell types (Figure 8).

Frontiers in Neural Circuits 07 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

FIGURE 6

Cbln4 expression in non-mammalian species. Cbln4 is expressed in certain subtypes of turtle SST neurons, but not in songbird SST neurons. Data

from Tosches et al. (2018) and Colquitt et al. (2021).

FIGURE 7

Evolutionary divergence of projection neuron morphology. Both turtle and mammalian projection neurons have a pyramidal morphology, but only

mammalian pyramidal neurons have a single apical dendrite. Songbird projection neurons have a stellate, not pyramidal morphology. Turtle and

mammalian neurons adapted from Larkum et al. (2008) (published under a Creative Commons License https://creativecommons.org/licenses/by-

nc-sa/3.0/). Songbird neuron adapted from Kornfeld et al. (2017) (published under a Creative Commons License https://creativecommons.org/

licenses/by/4.0/).

FIGURE 8

Phylogenetic inference of interneuron and pyramidal evolution. (A) Mice, humans, songbirds and turtles all have PV and SST interneurons. The most

likely explanation for these similarities is that the interneuron types were already present in the last common ancestor of these lineages. (B) Only

mammalian glutamateric neurons are known to exhibit dendritic plateau potentials that can elicit burst ﬁring. Other lineages probably lack this trait.

The most likely explanation is that dendritic bursting evolved only once, in the mammalian lineage.

Frontiers in Neural Circuits 08 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

3.3. Ancestral Elfn1 reconstruction

The expression of Elfn1 by zebra finch and turtle SST-like

neurons suggests these cells—and therefore the ancestral SST-like

cells— receive(d) facilitating inputs. But it is also possible that

the ancestral Elfn1 protein had different functional properties.

Previous work has used Elfn1 knockout (Sylwestrak and Ghosh,

2012;Dolan and Mitchell, 2013;Tomioka et al., 2014) and

targeted deletions (Dunn et al., 2018) to discover functionally

important domains of the mouse variant (Figure 9B). To determine

the evolution of Elfn1 at the resolution of individual sites, we

computationally reconstructed its ancestral state (Hochberg and

Thornton, 2017;Orlandi et al., 2023) (see Methods), starting from

the protein sequences of extant species (Figure 9A). Alignment

of the extant sequences revealed that on average across species

74.6% of the Elfn1 sites was identical to that of the mouse

protein (Figure 9C). Combining the sequence alignment with a

probabilistic model of sequence evolution (Jones et al., 1992) and

a phylogenetic species-tree allowed us to reconstruct the ancestral

protein (Figure 9D). The amount of conservation varied between

protein domains and extant species: the zebra finch and turtle

sequences were more similar to the ancestral sequence than the

mammalian sequences (Figure 9E). Future work could use the

reconstructed sequences to determine the evolutionary history and

the molecular mechanisms of short-term facilitation.

4. Discussion

The suspicious match between the synaptic properties of

PV and SST interneurons and the postsynaptic pyramidal cell

compartments suggests that these interneuron properties could

be the result of an adaptation to pyramidal cells. Here, we

evaluated this idea of interneurons being “fit to function" from an

evolutionary and developmental perspective, and showed that the

relevant interneuron properties predate those of pyramidal cells

both during development and in evolutionary history.

Two lines of evidence indicate that the development of PV

and SST interneurons is not induced by mature pyramidal cell

activity. First, interneurons become committed to a particular cell

type (e.g., PV or SST) before reaching the developing cortex.

Interneuron fate therefore cannot be influenced by the activity

of pyramidal cells. Second, at least some of the properties of

PV and SST interneurons that strongly shape their control of

pyramidal cells—short-term plasticity and output connectivity—

emerge before the maturation of pyramidal cell morphology and

dendritic activity (dendrite-dependent bursting). It should be

noted that other interneuron properties clearly are influenced

by pyramidal cell activity. Excitatory activity regulates both the

survival of interneurons (Denaxa et al., 2018), and the formation

of inhibitory synapses (García et al., 2015;Marques-Smith et al.,

2016). Specific types of excitatory neurons determine the laminar

allocation of interneurons (Lodato et al., 2011;Wester et al., 2019),

and their activity can even change the intrinsic properties of mature

interneurons (Dehorter et al., 2015). Cell-extrinsic cues therefore

play a role in the normal development of interneurons, but are

unlikely to determine their identity and the properties we focused

on here.

Analogous arguments suggest that the evolution of PV

and SST interneurons also cannot be driven by the dendritic

physiology of pyramidal cells. The lineages of birds, reptiles

and mammals diverged over 300 million years ago, yet they

all contain roughly similar interneuron types—evidence that

these types were already present in a common ancestor of the

three lineages. In contrast to interneurons, excitatory neurons

are not conserved, and therefore probably evolved later. The

second line of evolutionary evidence relates to two specific

aspects of interneuron diversity: short-term plasticity and output

connectivity. Recent scRNA-seq data (Tosches et al., 2018;Colquitt

et al., 2021) show that reptilian and songbird SST interneurons

express Elfn1, the gene that in mouse SST neurons is necessary

and sufficient for short-term facilitation. Certain reptilian, but not

songbird, SST subtypes also express Cbln4 that plays a role in

the synaptic specificity of mammalian SST cells (Favuzzi et al.,

2019).

These data suggest that ancestral interneurons already

comprised PV- and SST-like cell types characterized by some

of the genes for cell type-specific phenotypes in mammalian

interneurons. It does not, however, imply that these phenotypes

were actually present in ancestral cells. The expression of Elfn1,

for example, is not sufficient for facilitating inputs, as shown in

the case of VIP subtypes: Multipolar and bipolar VIP neurons

both express Elfn1, but only the multipolar subtype receives

facilitating excitation (Stachniak et al., 2019). It will therefore be

interesting to directly test the presence of PV- and SST-specific

phenotypes in reptiles and birds. If neither the reptile nor the

songbird homologue of SST interneurons receives facilitating

excitatory inputs, Elfn1 was likely reused for short-term facilitation

in mammals. The emergence of short-term facilitation in SST

neurons would then be an adaptation to pyramidal bursting,

co-opting pre-existing interneuron diversity for “pyramidal cell

purposes." The anatomical connectivity of interneurons might

similarly have been reused to control pyramidal cells. In the

mammalian brain, PV and SST interneurons inhibit not just the

somata and dendrites, respectively, of pyramidal cells but also of

non-pyramidal cells. Ancestral PV and SST interneurons might

therefore have specialized in compartment-specific inhibition, but

not of pyramidal cells for which their presynaptic dynamics are so

well-matched.

Although our results show that pyramidal cell bursting

is unlikely the driver of the differentiation of PV and SST

interneurons, this is not in conflict with the functional

interpretation of these cell types. In fact, an evolution of

active pyramidal cell dendrites before the presence of specialized

interneurons would have resulted in aberrant excitation, as seen,

e.g., in Elfn1 mutants (Dolan and Mitchell, 2013;Tomioka et al.,

2014). This suggests an alternative picture, in which excitatory

neurons can only evolve in a way that still allows the existing

interneurons to regulate their activity. This still leaves open the

question why interneuron diversity evolved in the first place, if

it was not for compartment-specific inhibition. Although it is

possible that the initial separation between PV and SST cell types

was selectively neutral, this is unlikely given their evolutionary

conservation. Instead, the existence of PV and SST cells presumably

offers advantages to mammalian and non-mammalian brains alike.

