Patch-seq: Advances and Biological Applications

Abstract

Multimodal analysis of gene-expression patterns, electrophysiological properties, and morphological phenotypes at the single-cell/single-nucleus level has been arduous because of the diversity and complexity of neurons. The emergence of Patch-sequencing (Patch-seq) directly links transcriptomics, morphology, and electrophysiology, taking neuroscience research to a multimodal era. In this review, we summarized the development of Patch-seq and recent applications in the cortex, hippocampus, and other nervous systems. Through generating multimodal cell type atlases, targeting specific cell populations, and correlating transcriptomic data with phenotypic information, Patch-seq has provided new insight into outstanding questions in neuroscience. We highlight the challenges and opportunities of Patch-seq in neuroscience and hope to shed new light on future neuroscience research.

Introduction

The advent of patch-clamp technology has revolutionized the study of cellular physiology and revealed how neuronal activity supports brain function at the single-cell level. Whole-cell recording has been applied to many regions of animal tissues, combined with optical technology, which reveals the dynamics of individual membrane potentials and the dynamics of individual membrane potentials. It has also shown other features of cellular activity, such as anatomical connections, genetic properties, and collective action related to brain function (Noguchi et al. 2021). However, traditional patch clamp technology cannot describe the profile of molecular characteristics of cells.

The single-cell/single-nucleus RNA sequencing (sc/snRNA-seq) technology has provided unprecedented information to study the expression of thousands of cells, identify cell subpopulations, analyze differential gene expression, and reconstruct the spatiotemporal dynamics of gene expression. Current sc/snRNA-seq has become one of the most powerful tools to characterize the transcriptome heterogeneity of single cells and reveal previously unreported cell types or states in complex tissues, providing new insights into cellular diversity and heterogeneity and tissue expression profiles (Jovic et al. 2022). Compared with patch clamp, the advantage of sc/snRNA-seq technology is the ability to obtain a large number of single-cell/single-nucleus transcriptome data simultaneously but lacking morphological and electrophysiological information.

By combining patch-clamp and sc/snRNA-seq, scientists have developed Patch-seq technology that simultaneously obtains electrophysiological, morphological, and gene expression data at a single neuron (Scala et al. 2021; Cadwell et al. 2020; Cadwell et al. 2017a, b). It is possible to correlate the transcriptomic features of a given neuron directly with other identifying characteristics of the same neuron, for example, the precise location, morphology, and intrinsic electrophysiological properties of the neuron (Gouwens et al. 2020). After some preliminary studies, Patch-seq was further optimized and successfully applied to the mouse cortex (Mahfooz and Ellender 2021; Oberst et al. 2019; Scala et al. 2019), hippocampus (Oláh et al. 2020; Qiu et al. 2012; Winterer et al. 2019), retina (Huang et al. 2022; Laboissonniere et al. 2019), hypothalamus (Wang et al. 2019; He et al. 2020), and other nervous systems.

Here, we focus on the history of Patch-seq and its applications in the nervous system. We discuss its applications in non-nervous systems and higher mammals and its potential application to neurodegenerative diseases.

The Development of the Patch-seq Technique

The patch clamp technique was a great invention in the late twentieth century, which can study the function of single or multiple ion channels in living tissues (Segev et al. 2016). In 1976, Erwin Neher and Bert Sakmann of Max Planck Institute of Biophysical Chemistry, Germany, used a two-electrode voltage clamp to fill the microelectrode with acetylcholine on frog muscle cells. The microelectrode was in close contact with the cell membrane, and the single-channel ion current activated by acetylcholine was recorded (Neher and Sakmann 1976). This is the first time a single-channel ionic current has been recorded in humans. In 1991, Pei et al. performed whole-cell recordings in the visual cortex of anesthetized cats (Pei et al. 1991). In 1994, whole-cell recordings were used to examine the effect of GABAergic inputs on the orientation selectivity of a cat's V1 pyramidal neurons under anesthesia. This became one of the most widely used configurations of the patch-clamp days after (Nelson et al. 1994).

Whole-cell recording has now been used in many areas of animal tissue. For example, the brain, including the cortex (Markram 1997; Zhu and Connors 1999; Brecht and Sakmann 2002; Manns et al. 2004; Arroyo et al. 2018), hippocampus (Joëls et al. 2003; Kawamura et al. 2016; Moriguchi 2011; Fujiwara-Tsukamoto and Isomura 2008; Isomura et al. 2008; Liu et al. 2017), cerebellum (Zhu et al. 2005; Ishikawa et al. 2015; Duguid et al. 2015, 2012; Rancz et al. 2007), retina (Walston et al. 2015; Arman and Sampath 2010; Werner et al. 2008; Wu 1987) and spinal cord (Deng and Xu 2012; Arulkandarajah et al. 2021; Wu et al. 2018; Jiang et al. 2017), etc. Although whole-cell recording can be used to record the electrical activity, synaptic response, and excitability of ion channels in individual neurons, the transcriptional information of cells is still unknown.

