An in vitro whole-cell electrophysiology dataset of human cortical neurons

Abstract Background Whole-cell patch-clamp electrophysiology is an essential technique for understanding how single neurons translate their diverse inputs into a functional output. The relative inaccessibility of live human cortical neurons for experimental manipulation has made it difficult to determine the unique features of how human cortical neurons differ from their counterparts in other species. Findings We present a curated repository of whole-cell patch-clamp recordings from surgically resected human cortical tissue, encompassing 118 neurons from 35 individuals (age range, 21–59 years; 17 male, 18 female). Recorded human cortical neurons derive from layers 2 and 3 (L2&3), deep layer 3 (L3c), or layer 5 (L5) and are annotated with a rich set of subject and experimental metadata. For comparison, we also provide a limited set of comparable recordings from 21-day-old mice (11 cells from 5 mice). All electrophysiological recordings are provided in the Neurodata Without Borders (NWB) format and are available for further analysis via the Distributed Archives for Neurophysiology Data Integration online repository. The associated data conversion code is made publicly available and can help others in converting electrophysiology datasets to the open NWB standard for general reuse. Conclusion These data can be used for novel analyses of biophysical characteristics of human cortical neurons, including in cross-species or cross-lab comparisons or in building computational models of individual human neurons.


*Denotes equal contribution
Dataset Description Intracellular electrophysiology, as performed via the whole-cell patch-clamp technique, is a hallmark method for characterizing the biophysical features of neurons. While there have been numerous datasets characterizing these features from cortical neurons in the rodent brain [1]- [4], there are comparatively fewer resources that provide high-quality whole-cell patch-clamp recordings from human cortical neurons due to the relative inaccessibility of human tissue.
However, collaborations between neurosurgeons and basic neuroscientists have recently made it possible to characterize living cortical neurons in brain slices immediately prepared from biopsies following routine neurosurgery [5]- [15]. Still, there remain relatively few datasets of human cortical neuron physiology that are openly-accessible and free for reuse to complement and compare to the the Allen Brain Institute Cell Types Database (http://celltypes.brainmap.org/) [7].
Here, we describe an openly-accessible dataset of electrophysiological recordings from human and mouse cortical neurons. The dataset encompasses 141 whole-cell patch-clamp recordings from surgically resected human tissue (129 cells from 36 individuals) or from 21-day old mice (12 cells from 5 mice). These datasets are made available in the Neurodata Without Borders (NWB) electrophysiology data format via the DANDI data archive. We provide morphological reconstructions for N = 7 cells, made available at NeuroMorpho.org. Each recording is made available with rich subject and experimental protocol metadata, enabling subsequent reuse and comparison with analogous datasets from other species and sources.  [19]. Patients underwent a standardized temporal, parietal or frontal lobectomy under general anesthesia using volatile anesthetics for seizure or tumor treatment [20,21]. Tissue was obtained from patients diagnosed with temporal (n=34), frontal (n=1) or parietal lobe (n=1) epilepsy or brain tumors (n=4) in 18 male and 18 female patients, age ranging 21 to 59 years (mean age ± SD: 40.22 ± 12.00).
The resected cortical tissue from the temporal lobe-middle temporal gyrus (MTG) exhibited no structural or functional abnormalities in preoperative MRI, and is considered "relatively healthy" by ourselves and others as it is located outside of the site of epileptogenesis [6,9,12,13].
Cortical tissue from the frontal cortex from patients with epilepsy are considered as "epileptogenic" tissue and are confirmed with independent electrocorticography (ECoG) (and are annotated as such in our metadata). For tumor cases, cortical tissue blocks were obtained from tissue at a distance from the main site of the tumor (i.e., such cortical tissue was not taken directly from the tumor itself).