An important example of an conserved function could be the

Frontiers in Neural Circuits 09 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

FIGURE 9

Reconstruction of ancestral Elfn1 protein. (A) Species-tree showing the phylogenetic relationships of the species whose Elfn1 homologs were used

to reconstruct the Elfn1 protein of the amniote ancestor. (B) Domain structure of mouse Elfn1 (Dolan et al., 2007;Dunn et al., 2018). LRR,

leucine-rich repeat; CT, C-terminal domain; FN3, ﬁbronectin type 3 domain; TM, transmembrane domain. (C) Per-site conservation across the tree

shown in (A), computed as the fraction of extant species that share the mouse amino acid at a given site. Dashed lines correspond to gaps. Mean

conservation: 0.746. (D) Posterior probability of ancestral protein. Gray: most likely (ML) sequence, red: 2nd most likely. Dashed line: cutoff for using

the 2nd most likely base in “altAll” sequence. Mean posterior: 0.986. (E) Multiple sequence alignment of protein domains shown in (A). Only the ﬁrst

two LRRs are shown for space reasons. Dots indicate identity to mouse site, dashes indicate gaps.

temporal coordination of inputs and outputs of pyramidal cells

based on oscillations (Bartos et al., 2002;Klausberger et al.,

2003).

Our findings have potential implications for the neuroscientific

interpretation of optimisation-based models of neural networks,

which have recently seen a renaissance (Mante et al., 2013;Yamins

et al., 2014;Richards et al., 2019;Saxe et al., 2019;Driscoll et al.,

2022). Most of these models describe neural data at the relatively

abstract level of dynamics and representations (Kriegeskorte and

Diedrichsen, 2019;Vyas et al., 2020). Recently, such deep network

models have also started to include circuit-level structure such as

separate excitatory and inhibitory populations (Song et al., 2016;

Naumann et al., 2022), different neuronal timescales (Kim et al.,

2019;Perez-Nieves et al., 2021), and short-term plasticity (Masse

et al., 2019;Keijser and Sprekeler, 2022). Deep learning is

therefore gradually making its way down from the level of

Frontiers in Neural Circuits 10 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

dynamical systems to that of circuits, potentially revealing

functional roles for circuit elements. Our findings highlight a

challenge to achieving this goal: Multiple circuit-level features—

such as the properties of interneuron and pyramidal cells—are

interdependent. The function of one feature might depend on

that of another and vice versa, raising the question which features

should be optimized (e.g., interneurons), and which should

be assumed as pre-exising constraints or opportunities (e.g.,

nonlinear PC dendrites). In other words, optimization-based

models face the challenge of modeling processes such as co-

evolution. Merging the functional and evolutionary/developmental

perspectives will therefore be an important challenge for

future work.

5. Methods

Code was written in Python [version (v) 3.10.8 (vanRossum,

1995)] and R [v4.2.1 (R Core Team, 2021)], based on practices

outlined in the Good Research Codebook (Mineault and Nozawa,

2021). Code for the transcriptomic analyses can be found at https://

github.com/JoramKeijser/interneuron_evolution (JoramKeijser,

2023a). Code for the protein reconstruction can be found at https://

github.com/JoramKeijser/elfn1_reconstruction (JoramKeijser,

2023b).

5.1. Datasets

We analyzed the following publicly available single cell

RNA sequencing data sets: mouse data from Tasic et al. (2018)

(downloaded from https://portal.brain-map.org/atlases-and-

data/rnaseq/mouse-v1-and-alm-smart-seq), human data from

Bakken et al. (2021) (downloaded from https://portal.brain-

map.org/atlases-and-data/rnaseq/human-m1-10x), zebra finch

data (downloaded from https://cloud.biohpc.swmed.edu/index.

php/s/nLicEtkmjGGmRF8?path=%2FHVC_RA), and turtle data

from Tosches et al. (2018) (downloaded from https://public.

brain.mpg.de/Laurent/ReptilePallium2018/). The paper’s code

repository contains a script for automatically downloading the

corresponding files.

For each data set, the starting point of our analysis was

a matrix of gene counts per cell, together with the clustering

of cells from the original publications. We converted each of

the datasets to Seurat [v4 (Hao et al., 2021)] and AnnData

[v0.8 (Virshup et al., 2021)] objects for downstream analysis in

Python and R, respectively. For visual comparison, we labeled

songbird and turtle cell clusters according to the most similar

mammalian interneuron subclass, as determined in the original

publications. This involved the merging of fine-level clusters that

presumably capture within-subclass differences. For each dataset,

we only visualized cells part of, or corresponding to, cortical

interneurons. In particular, we did not visualize the correlation of

the songbird GABAergic clusters 7, 8, and Pre, since these seem

homologous to mouse olfactory bulb interneurons (Colquitt et al.,

2021).

5.2. Dimensionality reduction and

clustering

We used AnnData and Scanpy [v1.9.1 (Wolf et al., 2018)]

to visualize the expression of the Elfn1 and Cbln4 genes. This

was done separately for each dataset. We first scaled the counts

from each cell to counts per 10 thousand (CP10K) to account for

differences in sequencing depth. We then used log plus one pseudo

count (log1p) as variance-stabilizing transformation. Finally, we

reduced the dimensionality of each dataset, by first finding highly

variable genes, performed PCA followed by UMAP (McInnes et al.,

2018). We used scanpy’s default parameters for each of these steps.

To investigate Cbln4 expression within the SST population, we

performed dimensionality reduction on all SST cells except long-

range projecting Chodl cells. Clustering was done using the Leiden

algorithm (Traag et al., 2019) with resolution 1.

5.3. Correlation analysis

We quantified the overall similarity of species-specific cell

clusters by replicating the correlation analysis from Tosches et al.

(2018) and Colquitt et al. (2021). We separately performed the

following analysis on GABAergic and glutamatergic cells, and only

compared zebra finch and mice. Specifically, we performed the

following steps.

1. Select genes to compare across species. For each species,

determine subclass-specific marker genes using Seurat’s

findAllMarkers (t-test, min.pct = 0.2, max.cells.per.ident = 200)

and retain genes with Bonferroni adjusted p-value below 0.05.

2. Intersect the two species-specific lists to find genes that are

differentially expressed in both species. This resulted in ∼500

genes, depending on the cell type.

3. Average counts within each cluster and transform to log scale

for variance-stabilization. Specifically, compute: log(1+x)+0.1,

with xthe average count.

4. Divide each gene’s value by its average across clusters to

obtain a “specificity score" invariant to a genes’ overall

expression (Tosches et al., 2018).

5. Compute the Pearson correlation between all pairs of mouse and

songbird clusters.

We visualized the result using the R package pheatmap [v1.10.12

(Kolde, 2012)].