The emergence of sequencing technology enables human beings to explore the essential differences of organisms directly. In the 1970s, Sanger and Coulson created the first generation of DNA sequencing methods—"Sanger sequencing," which promoted the progress of the human genome project (Sanger et al. 1977). However, due to its low flux and high cost, it is difficult to be applied widely. The second-generation sequencing technology, also known as Next Generation Sequencing (NGS) technology, emerged to improve the throughput of the first-generation Sequencing. Later, the research on nanopore single-molecule sequencing promoted the development of third-generation sequencing technology, reducing the cost of sequencing (Sanger et al. 1977; Pareek et al. 2011). However, a large amount of sequencing depth and sample number is still required, and the quality requirements for primers are high, so researchers began to look for more practical means.

To investigate the genetic coding of ion channels, researchers must isolate specific mRNAs from individual cells, a decidedly challenging task. With mRNAs accounting for only about 2% of cells and even fewer mRNAs coding for subunits of the relevant ion channels, isolation can be difficult. Although current PCR technology can detect individual molecules, the process also introduces a degree of contamination and signal-to-noise, requiring rigorous control and thorough documentation of the entire procedure.

At the end of the eighties the advent of molecular neurobiology with the genetic identification of many subunits of ionotropic receptors was an embarrassment for neurophysiologists which could not figure out the correct assembly of these numerous subunits to form a single heteromeric channel. The invention of scRT-PCR after patch-clamp was the elegant response to decipher the correct assembly in a single neuron. It was bringing together two technologies, patch-clamp and PCR in an unexpected way. A crucial connection between electrophysiology and molecular biology was made by this invention. Thanks to its simplicity scRT-PCR disseminated very quickly all over the world.

In the subsequent use of scRT-PCR, a large number of neurons in the retina and brain, as well as peripheral neuronal cells, including dopamine receptors, were studied (Yan and Surmeier 1997), and DABA receptors (Zhu et al. 1996). In the late 1980s, neurophysiologists also used this technique to study the correct subunit composition of receptors on cell membranes, and Lambolez et al. studied the subunit composition of AMPA receptors in large Purlirje cells (Lambolez et al. 1992; Yuste et al. 2020). Subsequent identified the subunits responsible for governing the voltage-dependence, ion selectivity, and kinetics of native glutamate, GABA, 5HT3, and nicotinic receptors (Klink et al. 2001; Tsuchiya et al. 1999; Rossier et al. 2015).

Since Jean Rossier lab applied scRT-PCR to the analysis of the expression profile of mRNA in neocortical interneurons. For years, neuroanatomists were trying to find a single marker defining a class of neurons, but coexistence of multiple transmitters ended this simple view (Vanderhaeghen et al. 1983). Present classification of interneurons is polythetic and relies on combination of multiple descriptors of a neural phenotype, anatomy, electrophysiology, embryonic origin, and the expression profile of mRNA. With scRT-PCR and more recently single cell RNA-Seq (scRNA-Seq) the large diversity of neocortical interneurons has been exposed leading to the characterization of at least 100 distinct classes of interneurons in the neocortex.

One of originality was to introduce non-hierarchic cluster analysis in the delineation of classes of interneurons (Cauli et al. 2000). In more recent publications, the specific roles of few classes have been defined; indeed, different classes of interneurons control local blood flow (Bruno Cauli et al. 2004), brain plasticity or long-term memory (Rossier et al. 2015).

Through scRT-PCR technology, researchers can obtain the target gene research, but it still cannot meet the need for neurobiologists to observe the epigenetic inheritance of proteins and genetic material, so sequencing technology has also entered the field of vision of biologists.

In 2009, Tang et al. proposed the single-cell RNA sequencing method for the first time (Tang et al. 2009). They sequenced the transcriptome of single blastomes and oocytes, making high-throughput RNA sequencing possible for the first time (Jovic et al. 2022). Since the introduction of scRNA-seq, the number of single-cell RNA sequencing experiments has dramatically increased, providing unprecedented opportunities to explore gene expression profiles at the single-cell level (Tang et al. 2009). ScRNA-seq has a higher resolution and is a powerful tool to identify and classify cell subsets compared to bulk RNA-seq. It allows us to understand cellular heterogeneity better, discover rare cell types, and place special cellular states in diseases (Zeisel et al. 2015; Li et al. 2023).