Consent to publish
Written informed consent were obtained from all study participants to use their tissue and to share the acquired data with anonymized demographic information -namely, subject age at time of surgery, sex, years of seizure, seizure frequency, secondarily generalized seizure frequency (using clinical records and epilepsy monitoring unit recordings), antiepileptic drug treatment, and type of seizure. Acute brain slice preparation Immediately following surgical human cortical resection, the cortical specimens were submerged in an ice-cold (~4°C) cutting solution that was continuously bubbled with carbogenated (95% O2/5% CO2) aCSF containing (in mM) sucrose 248, KCl 2, MgSO4.7H2O 3, CaCl2.2H2O 1, NaHCO3 26, NaH2PO4.H2O 1.25, D-glucose 10. The osmolarity was adjusted to 300-305 mOsm. Transverse brain slices (400 μm) were sectioned using a vibratome (Leica 1200 V) in cutting solution. Tissue slicing was performed perpendicular to the pial surface to help ensure that pyramidal cell dendrites were minimally truncated [6], [11], [17]. The cutting solution was the same as used for transport of tissue from the operating room to the laboratory.
Whole-cell patch-clamp recording from human and mice cortical slices

Axon Binary Format to Neurodata Without Borders file conversion
The x-to-nwb repository was used to convert current clamp recordings in axon binary format (ABF) to Neurodata Without Borders (NWB) format. Separate converters were used for files recorded using pClamp 9.0, which output ABFv1 files, and pClamp >10.0, which output ABFv2 files, to ensure valid conversions while incorporating the essential metadata. The key aspects of our usage of these data conversion computer scripts relate to defining which ABF channels correspond to stimulus and response traces and ensuring that appropriate scale and offset factors are applied properly upon conversion. We incorporated the ndx-dandi-icephys metadata extensions to allow for inclusion of user-defined 'Subject' and 'Lab' metadata fields to be able to include specific metadata including 'subject_id', 'age', 'species', 'cell_id' and 'tissue_sample_id'. and 'RMP' to record the resting membrane potential at the initial time of recording. Additional recording metadata was extracted directly from ABF files using custom scripts to extract the stimulus start and end times and the stimulus sampling rate.

Electrophysiology feature extraction
The Intrinsic Physiology Feature Extractor (IPFX) toolbox was used to extract features from converted NWB files [3], [20]. All experiments consisted of long-square hyperpolarizing and depolarizing current injections, and extracted features include subthreshold features (i.e., input resistance, sag ratio), action potential properties (i.e., action potential half-width, threshold time and voltage) derived from the rheobase spike as well as multi-action potential spike train features derived from the IPFX-defined "hero" sweep (i.e., adaptation index), as described previously [6]. Our included metadata files contain stimulus start and end times along with an IPFX-compatible stimulus description ontology file for reproducibility and to facilitate the feature extraction process.

Results
In Table 1, we summarize the three main axes differentiating the cells and recordings in this dataset. Namely, recordings differed by species (human versus mouse), by cortical layer of the cell body of the recorded neuron (Layer 2&3, Layer 3c, and Layer 5), and whether synaptic blockers were used in the external recording solution. Putative interneurons were identified by their action potential characteristics (large maximal firing rates and typically large spike after-hyperpolarization amplitudes) as described in Chameh, et al 2021. One reason why synaptic blockers were used is to make a subset of recordings more consistent with protocols used in other labs, such as the Allen Institute for Brain Sciences [3], [11].  and most depolarizing current injection (green).