5.4. Dataset integration

We used Seurat’s anchor-based integration (Stuart et al.,

2019) to integrate the zebra finch and mouse data. We did this

for GABAergic and glutamatergic neurons separately. First, we

jointly performed normalization and variance stabilization for each

dataset using Seurat’s scTransform (Hafemeister and Satija, 2019),

with the percentage of mitochondrial counts as covariate. Next,

we found the top 3,000 most variable features across datasets, and

used these to identify a set of anchors. These were then used to

integrate the datasets. Finally, we jointly analyzed the integrated

Frontiers in Neural Circuits 11 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

datasets using Seurat’s standard visualization pipeline: scaling and

centering, PCA, and UMAP.

5.5. Ancestral Elfn1 reconstruction

We used the Topiary pipeline (Orlandi et al., 2023) to

reconstruct the amino acid sequences of the ancestral Elfn1

protein based on sequences of extant species. To this end,

we first constructed a source dataset consisting of the Elfn1

sequences from Mus musculus (mouse), Homo sapiens (human),

Taeniopygia guttata (zebra finch), and Pelodiscus sinensis (Chinese

softshell turtle). Next, we used Topiary’s seed-to-alignment to find

sequence homologs, perform reciprocal BLAST (Altschul et al.,

1990) to predict their orthology, reduce sequence redundancy,

and align the remaining sequences using Muscle5 (Edgar, 2022).

This resulted in 62 aligned sequences that were used as input

to Topiary’s alignment-to-ancestors. This infers the maximum

likelihood (ML) gene tree, the ML substitution model, and

the ML ancestral sequences using RAxML-NG (Kozlov et al.,

2019). The posterior probability of an ancestral amino acid was

computed using the amino acid’s likelihood weighted by its prior

probability, normalized by the sum over all amino acids. Topiary

generates bootstrap replicates of the ML gene tree, and uses

GeneRax (Morel et al., 2020) to reconcile the gene tree with

the species tree. The number of bootstrap replicates was 700,

as automatically determined by the software. This number—but

not the reconstructed ML sequence—varied slightly between runs.

Finally, we used topiary’s bootstrap-reconcile that estimates the

branch support for the reconciled tree. The ancestral sequence

contained 16 ambiguous sites (based on a posterior probability

cutoff of 0.25). Besides the ML sequence, we also report a worst

case “altAll" sequence in which these ambiguous sites have been

replaced by the next most-likely amino acid. Branch support for

the amniote ancestor was 100/100, indicating very high confidence

in the existence of this ancestor, as expected. We aligned extant

and ancestral sequences using Muscle5, and visualized the resulting

alignment using the R package Ggmsa (Zhou et al., 2022).

Data availability statement

The original contributions presented in the study are included

in the article/supplementary material, further inquiries can be

directed to the corresponding authors.

Ethics statement

Ethical review and approval was not required for the

study on human participants in accordance with the local

legislation and institutional requirements. Written informed

consent for participation was not required for this study in

accordance with the national legislation and the institutional

requirements. Ethical review and approval was not required for

the animal study because the data was previously collected by

different authors, after approval of Animal Ethics Committees.

Author contributions

JK conceived of the project and analyzed the data.

HS supervised the project. JK and HS discussed the

interpretation of the data and wrote the manuscript. Both

authors contributed to the article and approved the submitted

version.

Acknowledgments

We thank Simon J. B. Butt, Loreen Hertäg, and members of the

Sprekeler Lab for comments on the manuscript.

Conﬂict of interest

The authors declare that the research was conducted in the

absence of any commercial or financial relationships that could be

construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those

of the authors and do not necessarily represent those of

their affiliated organizations, or those of the publisher,

the editors and the reviewers. Any product that may be

evaluated in this article, or claim that may be made by

its manufacturer, is not guaranteed or endorsed by the

publisher.

References

Alcamo, E. A., Chirivella, L., Dautzenberg, M. Dobreva, G., Fariñas, I., Grosschedl,

R., et al. (2008). Satb2 regulates callosal projection neuron identity in the developing

cerebral cortex. Neuron 57, 364–377. doi: 10.1016/j.neuron.2007.12.012

Altschul, S. F., Gish, W., Miller, W., Myers, E. W., and Lipman, D.

J. (1990). Basic local alignment search tool. J. Mol. Biol. 215, 403–410.

doi: 10.1016/S0022-2836(05)80360-2

Anderson, S., Eisenstat, D., Shi, L., and Rubenstein, J. (1997). Interneuron migration

from basal forebrain to neocortex: dependence on dlx genes. Science 278, 474–476.

doi: 10.1126/science.278.5337.474

Arendt, D., Musser, J. M., Baker, C. V., Bergman, A., Cepko, C., Erwin, D. H.,

et al. (2016). The origin and evolution of cell types. Nat. Rev. Genet. 17, 744–757.

doi: 10.1038/nrg.2016.127

Bakken, T. E., Jorstad, N. L., Hu, Q., Lake, B. B., Tian, W., Kalmbach, B. E., et al.

(2021). Comparative cellular analysis of motor cortex in human, marmoset and mouse.

Nature 598, 111–119. doi: 10.1038/s41586-021-03465-8

Bandler, R. C., Vitali, I., Delgado, R. N., Ho, M. C., Dvoretskova, E., Ibarra Molinas,

J. S., et al. (2021). Single-cell delineation of lineage and genetic identity in the mouse

brain. Nature 601, 404–409. doi: 10.1038/s41586-021-04237-0

Frontiers in Neural Circuits 12 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

Bartos, M., Vida, I., Frotscher, M., Meyer, A., Monyer, H., Geiger, J. R.,

et al. (2002). Fast synaptic inhibition promotes synchronized gamma oscillations

in hippocampal interneuron networks. Proc. Nat. Acad. Sci. 99, 13222–13227.

doi: 10.1073/pnas.192233099

Bathellier, B., Margrie, T. W., and Larkum, M. E. (2009). Properties of piriform

cortex pyramidal cell dendrites: implications for olfactory circuit design. J. Neurosci.

29, 12641–12652. doi: 10.1523/JNEUROSCI.1124-09.2009

Beaulieu-Laroche, L., Brown, N. J., Hansen, M., Toloza, E. H., Sharma, J., Williams,

Z. M., et al. (2021). Allometric rules for mammalian cortical layer 5 neuron biophysics.

Nature 600, 274–278. doi: 10.1038/s41586-021-04072-3

Beierlein, M., and Connors, B. W. (2002). Short-term dynamics of thalamocortical

and intracortical synapses onto layer 6 neurons in neocortex. J. Neurophysiol. 88,

1924–1932. doi: 10.1152/jn.2002.88.4.1924

Benezra, S. E., Narayanan, R. T., Egger, R., Oberlaender, M., and Long, M. A.

(2018). Morphological characterization of hvc projection neurons in the zebra finch

(Taeniopygia guttata). J. Comp. Neurol. 526, 1673–1689. doi: 10.1002/cne.24437

Berger, T. K., Silberberg, G., Perin, R., and Markram, H. (2010). Brief

bursts self-inhibit and correlate the pyramidal network. PLoS Biol. 8, e1000473.

doi: 10.1371/journal.pbio.1000473

Bhatia, A., Moza, S., and Bhalla, U. S. (2019). Precise excitation-inhibition balance

controls gain and timing in the hippocampus. Elife 8, e43415. doi: 10.7554/eLife.