Since the start of the twenty-first century, Sc/snRNA-seq technology has progressed rapidly. Additionally, the partial sequence of complete proteins in a single neuron can be predicted from Sc/snRNA-seq data, providing a more robust tool for exploring neuronal differentiation, function, and network architecture. Jean Rossier have teamed up with scientists at the Allen Institute in a large scRNA-Seq study describing the wide expression of neuropeptides in neurons of the cerebral cortex. Each single cortical neuron expresses at least one neuropeptide and most cortical neurons co-express several neuropeptides. Importantly, the study identified a higher abundance and density of neuropeptides in the cortical circuit, presenting a novel direction for future research on cortical circuits (Smith et al. 2019).

Sc/snRNA-seq has become influential in discovering new or rare cell types and subtypes from different tissues and, for example, identified 78 endothelial phenotypes with varying profiles of the transcriptome, including Aqp7 + intestinal capillaries and angiogenic ECs in healthy tissues by analyzing single-cell transcripts of more than 30,000 endothelial cells (ECs) from 11 mouse organs, which provided the first single-cell transcriptome profile of mouse endothelial cells for future studies (Kalucka et al. 2020). Peng et al. used 165,000 sc/snRNA-seq maps to build a nonhuman primate retina map. In macaque monkeys, they performed a comprehensive cell classification of the fovea and peripheral retina, identifying about 60 cell types (Peng et al. 2019). Comparing macaque retinal types to mouse retinal types, a close correspondence was found between mouse and macaque PR, BC, and AC. Furthermore, by sc/snRNA-seq, more RGC subtypes were found. Karthik Shekhar et al. divided the collected 25,000 mouse retinal bipolar cells into 15 types, including 13 previously reported and 2 novel types, and established the most comprehensive transcriptome map of mouse bipolar cells (Shekhar et al. 2016). In the non-human primate study, Hao et al. (2022) analyzed 207,785 cells from the adult macaque hippocampus using an optimized sc/snRNA-seq pipeline (Wei et al. 2023) and identified 34 cell populations containing all major hippocampal cell types. Wei et al. employed a modified sc/snRNA sequencing technique to construct a comprehensive dataset of the macaque visual cortex, encompassing major cortical cell classes such as 25 excitatory neuron types, 37 inhibitory neuron types, and all glial cell types (Wei et al. 2022). The study identified specific markers for different layers of the cortex, including HPCAL1 and NXPH4. Additionally, two unique cell types were discovered: an excitatory neuron type expressing NPY and the dopamine receptor D3 gene and a primate-specific sensory neuron type, OSTN + , which is activity-dependent. Lau et al. described high-resolution sc/snRNA-seq profiles of cells derived from cynomolgus monkey testes, identifying genetic signatures that define spermatogonia populations (Lau et al. 2020). Therefore, sc/snRNA-seq technology has become an important method to study the map of animals, even humans and plants, which can provide important targets for clinical disease treatment.

In summary, developing sc/snRNA-seq technology has become a powerful tool to reveal the heterogeneity and complexity of different tissues at the cell level. Sc/snRNA-seq generates data with a high-dimensional quantitative expression of numerous genes for each neuron, and it is highly scalable as scRNA-seq data can be gathered from a vast number of individual cells, ranging from thousands to millions (Zeisel et al. 2018; Gouwens et al. 2020). Although it identifies subpopulations of cells within a tissue and provides information about expression patterns, it does not capture their spatial distribution or reveal the local communication networks between cells. However, it is still impossible to directly link the transcriptome with phenotypes such as morphology and electrophysiology. Future advances in technologies, linking the transcriptomic profile of neuronal cells to their neurophysiological and morphological phenotypes, contribute to creating an adequate neuronal taxonomy and open up new avenues to investigate neuroscience.

The Advances in Patch-seq Methods

To obtain the electrophysiological, morphological, and transcriptome information of neurons in the same cell to describe the characteristics of neurons more accurately, neuroscientists proposed Patch-seq technology by combining scRNA-seq technology and immunohistochemistry technology (Cadwell et al. 2016; János Fuzik et al. 2016). Patch-seq technology has been successfully applied to the cortex (Mahfooz and Ellender 2021; Oberst et al. 2019; Scala et al. 2019), hippocampus (Oláh et al. 2020; Qiu et al. 2012; Winterer et al. 2019), retina (Laboissonniere et al. 2019; Huang et al. 2022), and cultured neurons (van den Hurk et al. 2018) (Table 1).

Table 1.