Dataset
In Fig. 3A,  To compare the effect of solution used for the brain slice preparation on intrinsic properties, we compared the input resistance and sag ratio recorded following preparation in either solution.
Due to the confounding of cutting or slice preparation solution used and cortical layer resected we highlight recordings using synaptic blockers in aCSF recording solution from L2/3 and L5 of human cortex. In Fig. 4A, we highlight a consistent trend of higher measured input resistance in the recordings made following preparation in the NMDG recovery solution compared to the sucrose solution and note that the comparison in L5 differed significantly (L2/3   To illustrate the rich diversity of the metadata for each of the human recordings, in Fig. 5 we highlight specific comparisons of input resistance and sag ratio measurements recorded in regular aCSF across demographic conditions. Specifically, we focus on the input resistance as a fundamental passive electrophysiological property and the sag ratio as an active property that has previously been used to distinguish between subtypes of human neurons [6], [11]. In Fig. 5B, 5E we compare distributions of these electrophysiological features measured in male and female patients (Fig. 5B, E); whether the recorded cell was resected from patients with a tumor or not (Fig. 5 C, F); and across the three different brain lobes from which neuronal tissue was resected (Fig. 5A, D). Note that all cells from frontal and parietal cortex were recorded from tissue resected near the site of the epileptogenic focus, whereas all cells from temporal cortex were recorded distal from the epileptogenic focus (with the exception of 1 subject). Additionally, we illustrate the relationship of the input resistance and sag ratio against both patient age at time of surgical resection (Fig. 6A, B) and years of seizure experienced by the patient prior to the surgical intervention (Fig. 6C, D).  Other associated metadata for recordings can be found at the following link: https://github.com/derekhoward/nwb_conversion/blob/master/data/processed/meta/all_meta_cle aned.csv

Application Scenarios
The recordings in this database permit the quantification of biophysical properties from a diverse set of neurons, including human and mouse neurons with a well described set of metadata.
Independent variables collected include age, sex, seizure history and cortical layer from which the tissue was resected. Additionally, experiments on the human neurons were performed with the use of synaptic blockers and without, allowing for comparisons and integration with other intrinsic electrophysiological databases comprising patch-clamp recordings, including from the the Allen Cell types database (https://celltypes.brain-map.org/).
These data from current-clamp experiments are particularly beneficial for the development of conductance-based models of human neurons [11], [21], [22]. In particular, we highlight that in some instances, it may be more suitable to constrain biophysical models to human data in the presence of synaptic blockers, that is to say, when background synaptic activity is having a significant effect on input resistance measurements. The voltage responses can be used as a training set to constrain biophysical models when combined and integrated with other publicly available databases that provide relevant morphologies and channel kinetics, such as NeuroMorpho.org, Channelpedia or ICGenealogy [23]- [25]. Moreover, fitting biophysical models to data that is grouped based on demographic information can allow for cross-group comparisons using in silico approaches. Usage of these models in neuronal or circuit simulations can thus help to further predict and unveil the potential effects of differences in neuronal physiology across demographic groupings [26], [27].

Discussion and Limitations
The repository provided is focused mainly on cortical neurons derived from human tissue. There are comparatively fewer recordings for analyses of mouse neuron function provided and all of these were performed using synaptic blockers that were shown to have a baseline effect on input resistance.
The recordings from human specimens derived from tissue during the surgical resection of diseased tissue for patients with intractable epilepsy or brain tumors. Along with having suffered seizures for an extended period of time, the patients may have concurrently taken one or a variety of anti-epileptic drugs that could have affected baseline neuronal excitability characteristics.
While these data were collected for the purpose of characterizing intrinsic properties of human neocortical neurons, we note that they were collected using different sets of experimental conditions, including those related to different recording solutions as well as cutting solutions.
Our analyses suggest such experimental condition differences likely contribute to differences in downstream electrophysiological properties, such as input resistances being higher among cells recorded in the presence of synaptic blockers of excitatory and inhibitory neurotransmission.
The observed effect of synaptic blockers on the input resistance is due to reduction of overall membrane permeability as a consequence of the block of both excitatory and inhibitory conductances but does not distinguish whether they have equal effect [28]. In contrast, the comparisons of electrophysiological measures following different cutting solutions highlight potential effects on neuronal excitability. We observe variability in the measured input resistance but consider that these effects may be due to changes in conductivity across the membrane or also experimental biases in selection of healthy neurons for patch-clamp protocol due to differential response to solutions of different osmolarity. Furthermore, there exists contrasting results in the literature on the effects of NMDG on neuronal excitability and synaptic transmission which shows the context-dependence of the many experimental variables [29], [30]. Taken together, such potential differences in electrophysiological characteristics due to experimental conditions are important to consider when re-using these data in downstream analyses.