43415

Blanton, M. G., Shen, J. M., and Kriegstein, A. R. (1987). Evidence for the

inhibitory neurotransmitter γ-aminobutyric acid in aspiny and sparsely spiny

nonpyramidal neurons of the turtle dorsal cortex. J. Comp. Neurol. 259, 277–297.

doi: 10.1002/cne.902590208

Butt, S. J., Fuccillo, M., Nery, S., Noctor, S., Kriegstein, A., Corbin, J. G., et al. (2005).

The temporal and spatial origins of cortical interneurons predict their physiological

subtype. Neuron 48, 591–604. doi: 10.1016/j.neuron.2005.09.034

Butt, S. J., Sousa, V. H., Fuccillo, M. V., Hjerling-Leffler, J., Miyoshi, G., Kimura,

S., et al. (2008). The requirement of nkx2-1 in the temporal specification of cortical

interneuron subtypes. Neuron 59, 722–732. doi: 10.1016/j.neuron.2008.07.031

Caillard, O., Moreno, H., Schwaller, B., Llano, I., Celio, M. R., Marty, A., et al. (2000).

Role of the calcium-binding protein parvalbumin in short-term synaptic plasticity.

Proc. Nat. Acad. Sci. 97, 13372–13377. doi: 10.1073/pnas.230362997

Campagnola, L., Seeman, S. C., Chartrand, T., Kim, L., Hoggarth, A., Gamlin, C.,

et al. (2022). Local connectivity and synaptic dynamics in mouse and human neocortex.

Science 375, eabj5861. doi: 10.1126/science.abj5861

Colquitt, B. M., Merullo, D. P., Konopka, G., Roberts, T. F., and Brainard, M.

S. (2021). Cellular transcriptomics reveals evolutionary identities of songbird vocal

circuits. Science 371, eabd9704. doi: 10.1126/science.abd9704

Connors, B. W., and Kriegstein, A. (1986). Cellular physiology of the turtle visual

cortex: distinctive properties of pyramidal and stellate neurons. J. Neurosci. 6, 164–177.

doi: 10.1523/JNEUROSCI.06-01-00164.1986

Cristo, G. D., Wu, C., Chattopadhyaya, B., Ango, F., Knott, G., Welker, E.,

et al. (2004). Subcellular domain-restricted gabaergic innervation in primary visual

cortex in the absence of sensory and thalamic inputs. Nat. Neurosci. 7, 1184–1186.

doi: 10.1038/nn1334

Dard, R. F., Leprince, E., Denis, J., Balappa, S. R., Suchkov, D., Boyce, R., et al. (2022).

The rapid developmental rise of somatic inhibition disengages hippocampal dynamics

from self-motion. Elife 11, e78116. doi: 10.7554/eLife.78116.sa2

Dehorter, N., Ciceri, G., Bartolini, G., Lim, L., del Pino, I., and Marín, O. (2015).

Tuning of fast-spiking interneuron properties by an activity-dependent transcriptional

switch. Science 349, 1216–1220. doi: 10.1126/science.aab3415

Denaxa, M., Neves, G., Rabinowitz, A., Kemlo, S., Liodis, P., Burrone, J., et al. (2018).

Modulation of apoptosis controls inhibitory interneuron number in the cortex. Cell

Rep. 22, 1710–1721. doi: 10.1016/j.celrep.2018.01.064

Devoogd, T. J., and Nottebohm, F. (1981). Sex differences in dendritic morphology

of a song control nucleus in the canary: a quantitative golgi study. J. Comp. Neurol. 196,

309–316. doi: 10.1002/cne.901960209

Dolan, J., and Mitchell, K. J. (2013). Mutation of elfn1 in mice causes seizures and

hyperactivity. PLoS ONE 8, e80491. doi: 10.1371/journal.pone.0080491

Dolan, J., Walshe, K., Alsbury, S., Hokamp, K., O’Keeffe, S., Okafuji, T., et al.

(2007). The extracellular leucine-rich repeat superfamily; a comparative survey and

analysis of evolutionary relationships and expression patterns. BMC Genomics 8, 1–24.

doi: 10.1186/1471-2164-8-320

Driscoll, L., Shenoy, K., and Sussillo, D. (2022). Flexible multitask

computation in recurrent networks utilizes shared dynamical motifs. bioRxiv.

doi: 10.1101/2022.08.15.503870

Dunn, H. A., Patil, D. N., Cao, Y., Orlandi, C., and Martemyanov, K. A. (2018).

Synaptic adhesion protein elfn1 is a selective allosteric modulator of group iii

metabotropic glutamate receptors in trans. Proc. Nat. Acad. Sci. 115, 5022–5027.

doi: 10.1073/pnas.1722498115

Edgar, R. C. (2022). Muscle5: high-accuracy alignment ensembles enable unbiased

assessments of sequence homology and phylogeny. Nat. Commun. 13, 6968.

doi: 10.1038/s41467-022-34630-w

Favuzzi, E., Deogracias, R., Marques-Smith, A., Maeso, P., Jezequel, J., Exposito-

Alonso, D., et al. (2019). Distinct molecular programs regulate synapse specificity in

cortical inhibitory circuits. Science 363, 413–417. doi: 10.1126/science.aau8977

Ferguson, K. A., and Cardin, J. A. (2020). Mechanisms underlying gain modulation

in the cortex. Nat. Rev. Neurosci. 21, 80–92. doi: 10.1038/s41583-019-0253-y

Fishell, G., and Kepecs, A. (2020). Interneuron types as attractors and controllers.

Annu. Rev. Neurosci. 43, 1–30. doi: 10.1146/annurev-neuro-070918-050421

Flames, N., Pla, R., Gelman, D. M., Rubenstein, J. L., Puelles, L., and

Marín, O. (2007). Delineation of multiple subpallial progenitor domains by the

combinatorial expression of transcriptional codes. J. Neurosci. 27, 9682–9695.

doi: 10.1523/JNEUROSCI.2750-07.2007

Fogarty, M., Grist, M., Gelman, D., Marín, O., Pachnis, V., and Kessaris,

N. (2007). Spatial genetic patterning of the embryonic neuroepithelium generates

gabaergic interneuron diversity in the adult cortex. J. Neurosci. 27, 10935–10946.

doi: 10.1523/JNEUROSCI.1629-07.2007

Franceschetti, S., Sancini, G., Panzica, F., Radici, C., and Avanzini, G. (1998).

Postnatal differentiation of firing properties and morphological characteristics in

layer v pyramidal neurons of the sensorimotor cortex. Neuroscience 83, 1013–1024.

doi: 10.1016/S0306-4522(97)00463-6

Fu, Y., Tucciarone, J. M., Espinosa, J. S., Sheng, N., Darcy, D. P., Nicoll, R. A.,

et al. (2014). A cortical circuit for gain control by behavioral state. Cell 156, 1139–1152.

doi: 10.1016/j.cell.2014.01.050

García, N. V. D. M., Priya, R., Tuncdemir, S. N., Fishell, G., Karayannis, T. (2015).

Sensory inputs control the integration of neurogliaform interneurons into cortical

circuits. Nat. Neurosci. 18, 393–401. doi: 10.1038/nn.3946

Gentet, L. J., Kremer, Y., Taniguchi, H., Huang, Z. J., Staiger, J. F., Petersen, C.