Summary of the application of Patch-seq in neuroscience

Species	Tissue	Cell type	Major findings	Publication
Mouse	Visual cortex	Light-sensitive neurons	V1 neurons with a high expression of Rtn4r and Rgs7	Liu et al. (2020)
Mouse	Visual cortex	GABAergic neurons	28 met-types of cortical interneurons	Gouwens et al. (2020)
Mouse	Retina	Retinal ganglion cells	Prph, Ctxn3, and Prkcq as potential candidates for ipRGC classification	Laboissonniere et al. (2019)
Mouse	Retina	Retinal ganglion cells	Vat1l, Slitrk6, and Lmo7 differentially expressed among ON, OFF, and ON–OFF RGCs	Huang et al. (2022)
Mouse	Hippocampus	Pyramidal cells and interneurons	CAMs (Cell Adhesion Molecules) should differ among cell types, including RS-INT, FS-INT, CA1-PYR, BS-PYR, and RS-PYR cells	Földy et al. (2016)
Mouse	Hippocampus	VIP + projecting interneurons	VIP + cell gene expression (Chrna4, Adrb1 and so on) and proenkephalin being identified as an additional molecular marker of this VIP + cell type	Luo et al. (2019)
Mouse	Hippocampus	Interneurons	OLMs express Npy, OLM neurons lacked expression of the Htr3a gene, OLMs expressed MGE‐associated Lhx6, Satb1 and Sox6 transcription factors	Winterer et al. (2019)
Rat	Hippocampus	CCK + interneurons	TOR and RS were identified as two firing phenotypes of CCK + INs, the firing phenotypes were correlated with the presence of distinct isoforms of Kv4 auxiliary subunits (KChIP1 vs. KChIP4e and DPP6S)	Oláh et al. (2020)
Mouse	Neocortex	Layer 4 sensory cortex neurons	All excitatory neurons are pyramidal and all SOM + interneurons are Martinotti cells in V1, and in S1 the excitatory neurons are stellate and SOM + interneurons are non-Martinotti	Scala et al. (2019)
Mouse	Neocortex	Excitatory neurons	Translaminar clones labeled at E10.5 are composed of diverse classes of excitatory neurons (IT, PT, CT)	Cadwell et al. (2020)
Human	Neocortex	Pyramidal neurons	Three distinct types of human L5 neurons, respectively ET-like, IT-like 1 and IT-like 2	Kalmbach et al. (2021)
Human	Neocortex	Interneurons	Mapping to two cortical GABAergic SST t-types (SST CALB1 and SST ADGRG6)	Lee et al. (2023)
Human	Neocortex	L1 interneurons	Two human cell types with specialized phenotypes were identified (MC4R rosehip cells and the bursting PAX6 TNFAIP8L3 t-type)	Chartrand et al. 2023
Human	Neocortex	Glutamatergic neurons	Five human supragranular neuron t-types (LTK, GLP2R, FREM3, CARM1P1, and COL22A1) have corresponding morphology, physiology, and transcriptome phenotypes	Berg et al. 2021
Mouse	Primary motor cortex	Cortical neurons	Providing a morpho-electric annotation of almost all transcriptomically defined neural cell types (Vip, Pvalb, Sst and so on)	Scala et al. (2021)
Rat	Brainstem	Brainstem serotonin neurons	Identified six candidate gene markers for CO2-sensitive 5-HT neurons, including CD46 and Iba57	Mouradian et al. (2022)
Mouse	Hypothalamus	POA neurons	Identified Ptgds as a genetic marker of temperatum-sensitive POA neurons and L-PGDS expression as a molecular indicator	Wang et al. (2019)

The process of Patch-seq includes patch-clamp recording, sample collection, cDNA amplification and purification, library construction and sequencing, patch-clamp sequencing data analysis, and subsequent cell morphology recovery (Fig. 1).

Fig. 1.

Workflow of Patch-seq technique. Access to the intracellular compartments of individual neurons was obtained by whole-cell patch clamp, and the electrical properties of the cells were recorded. The intracellular contents were aspirated into a patch pipettor and collected in PCR tubes for downstream RNA sequencing. Tissue sections, which retained folded cell bodies and fine processes of the cells, were subjected to immunohistochemical staining to visualize the complex morphology of the cells

The first step of Patch-seq is patch-clamp recording, which records the electrophysiology of neurons from acute brain slices. To facilitate the aspiration of cell contents into the pipet, a slightly larger tips glass electrode is usually used to complete 10–15 min of patch-clamp electrophysiological. Typically, it includes firing pattern, spontaneous synaptic activity, and action potential threshold, waveform, pattern, etc. In addition, the glass electrode is loaded with less than the volume of internal solution used for ordinary electrophysiological recording. This helps avoid RNA dilution and reduces the sample’s surface area in contact with the glass electrode surface to reduce the loss of mRNA (Lipovsek et al. 2021; van den Hurk et al. 2018; Arman and Sampath 2010). After the electrophysiological recording, the cell contents were aspirated into the recording glass electrode by applying negative pressure. At the same time, a stable electrical seal was maintained between the pipette and the cell membrane, thus avoiding contamination of the extracellular solution.

The aspirated cell contents were then deposited into the lysis buffer or buffer containing RT reaction components, and the obtained RNA was immediately reverse-transcribed into cDNA. After cDNA quality control, appropriate concentrations and lengths of cDNA are selected for the next step of library construction. After PCR amplification, the library was constructed, and the cDNA library was sequenced. Because RNA is highly susceptible to degradation and the presence of small amounts of mRNA per cell, the Patch-seq protocol emphasizes the importance of maintaining strict RNase-free conditions during sample collection and before reverse transcription (RT). This includes careful cleaning of any surfaces that may come into contact with mRNA samples or internal solutions using RNase decontamination solutions and maintaining strict RNase-free conditions when preparing all pre-RT solutions (Lipovsek et al. 2021).

The data analysis workflow will depend on the specific research purpose, such as mapping Patch-seq transcriptome data into existing transcription profiles to study cell types (Lipovsek et al. 2021; Scala et al. 2019). The results were counted by software to obtain a transcriptome gene expression matrix. Cluster analysis was performed on the data, and machine learning and molecular informatics statistical methods were used to align the patch clamp sequencing data with the cell classification based on transcriptome definition established by high-throughput sequencing to find the position of the patch clamp sequencing data in the cell type map obtained by sc/snRNA-seq.

In addition, in Patch-seq, brain slices were fixed using a fixative after a recording was completed, followed by staining. To increase the success rate of staining, we controlled the osmotic pressure of the fluid in the electrode to allow rapid diffusion of biotin within the cell. The formulation of the fixative was also modified to allow rapid tissue fixation. Finally, the morphology of the successfully stained cells was reconstructed under a special microscope.

The Application of Patch-seq in Cortex

Large-scale sc/snRNA-seq studies can describe hundreds of transcriptome cell types defined by cluster analysis. Still, many of these cell types' corresponding morphological and electrophysiological phenotypes of many of these cell types have not been characterized, including many rare transcriptome cell types or newly defined cell types (Fig. 2). Patch-seq can combine sc/snRNA-seq to link the morphology, electrophysiology, and transcriptomics of single cells, allowing researchers to study the gene expression profile of single neurons and identify subpopulations of cells with different functional characteristics. For example, through characterized more than 1300 excitatory and inhibitory neurons in the mouse primary motor cortex, providing morphological, electrical annotations of almost all transcriptomically defined nerve cell types (Scala et al. 2021). Moreover, it was found that in the broad-spectrum transcriptome family (such as Vip, Pvalb, and Sst family), the morpho-electrophysiological characteristics of cells did not overlap with each other, and adjacent transcriptome cell types showed continuous but no obvious boundary changes in morpho-electrophysiological characteristics. A well-structured transcriptomic and morpho-electrical landscapes within families have been constructed. In addition, in the study of neurons in layer 4 (L4) of the neocortex, they found that in V1, all excitatory neurons were pyramidal, and all somatostatin-positive (SOM +) non-fast-spiking interneurons were Martinotti cells (Scala et al. 2019). In contrast, in the somatosensory cortex (S1), excitatory neurons were mostly stellate, and SOM + interneurons were non-Martinotti. They demonstrated that different cortical regions and transcriptome types have different morphological and electrophysiological phenotypes, and the circuit motifs of different cell types were mapped (Scala et al. 2019).

Fig. 2.

Application of Patch-seq and progress of Patch-seq technology. Patch-seq technology enables the compilation and integration of multimodal cell type atlases. The molecular maker can be identified. Patch-seq can also identify subpopulations of cells with different functional characteristics and species differences. Selected figure adapted from Huang et al. 2022, Allen et al. 2022, and Lee et al. 2023