C., et al. (2012). Unique functional properties of somatostatin-expressing gabaergic

neurons in mouse barrel cortex. Nat. Neurosci. 15, 607–612. doi: 10.1038/nn.3051

Gidon, A., Zolnik, T. A., Fidzinski, P., Bolduan, F., Papoutsi, A., Poirazi, P., et al.

(2020). Dendritic action potentials and computation in human layer 2/3 cortical

neurons. Science 367, 83–87. doi: 10.1126/science.aax6239

Goldberg, J. H., Lacefield, C. O., and Yuste, R. (2004). Global dendritic calcium

spikes in mouse layer 5 low threshold spiking interneurones: implications for control

of pyramidal cell bursting. J. Physiol. 558, 465–478. doi: 10.1113/jphysiol.2004.064519

Gouwens, N. W., Sorensen, S. A., Baftizadeh, F., Budzillo, A., Lee, B. R., Jarsky,

T., et al. (2020). Integrated morphoelectric and transcriptomic classification of cortical

gabaergic cells. Cell, 183, 935–953. doi: 10.1016/j.cell.2020.09.057

Hafemeister, C., and Satija, R. (2019). Normalization and variance stabilization of

single-cell rna-seq data using regularized negative binomial regression. Genome Biol.

20, 1–15. doi: 10.1186/s13059-019-1874-1

Hansen, D. V., Lui, J. H., Flandin, P., Yoshikawa, K., Rubenstein, J. L.,

Alvarez-Buylla, A., et al. (2013). Non-epithelial stem cells and cortical interneuron

production in the human ganglionic eminences. Nat. Neurosci. 16, 1576–1587.

doi: 10.1038/nn.3541

Hao, Y., Hao, S., Andersen-Nissen, E., Mauck, I. I. I., Zheng, W. M., Butler, S.,

et al. (2021). Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587.

doi: 10.1016/j.cell.2021.04.048

He, M., Tucciarone, J., Lee, S., Nigro, M. J., Kim, Y., Levine, J. M., et al. (2016).

Strategies and tools for combinatorial targeting of gabaergic neurons in mouse cerebral

cortex. Neuron 91, 1228–1243. doi: 10.1016/j.neuron.2016.08.021

Hertäg, L., and Clopath, C. (2022). Prediction-error neurons in circuits with

multiple neuron types: formation, refinement, and functional implications. Proc. Nat.

Acad. Sci. 119, e2115699119. doi: 10.1073/pnas.2115699119

Hobert, O. (2008). Regulatory logic of neuronal diversity: terminal

selector genes and selector motifs. Proc. Nat. Acad. Sci. 105, 20067–20071.

doi: 10.1073/pnas.0806070105

Hochberg, G. K., and Thornton, J. W. (2017). Reconstructing ancient proteins to

understand the causes of structure and function. Annu. Rev. Biophys. 46, 247–269.

doi: 10.1146/annurev-biophys-070816-033631

Hu, J. S., Vogt, D., Lindtner, S., Sandberg, M., Silberberg, S. N., Rubenstein,

J. L., et al. (2017). Coup-tf1 and coup-tf2 control subtype and laminar

identity of mge-derived neocortical interneurons. Development 144, 2837–2851.

doi: 10.1242/dev.150664

Inan, M., Welagen, J., and Anderson, S. A. (2012). Spatial and temporal bias in the

mitotic origins of somatostatin-and parvalbumin-expressing interneuron subgroups

and the chandelier subtype in the medial ganglionic eminence. Cereb. Cortex 22,

820–827. doi: 10.1093/cercor/bhr148

Jiang, X., Shen, S., Cadwell, C. R., Berens, P., Sinz, F., Ecker, A. S., et al. (2015).

Principles of connectivity among morphologically defined cell types in adult neocortex.

Science 350, aac9462. doi: 10.1126/science.aac9462

Frontiers in Neural Circuits 13 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

Johenning, F. W., Beed, P. S., Trimbuch, T., Bendels, M. H., Winterer, J., Schmitz, D.,

et al. (2009). Dendritic compartment and neuronal output mode determine pathway-

specific long-term potentiation in the piriform cortex. J. Neurosci. 29, 13649–13661.

doi: 10.1523/JNEUROSCI.2672-09.2009

Jones, D. T., Taylor, W. R., and Thornton, J. M. (1992). The rapid generation

of mutation data matrices from protein sequences. Bioinformatics 8, 275–282.

doi: 10.1093/bioinformatics/8.3.275

JoramKeijser (2023a). JoramKeijser/interneuron_evolution: Code After Revision,

March 2023.

JoramKeijser (2023b). JoramKeijser/interneuron_evolution: Code After Revision,

March 2023.

Keijser, J., and Sprekeler, H. (2022). Optimizing interneuron circuits for

compartment-specific feedback inhibition. PLoS Comput. Biol. 18, e1009933.

doi: 10.1371/journal.pcbi.1009933

Keller, G. B., and Mrsic-Flogel, T. D. (2018). Predictive processing: a canonical

cortical computation. Neuron 100, 424–435. doi: 10.1016/j.neuron.2018.10.003

Kepecs, A., and Fishell, G. (2014). Interneuron cell types are fit to function. Nature

505, 318–326. doi: 10.1038/nature12983

Kim, R., Li, Y., and Sejnowski, T. J. (2019). Simple framework for constructing

functional spiking recurrent neural networks. Proc. Nat. Acad. Sci. 116, 22811–22820.

doi: 10.1073/pnas.1905926116

Klausberger, T., Magill, P. J., Márton, L. F., Roberts, J. D. B., Cobden, P. M., Buzsáki,

G., et al. (2003). Brain-state-and cell-type-specific firing of hippocampal interneurons

in vivo.Nature 421, 844–848. doi: 10.1038/nature01374

Kolde, R. (2012). Pheatmap: Pretty Heatmaps. R package version, 1, 726.

Kornfeld, J., Benezra, S. E., Narayanan, R. T., Svara, F., Egger, R., Oberlaender, M.,

et al. (2017). Em connectomics reveals axonal target variation in a sequence-generating

network. Elife 6, e24364. doi: 10.7554/eLife.24364.020

Kozlov, A. M., Darriba, D., Flouri, T., Morel, B., and Stamatakis, A. (2019).

Raxml-ng: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic

inference. Bioinformatics 35, 4453–4455. doi: 10.1093/bioinformatics/btz305

Kriegeskorte, N., and Diedrichsen, J. (2019). Peeling the onion

of brain representations. Annu. Rev. Neurosci. 42, 407–432.

doi: 10.1146/annurev-neuro-080317-061906

Kumar, A., Schiff, O., Barkai, E., Mel, B. W., Poleg-Polsky, A., Schiller, J., et al.

(2018). Nmda spikes mediate amplification of inputs in the rat piriform cortex. Elife

7, e38446. doi: 10.7554/eLife.38446.018

Larkum, M. (2013). A cellular mechanism for cortical associations: an

organizing principle for the cerebral cortex. Trends Neurosci. 36, 141–151.

doi: 10.1016/j.tins.2012.11.006

Larkum, M. E., Watanabe, S., Lasser-Ross, N., Rhodes, P., and Ross, W. N.