Similarly, Cadwell et al. studied cell types and associated circuit diagrams of neocortical neurons (Cadwell et al. 2016) in mice and clonally related excitatory neurons (Cadwell et al. 2020). They found that translaminar clones labeled at embryonic day 10.5 (E10.5) comprise diverse transcriptomic subtypes of excitatory neurons, such as IT, PT, and CT. The mouse visual cortex analyzed the transcriptome and intrinsic physiological properties of more than 4200 GABAergic neurons, reconstructing the local morphology of 517 among them. These neurons are classified into 28 types based on morphological, electrophysiological, and transcriptomic data (Gouwens et al. 2020). In the primary somatosensory cortex, the division of layer 1 and layer 2 cortical interneurons and pyramidal neurons into different nerve subtypes was compatible with later morphological analyses, and a quantitative data set was generated (János Fuzik et al. 2016). In addition to studies in rodent cortex, Patch-seq has also been applied to biopsy tissues removed from surgical patients. For example, in humans' neocortical layer 1 interneurons of humans, subclasses of neurons characterized by their transcriptomes display distinct morpho-electrical phenotypes. Specifically, two distinct cell types, namely MC4R rosehip cells and the bursting PAX6 TNFAIP8L3 t-type, were identified, each exhibiting specialized phenotypic characteristics (Chartrand et al. 2023). Patch-seq data obtained from neurons in layers 1, 2, 3, and 5 of the human cortex were integrated into the human transcriptome cell reference map. The electrophysiological and morphological features are then assigned to the mapped T-type (Berg et al. 2021; Kalmbach et al. 2021; Chartrand et al. 2023). Human-specific double bouquet cells were identified, consistent with two cortical GABAergic SST t-types (SST CALB1 and SST ADGRG6) (Lee et al. 2023). Studied glutamatergic neurons in the supragranular layer of the human neocortex and identified a strong correspondence between morphological, physiological, and transcriptomic phenotypes of five neuronal types (LTK、GLP2R、FREM3、CARM1P1, and COL22A1 types) on glutamatergic granules (Berg et al. 2021). These results demonstrate the diversity of cell types in the human cortex and provide a structural basis for studying the complexity of the human cortex.

Therefore, Patch-seq can not only study the type of single cell but also integrate the morphology, electrophysiology, and gene expression of single cells to generate a multidimensional map of neuronal cells, which provides a basis for studying different cell types in different cortex regions.

In addition, Patch-seq can also identify molecular markers to target certain cell types for further study and elucidate the structure and function of neuronal circuits, generating new hypotheses about their differentiation. For example, in the study of photosensitive neurons in the visual cortex of mice, the three-dimensional (3-D) morphological characteristics of neurons were reconstructed and indicated that V1 neurons with a high expression of Rtn4r and Rgs7. It is demonstrated that synaptic transmission strength plays a key role in neurons of light-sensing V1 (Liu et al. 2020). Fuzik et al.studied the indusium griseum of mouse cortical glutamatergic neurons and showed that IG is a quasi-continuum of glutamatergic neurons (IG-VGLUT1 +), confirming neuronal connections associated with glutamatergic neurons and olfactory circuits (Janos Fuzik et al. 2019). In contrast, in humans, the neocortex is located in the ventricular and outer subventricular zones (VZ and OSVZ, respectively), identified ITGB4 and ANGPTL4 as candidate markers for dense spherical and dense smooth astrocytes. This study highlights the complex role of cell lineage in diversifying human neocortical astrocytes (Allen et al. 2022).

Applying Patch-seq technology can also highlight the conservation and divergence between species and cortical regions. Used Patch-seq technology to identify corticospinal neurons (Betz) in layer 5 of the primary motor cortex in non-human primates and humans and determined the morphological structure of Betz cells and showed that Betz cells had extratelencephalic-like physiology (Bakken et al. 2021). These findings identify and highlight the diversity of mammalian cell types and the specificity of the functions of different cell types. Analysis of L5 ET neurons from neurosurgical human and rodent neocortexs identified three distinct types, respectively ET-like, IT-like 1, and IT-like 2, highlighting the general conserved nature of the core-defining features of L5 ET and L5 IT neurons and phenotypic differences may be related to functional specialization of the human neocortex (Kalmbach et al. 2021). Indeed, observing a 'rosehip' cell type found in human and not mouse neocortex emphasizes the importance of studying human L1 neocortical interneurons to uncover potential species-specific specializations (Chartrand et al. 2023).

Patch-seq technology can also be used to study the structure and function of neural circuits. Patch-seq conducted on brain slices can be combined with optical tools or synapse-specific transneuronal tracking methods to query the transcriptome of neurons that receive input from or project to specific brain regions. For example, Pfeffer et al.used Patch-seq to explore and compare the cell types and spatial distribution of GABAergic neurons such as PV, SOM, and VIP, identified their differentially expressed genes of the mouse visual cortex, and demonstrated that L5 neurons that project to the frontal lobe or contralateral cortex are molecularly different (Pfeffer and Beltramo 2017). However, the gene expression patterns of L2 neurons are similar in their location, projection, or function. Ellender et al.studied excitatory neurons in the mouse somatosensory cortex and found that apical intermediate progenitors (aIPs) give rise to a limited number of upper cortical neurons that can share many functional and morphological properties with neighboring excitatory neurons from different progenitor pools (Ellender et al. 2019). However, there are significant differences in their postsynaptic partners when compared with neighboring neurons in the same layer.

Therefore, the development and application of Patch-seq technology are critical for understanding the relationship between gene expression in individual neurons and specific integration of circuits, providing new insights into the regulation of different brain regions and highlighting the importance of understanding the molecular and cellular basis of the brain.

The Application of Patch-seq in the Hippocampus

Patch-seq can be used to integrate multimodal atlases of cell diversity in the hippocampus. The hippocampal structure is also a complex multi-brain region structure, and neurons in different brain regions are connected to form a complex neural network. Patch-seq targeting can be used to record some cell types with specific functions, and the differences in transcriptome and morphology between them can also be studied. For example, PV neurons in mice hippocampus in the study of hippocampal brain slices revealed the differences in morphology, electrophysiology, and transcriptome of PV neurons at different developmental stages (Que et al. 2020). Recorded CCK + interneurons in the rat hippocampus and identified two different types (TOR and RS) of CCK + neurons with different excitability properties (Oláh et al. 2020). In the study of oriens-lacunosum moleculare (OLM) interneurons in the mouse hippocampus, two different functional subtypes of OLM (SstCre-OLM and Htr3aCre-OLM) were also found, which proved that Npy expression is a characteristic of OLM neurons in the mouse hippocampus (Winterer et al. 2019). Cells in the pathway of hippocampal FS interneurons (FS-int), RS-INT, CA1-PYR cells, and subiculum RS-PYR and BS-PYR cells and their gene expression profiles were analyzed (Földy et al. 2016). Patch-seq can also be used to study the projection of hippocampal neurons. In the study of mouse hippocampal VIP + projection interneurons explored the VIP + cell gene expression including Chrna4, Adrb1, and proenkephalin being identified as an additional molecular marker of this VIP + cell type, it was proved that VIP + hypochoritic projection cells share molecular properties with other VIP + and long-distance projection GABAergic neurons. This is important for the specific functions of these cells associated with their local and remote projection modes (Luo et al. 2019).

In total, subtle differences in transcriptomic profile may profoundly affect neuron morphology and function in the hippocampus. Using Patch-seq to identify a correlation between transcriptomes and sensory response type, neuronal position, and projection pattern in the hippocampus may provide new insight into the molecular basis in the hippocampus.

The Advances of Patch-seq in Other Fields

In addition to its application in the cortex and hippocampus, Patch-seq can also be applied in the retina, hypothalamus, olfactory bulb, and some neurons cultured in vitro in the nervous system. In the retina, the transcriptomic, morphological, and functional features of 472 high-quality RGCs were characterized by Patch-seq, providing functional and morphological annotations for many of the transcriptomics-defined cell types of the previously established RGC atlas. They identified Vat1l, Slitrk6, and Lmo7 differentially expressed among ON, OFF, and ON–OFF RGCs (Huang et al. 2022). Studies in the brain stem identified six candidate gene markers for CO₂-sensitive 5-HT neurons and two selected candidate genes (CD46 and Iba57) were all expressed in 5-HT neurons identified by in situ mRNA hybridization (Mouradian et al. 2022). Studies in the hypothalamus identified Ptgds as a genetic marker of temperature-sensitive POA neurons, and the increased level of L-PGDS expression in POA neurons that are sensitive to temperature can serve as a molecular indicator for further exploration of the neural pathway involved in thermoregulation (Wang et al. 2019). Mechanisms by which hypothalamic glucose-inhibitory neurons (GI-ERα^vlVMH) or glucose-excitatory neurons (GE-ERα^vlVMH) sense and respond to glucose fluctuations have also been identified (He et al. 2020). Using Patch-Seq in the striatum, Pthlh (the gene encoding for parathyroid hormone-related protein) cells were found to exhibit a continuum of electrophysiological properties associated with Pvalb expression. In addition, there are significant molecular differences related to differences in electrophysiological properties between Pvalb-expressing cells in the striatum and Pvalb-expressing cells in the cortex (Muñoz-Manchado et al. 2018). Similarly, the Patch-seq whole-transcriptome analysis approach has been validated by performing electrophysiological and genome-wide analyses on embryos from healthy subjects and human neurons generated from induced pluripotent stem cells (ESCs/iPSCs) (van den Hurk et al. 2018). Using weighted gene co-expression network analysis (WGCNA), gene clusters highly correlated with neuronal maturation as judged by electrophysiological features were identified. A strong link between neuronal maturation and genes involved in ubiquitination and mitochondrial function was revealed. In addition, a list of candidate genes was identified that have the potential to serve as biomarkers of neuronal maturation. The combination of electrophysiological recordings and single-cell transcriptome analysis will be a powerful tool to reveal the molecular logic of neural circuit function in the future (Chen et al. 2016). Studies in human iPSC-derived neurons and astrocytes have revealed a novel biomarker, GDAP1L1, that can effectively predict which neurons are highly functional. These biomarkers allow the functional classification of large numbers of neurons without patch-clamp and can be used to stratify functional heterogeneity (Bardy et al. 2016). The same study on astrocytes, which applied Patch-seq to the molecular analysis of electrophysiological characteristic Ascl1- and Neurog2-iNs, showed no overall correlation between the functional properties of heterology-induced neurons and the transcriptome, except for K⁺ channels (Kempf et al. 2021). In mouse DRG-cultured neurons, significant transcriptional changes were identified in single mechanosensitive DRG neurons expressing Piezo2, and differential gene expression was found across mechanosensory neuron types (Parpaite et al. 2021). Combining MS current recordings in DRG neurons with sc/snRNA-seq, these data make it possible to precisely distribute four different types of MS currents (NF200+ , IB4+ , Non-peptidergic, CGRP+) in molecularly defined populations of DRG neurons. Furthermore, Piezo2, Trpc1, and Fam155a are candidate genes for mechanotransduction complexes providing an open resource for further exploring cell type-specific determinants of mechanosensory properties (Michel et al. 2020).