(2008). Dendritic properties of turtle pyramidal neurons. J. Neurophysiol. 99, 683–694.

doi: 10.1152/jn.01076.2007

Larkum, M. E., Zhu, J. J., and Sakmann, B. (1999). A new cellular mechanism

for coupling inputs arriving at different cortical layers. Nature 398, 338–341.

doi: 10.1038/18686

Ledderose, J. M., Zolnik, T. A., Toumazou, M., Trimbuch, T., Rosenmund, C.,

Eickholt, B. J., et al. (2022). Input to cortical layer 1 of somatosensory cortex. bioRxiv.

[preprint]. doi: 10.1101/2021.11.26.469979v3

Letzkus, J. J., Kampa, B. M., and Stuart, G. J. (2006). Learning rules for spike

timing-dependent plasticity depend on dendritic synapse location. J. Neurosci. 26,

10420–10429. doi: 10.1523/JNEUROSCI.2650-06.2006

Letzkus, J. J., Wolff, S. B., and Lüthi, A. (2015). Disinhibition, a circuit

mechanism for associative learning and memory. Neuron 88, 264–276.

doi: 10.1016/j.neuron.2015.09.024

Lim, L., Mi, D., Llorca, A., and Marín, O. (2018a). Development and

functional diversification of cortical interneurons. Neuron 100, 294–313.

doi: 10.1016/j.neuron.2018.10.009

Lim, L., Pakan, J. M., Selten, M. M., Marques-Smith, A., Llorca, A., Bae, S. E., et al.

(2018b). Optimization of interneuron function by direct coupling of cell migration and

axonal targeting. Nat. Neurosci. 21, 920–931. doi: 10.1038/s41593-018-0162-9

Lodato, S., Molyneaux, B. J., Zuccaro, E., Goff, L. A., Chen, H.-H., Yuan, W., et al.

(2014). Gene co-regulation by fezf2 selects neurotransmitter identity and connectivity

of corticospinal neurons. Nat. Neurosci. 17, 1046–1054. doi: 10.1038/nn.3757

Lodato, S., Rouaux, C., Quast, K. B., Jantrachotechatchawan, C., Studer, M.,

Hensch, T. K., et al. (2011). Excitatory projection neuron subtypes control the

distribution of local inhibitory interneurons in the cerebral cortex. Neuron 69, 763–779.

doi: 10.1016/j.neuron.2011.01.015

Lovett-Barron, M., Turi, G. F., Kaifosh, P., Lee, P. H., Bolze, F., Sun, X.-H., et al.

(2012). Regulation of neuronal input transformations by tunable dendritic inhibition.

Nat. Neurosci. 15, 423–430. doi: 10.1038/nn.3024

Ma, T., Wang, C., Wang, L., Zhou, X., Tian, M., Zhang, Q., et al. (2013). Subcortical

origins of human and monkey neocortical interneurons. Nat. Neurosci. 16, 1588–1597.

doi: 10.1038/nn.3536

Mante, V., Sussillo, D., Shenoy, K. V., and Newsome, W. T. (2013). Context-

dependent computation by recurrent dynamics in prefrontal cortex. Nature 503, 78–84.

doi: 10.1038/nature12742

Marín, O., Rubenstein, J. L. (2001). A long, remarkable journey: tangential

migration in the telencephalon. Nat. Rev. Neurosci. 2, 780–790. doi: 10.1038/35097509

Marques-Smith, A., Lyngholm, D., Kaufmann, A.-K., Stacey, J. A., Hoerder-

Suabedissen, A., Becker, E. B., et al. (2016). A transient translaminar gabaergic

interneuron circuit connects thalamocortical recipient layers in neonatal

somatosensory cortex. Neuron 89, 536–549. doi: 10.1016/j.neuron.2016.01.015

Masse, N. Y., Yang, G. R., Song, H. F., Wang, X.-J., and Freedman, D. J. (2019).

Circuit mechanisms for the maintenance and manipulation of information in working

memory. Nat. Neurosci. 22, 1159–1167. doi: 10.1038/s41593-019-0414-3

Mayer, C., Hafemeister, C., Bandler, R. C., Machold, R., Brito, R. B., Jaglin, X., et al.

(2018). Developmental diversification of cortical inhibitory interneurons. Nature 555,

457–462. doi: 10.1038/nature25999

McInnes, L., Healy, J., and Melville, J. (2018). Umap: uniform manifold

approximation and projection for dimension reduction. arXiv. [preprint].

doi: 10.21105/joss.00861

McKenzie, M. G., Cobbs, L. V., Dummer, P. D., Petros, T. J., Halford, M. M.,

Stacker, S. A., et al. (2019). Non-canonical wnt signaling through ryk regulates the

generation of somatostatin-and parvalbumin-expressing cortical interneurons. Neuron

103, 853–864. doi: 10.1016/j.neuron.2019.06.003

Mi, D., Li, Z., Lim, L., Li, M., Moissidis, M., Yang, Y., et al. (2018). Early

emergence of cortical interneuron diversity in the mouse embryo. Science 360, 81–85.

doi: 10.1126/science.aar6821

Mineault, P., and Nozawa, K. (2021). patrickmineault/codebook: 1.0.0.

Miyoshi, G., Hjerling-Leffler, J., Karayannis, T., Sousa, V. H., Butt, S. J., Battiste,

J., et al. (2010). Genetic fate mapping reveals that the caudal ganglionic eminence

produces a large and diverse population of superficial cortical interneurons. J. Neurosci.

30, 1582–1594. doi: 10.1523/JNEUROSCI.4515-09.2010

Morel, B., Kozlov, A. M., Stamatakis, A., and Szöll˝

osi, G. J. (2020). Generax:

a tool for species-tree-aware maximum likelihood-based gene family tree inference

under gene duplication, transfer, and loss. Mol. Biol. Evol. 37, 2763–2774.

doi: 10.1093/molbev/msaa141

Murayama, M. Pérez-Garci, E., Nevian, T., Bock, T., Senn, W., Larkum, M. E.

(2009). Dendritic encoding of sensory stimuli controlled by deep cortical interneurons.

Nature 457, 1137–1141. doi: 10.1038/nature07663

Naud, R., Bathellier, B., and Gerstner, W. (2013). Spike timing prediction with active

dendrites. arXiv. [preprint]. doi: 10.48550/arXiv.1311.3586

Naud, R., and Sprekeler, H. (2018). Sparse bursts optimize information

transmission in a multiplexed neural code. Proc. Nat. Acad. Sci. 115, E6329–E6338.

doi: 10.1073/pnas.1720995115

Naumann, L. B., Keijser, J., and Sprekeler, H. (2022). Invariant neural

subspaces maintained by feedback modulation. Elife 11, e76096. doi: 10.7554/eLife.

76096.sa2

Nery, S., Fishell, G., and Corbin, J. G. (2002). The caudal ganglionic eminence is a

source of distinct cortical and subcortical cell populations. Nat. Neurosci. 5, 1279–1287.

doi: 10.1038/nn971

Nigro, M. J., Hashikawa-Yamasaki, Y., and Rudy, B. (2018). Diversity and

connectivity of layer 5 somatostatin-expressing interneurons in the mouse barrel

cortex. J. Neurosci. 38, 1622–1633. doi: 10.1523/JNEUROSCI.2415-17.2017

Nomura, T., Yamashita, W., Gotoh, H., and Ono, K. (2018). Species-specific

mechanisms of neuron subtype specification reveal evolutionary plasticity of

amniote brain development. Cell Rep. 22, 3142–3151. doi: 10.1016/j.celrep.2018.