Conclusions and Future Directions

Patch-seq techniques enable measuring multiple characteristics of individual neuronal cells, including transcriptomics, morphology, and electrophysiology in complex brains, to build single-cell multimodal data linking the transcriptomic profile of neuronal cells to their neurophysiological and morphological phenotypes.

Alternatively, we can record known cell types and classify them into different subtypes based on morphology and transcriptomics. It may be possible to link gene expression to a specific function, identifying within a single cell the underlying mechanisms of functional differences in other similar cell types. In combination with transgenic, viral, and optogenetic techniques, explore the transcriptional signature of neurons with specific developmental lineages, projecting to neurons in specific brain regions or receiving input from common brain regions (Cadwell et al. 2017a, b; Lipovsek et al. 2021; van den Hurk et al. 2018). This technology has great potential to solve many problems in neuroscience. It has made great contributions to identifying cell types and subtypes and has extraordinary prospects for exploring different regions of different species using this technology. Nonetheless, several important challenges remain, and future experimental questions will undoubtedly drive innovation in this protocol.

First, although Patch-seq has been extensively studied in the nervous system under physiological conditions, few studies have been conducted in neurodegenerative diseases. Based on retrieval, it was found that Patch-seq was performed in an in vitro HD model to study the effects of mutant huntingtin (Htt) on synaptic transmission and gene transcription in individual striatal neurons (Paraskevopoulou et al. 2021). Similarly, reprogrammed midbrain neurons from 42 individuals (80 batches, 5,315 single-cell RNA-seq, and 44 Patch-seq samples) to study the transcriptome signature of PD, and confirmed a synaptic impairment with patch-clamping and identified pesticides and endoplasmic reticulum stressors as the most significant gene-chemical interactions in PD (van den Hurk et al. 2022). These studies demonstrate that future Patch-seq research in neurodegenerative diseases can be applied to the molecular mechanisms of complex neurological diseases at the single-cell level and provide a platform for screening treatments for this disease.

Secondly, it is of great significance to establish multimodal maps covering the whole non-human primate brain, for monkeys share many physiological and anatomical features with humans, such as the organization of the brain. However, due to the scarcity of primate tissues, especially non-human primate neurodegenerative disease model tissues, most of the research on Patch-seq has focused on mice.

Finally, in addition to applications in the nervous system, there are also some applications in non-nervous systems, such as islets. Islet cell function and its transcriptome were mapped using Patch-seq in cells collected from the human donor pancreas of diabetic patients. At the same time, in another study, maturation, respiration, and receptor expression were linked to pancreatic α-cell function, providing a valuable resource for exploring islet physiological and genetic dysfunction in health and disease (Camunas-Soler et al. 2020; Dai et al. 2022).In addition, Patch-seq can also be applied to non-neuronal cells, such as iPSC-derived cardiomyocytes. Therefore, the broader application of Patch-seq in non-nervous systems is also significant.

Overall, Patch-seq is a versatile method that has been applied in many ways, and multimodal approaches to study cell type and function are becoming the norm. Future research on human and non-human primate tissues, neurodegenerative diseases, and non-nervous systems, developing genetic manipulation tools suitable for different regions and tissues, and integrating them into a huge data set will be of immeasurable value to promote our neuroscience research.

Author Contributions

LS conceived the review. LS, SMT, ZW, LY, TL and HZZ prepared the manuscript and table. All authors read and approved the final version of the manuscript.

Funding

This research is supported by research grants from the Natural Science Foundation of China (Grant Nos. 81961128021 and 81870682), the National Key R&D Program of China (Grant No. 2022YEF0203200), the Guangdong Provincial Key R&D Programs (Grant No. 2018B030335001) and the Science and Technology Program of Guangzhou (Grant Nos. 202007030011 and 202007030010).

Data Availability

Enquiries about data availability should be directed to the authors.

Declarations

Competing Interests

The authors declare no competing financial interests.