02.086

Orlandi, K. N., Phillips, S. R., Sailer, Z. R., Harman, J. L., and Harms, M. J. (2023).

Topiary: pruning the manual labor from ancestral sequence reconstruction. Protein Sci.

32, e4551. doi: 10.1002/pro.4551

Perez-Nieves, N., Leung, V. C., Dragotti, P. L., and Goodman, D. F.

(2021). Neural heterogeneity promotes robust learning. Nat. Commun. 12, 1–9.

doi: 10.1038/s41467-021-26022-3

Petreanu, L., Huber, D., Sobczyk, A., and Svoboda, K. (2007). Channelrhodopsin-2-

assisted circuit mapping of long-range callosal projections. Nat. Neurosci. 10, 663–668.

doi: 10.1038/nn1891

Pouille, F., and Scanziani, M. (2004). Routing of spike series by dynamic circuits in

the hippocampus. Nature 429, 717–723. doi: 10.1038/nature02615

R Core Team. (2021). R: A Language and Environment for Statistical Computing.

Vienna: R Foundation For Statistical Computing.

Rakic, P. (1971). Neuron-glia relationship during granule cell migration in

developing cerebellar cortex. a golgi and electonmicroscopic study in macacus rhesus.

J. Comp. Neurol. 141, 283–312. doi: 10.1002/cne.901410303

Reiner, A. (1993). Neurotransmitter organization and connections of turtle cortex:

implications for the evolution of mammalian isocortex. Comp. Biochem. Physiol. A

Physiol. 104, 735–748. doi: 10.1016/0300-9629(93)90149-X

Frontiers in Neural Circuits 14 frontiersin.org

Keijser and Sprekeler 10.3389/fncir.2023.1172464

Reyes, A., Lujan, R., Rozov, A., Burnashev, N., Somogyi, P., Sakmann, B., et al.

(1998). Target-cell-specific facilitation and depression in neocortical circuits. Nat.

Neurosci. 1, 279–285. doi: 10.1038/1092

Richards, B. A., Lillicrap, T. P., Beaudoin, P., Bengio, Y., Bogacz, R., Christensen,

A., et al. (2019). A deep learning framework for neuroscience. Nat. Neurosci. 22,

1761–1770. doi: 10.1038/s41593-019-0520-2

Romand, S., Wang, Y., Toledo-Rodriguez, M., and Markram, H. (2011).

Morphological development of thick-tufted layer v pyramidal cells in the rat

somatosensory cortex. Front. Neuroanat. 5, 5. doi: 10.3389/fnana.2011.00005

Sadeh, S., and Clopath, C. (2021). Inhibitory stabilization and cortical computation.

Nat. Rev. Neurosci. 22, 21–37. doi: 10.1038/s41583-020-00390-z

Saxe, A. M., McClelland, J. L., and Ganguli, S. (2019). A mathematical theory of

semantic development in deep neural networks. Proc. Nat. Acad. Sci. 116, 11537–11546.

doi: 10.1073/pnas.1820226116

Scala, F., Kobak, D., Shan, S., Bernaerts, Y., Laturnus, S., Cadwell, C. R., et al.

(2019). Layer 4 of mouse neocortex differs in cell types and circuit organization between

sensory areas. Nat. Commun. 10, 4174. doi: 10.1038/s41467-019-12058-z

Schmitz, M. T., Sandoval, K., Chen, C. P., Mostajo-Radji, M. A., Seeley, W. W.,

Nowakowski, T. J., et al. (2022). The development and evolution of inhibitory neurons

in primate cerebrum. Nature 603, 871–877. doi: 10.1038/s41586-022-04510-w

Shi, Y., Wang, M., Mi, D., Lu, T., Wang, B., Dong, H., et al. (2021). Mouse

and human share conserved transcriptional programs for interneuron development.

Science 374, eabj6641. doi: 10.1126/science.abj6641

Silberberg, G., and Markram, H. (2007). Disynaptic inhibition between

neocortical pyramidal cells mediated by martinotti cells. Neuron 53, 735–746.

doi: 10.1016/j.neuron.2007.02.012

Sjostrom, P. J., Rancz, E. A., Roth, A., and Hausser, M. (2008). Dendritic excitability

and synaptic plasticity. Physiol. Rev. 88, 769–840. doi: 10.1152/physrev.00016.2007

Song, H. F., Yang, G. R., and Wang, X.-J. (2016). Training excitatory-inhibitory

recurrent neural networks for cognitive tasks: a simple and flexible framework. PLoS

Comput. Biol. 12, e1004792. doi: 10.1371/journal.pcbi.1004792

Stachniak, T. J., Kastli, R., Hanley, O., Argunsah, A. Ö., van der Valk, E.

G. T., Kanatouris, G., et al. (2021). Postmitotic prox1 expression controls the

final specification of cortical vip interneuron subtypes. J. Neurosci. 41, 8150–8162.

doi: 10.1523/JNEUROSCI.1021-21.2021

Stachniak, T. J., Sylwestrak, E. L., Scheiffele, P., Hall, B. J., and Ghosh, A.

(2019). Elfn1-induced constitutive activation of mglur7 determines frequency-

dependent recruitment of somatostatin interneurons. J. Neurosci. 39, 4461–4474.

doi: 10.1523/JNEUROSCI.2276-18.2019

Stuart, T., Butler, A., Hoffman, P., Hafemeister, C., Papalexi, E., Mauck, I. I. I.,

et al. R. (2019). Comprehensive integration of single-cell data. Cell 177, 1888–1902.

doi: 10.1016/j.cell.2019.05.031

Sussel, L., Marin, O., Kimura, S., and Rubenstein, J. (1999). Loss of nkx2, 1.

homeobox gene function results in a ventral to dorsal molecular respecification within

the basal telencephalon: evidence for a transformation of the pallidum into the

striatum. Development 126, 3359–3370. doi: 10.1242/dev.126.15.3359

Sylwestrak, E. L., and Ghosh, A. (2012). Elfn1 regulates target-specific

release probability at ca1-interneuron synapses. Science 338, 536–540.

doi: 10.1126/science.1222482

Tanay, A., and Sebé-Pedrós, A. (2021). Evolutionary cell type mapping with single-

cell genomics. Trends Genet. 37, 919–932. doi: 10.1016/j.tig.2021.04.008

Tang, F., Barbacioru, C., Wang, Y., Nordman, E., Lee, C., Xu, N., et al. (2009).

mrna-seq whole-transcriptome analysis of a single cell. Nat. Methods 6, 377–382.

doi: 10.1038/nmeth.1315

Tasic, B., Menon, V., Nguyen, T. N., Kim, T. K., Jarsky, T., Yao, Z., et al. (2016). Adult

mouse cortical cell taxonomy revealed by single cell transcriptomics. Nat. Neurosci. 19,

335–346. doi: 10.1038/nn.4216

Tasic, B., Yao, Z., Graybuck, L. T., Smith, K. A., Nguyen, T. N., Bertagnolli, D., et al.

(2018). Shared and distinct transcriptomic cell types across neocortical areas. Nature

563, 72–78. doi: 10.1038/s41586-018-0654-5

Tomioka, N. H., Yasuda, H., Miyamoto, H., Hatayama, M., Morimura, N.,

Matsumoto, Y., et al. (2014). Elfn1 recruits presynaptic mglur7 in trans and its loss

results in seizures. Nat. Commun. 5, 1–16. doi: 10.1038/ncomms5501

Tosches, M. A. (2021a). Different origins for similar brain circuits. Science 371,

676–677. doi: 10.1126/science.abf9551

Tosches, M. A. (2021b). From cell types to an integrated understanding of brain

evolution: the case of the cerebral cortex. Annu. Rev. Cell Dev. Biol. 37, 495–517.

doi: 10.1146/annurev-cellbio-120319-112654

Tosches, M. A., Yamawaki, T. M., Naumann, R. K., Jacobi, A. A., Tushev,

G., Laurent, G., et al. (2018). Evolution of pallium, hippocampus, and cortical

cell types revealed by single-cell transcriptomics in reptiles. Science 360, 881–888.

doi: 10.1126/science.aar4237

Traag, V. A., Waltman, L., and Van Eck, N. J. (2019). From louvain

to leiden: guaranteeing well-connected communities. Sci. Rep. 9, 5233.

doi: 10.1038/s41598-019-41695-z

Tremblay, R., Lee, S., and Rudy, B. (2016). Gabaergic interneurons in

the neocortex: from cellular properties to circuits. Neuron 91, 260–292.

doi: 10.1016/j.neuron.2016.06.033

Tsodyks, M., Pawelzik, K., and Markram, H. (1998). Neural networks with dynamic

synapses. Neural Comput. 10, 821–835. doi: 10.1162/089976698300017502

Udakis, M., Pedrosa, V., Chamberlain, S. E., Clopath, C., and Mellor, J. R. (2020).

Interneuron-specific plasticity at parvalbumin and somatostatin inhibitory synapses

onto ca1 pyramidal neurons shapes hippocampal output. Nat. Commun. 11, 1–17.

doi: 10.1038/s41467-020-18074-8

vanRossum, G. (1995). Python Reference Manual. Ashburn, VA: Department of

Computer Science [CS], (R 9525).

Virshup, I., Rybakov, S., Theis, F. J., Angerer, P., and Wolf, F. A. (2021). anndata:

Annotated Data. bioRxiv. doi: 10.1101/2021.12.16.473007

Vogels, T. P., Sprekeler, H., Zenke, F., Clopath, C., and Gerstner, W. (2011).

Inhibitory plasticity balances excitation and inhibition in sensory pathways and

memory networks. Science 334, 1569–1573. doi: 10.1126/science.1211095

Vyas, S., Golub, M. D., Sussillo, D., and Shenoy, K. V. (2020). Computation

through neural population dynamics. Annu. Rev. Neurosci. 43, 249.

doi: 10.1146/annurev-neuro-092619-094115

Wamsley, B., and Fishell, G. (2017). Genetic and activity-dependent

mechanisms underlying interneuron diversity. Nat. Rev. Neurosci. 18, 299–309.

doi: 10.1038/nrn.2017.30

Wehr, M., and Zador, A. M. (2003). Balanced inhibition underlies

tuning and sharpens spike timing in auditory cortex. Nature 426, 442–446.

doi: 10.1038/nature02116

Wester, J. C., Mahadevan, V., Rhodes, C. T., Calvigioni, D., Venkatesh,

S., Maric, D., et al. (2019). Neocortical projection neurons instruct inhibitory

interneuron circuit development in a lineage-dependent manner. Neuron 102,

960–975. doi: 10.1016/j.neuron.2019.03.036

Wichterle, H., Turnbull, D. H., Nery, S., Fishell, G., and Alvarez-Buylla, A.

(2001). In utero fate mapping reveals distinct migratory pathways and fates of

neurons born in the mammalian basal forebrain. Development 128, 3759–3771.

doi: 10.1242/dev.128.19.3759

Williams, L. E., and Holtmaat, A. (2019). Higher-order thalamocortical inputs

gate synaptic long-term potentiation via disinhibition. Neuron 101, 91–102.

doi: 10.1016/j.neuron.2018.10.049

Williams, S. R., and Stuart, G. J. (1999). Mechanisms and consequences of action

potential burst firing in rat neocortical pyramidal neurons. J. Physiol. 521(Pt 2), 467.

doi: 10.1111/j.1469-7793.1999.00467.x

Wolf, F. A., Angerer, P., and Theis, F. J. (2018). Scanpy: large-scale single-cell gene

expression data analysis. Genome Biol. 19, 1–5. doi: 10.1186/s13059-017-1382-0

Wonders, C. P., and Anderson, S. A. (2006). The origin and specification of cortical

interneurons. Nat. Rev. Neurosci. 7, 687–696. doi: 10.1038/nrn1954

Wonders, C. P., Taylor, L., Welagen, J., Mbata, I. C., Xiang, J. Z., Anderson, S. A.,

et al. (2008). A spatial bias for the origins of interneuron subgroups within the medial

ganglionic eminence. Dev. Biol. 314, 127–136. doi: 10.1016/j.ydbio.2007.11.018

Wu, Y. K., Miehl, C., and Gjorgjieva, J. (2022). Regulation of circuit organization

and function through inhibitory synaptic plasticity. Trends Neurosci. 45, 884–898.

doi: 10.1016/j.tins.2022.10.006

Xu, Q., Cobos, I., De La Cruz, E., Rubenstein, J. L., and Anderson, S.

A. (2004). Origins of cortical interneuron subtypes. J. Neurosci. 24, 2612–2622.

doi: 10.1523/JNEUROSCI.5667-03.2004

Yamins, D. L., Hong, H., Cadieu, C. F., Solomon, E. A., Seibert, D., DiCarlo, J. J.,

et al. (2014). Performance-optimized hierarchical models predict neural responses in

higher visual cortex. Proc. Nat. Acad. Sci. 111, 8619–8624. doi: 10.1073/pnas.14031

12111

Zeisel, A., Muñoz-Manchado, A. B., Codeluppi, S., Lönnerberg, P., La Manno, G.,

Juréus, A., et al. (2015). Cell types in the mouse cortex and hippocampus revealed by

single-cell rna-seq. Science 347, 1138–1142. doi: 10.1126/science.aaa1934

Zhou, L., Feng, T., Xu, S., Gao, F., Lam, T. T., Wang, Q., et al. (2022). ggmsa: a visual

exploration tool for multiple sequence alignment and associated data. Brief. Bioinform.

23, bbac222. doi: 10.1093/bib/bbac222

Zhu, J. J. (2000). Maturation of layer 5 neocortical pyramidal neurons: amplifying

salient layer 1 and layer 4 inputs by ca2+ action potentials in adult rat tuft dendrites. J.

Physiol. 526, 571–587. doi: 10.1111/j.1469-7793.2000.00571.x

Frontiers in Neural Circuits 15 frontiersin.org