A transgenic mouse embryonic stem cell line for puromycin selection of V0V interneurons from heterogenous induced cultures

Background Spinal interneurons (INs) relay sensory and motor control information between the brain and body. When this relay circuitry is disrupted from injury or disease, it is devastating to patients due to the lack of native recovery in central nervous system (CNS) tissues. Obtaining a purified population of INs is necessary to better understand their role in normal function and as potential therapies in CNS. The ventral V0 (V0V) INs are excitatory neurons involved in locomotor circuits and are thus of interest for understanding normal and pathological spinal cord function. To achieve scalable amounts of V0V INs, they can be derived from pluripotent sources, such as mouse embryonic stem cells (mESCs), but the resultant culture is heterogenous, obscuring the specific role of V0V INs. This study generated a transgenic mESC line to enrich V0V INs from induced cultures to allow for a scalable, enriched population for future in vitro and in vivo studies. Methods The transgenic Evx1-PAC mESC line was created by CRISPR-Cas9-mediated insertion of puromycin-N-acetyltransferase (PAC) into the locus of V0V IN marker Evx1. Evx1 and PAC mRNA expression were measured by qPCR. Viability staining helped establish the selection protocol for V0V INs derived from Evx1-PAC mESCs inductions. Immunostaining was used to examine composition of selected inductions. Cultures were maintained up to 30 days to examine maturation by expression of mature/synaptic markers, determined by immunostaining, and functional activity in co-cultures with selected motor neurons (MNs) and V2a INs on microelectrode arrays (MEAs). Results V0V IN inductions were best selected with 4 µg/mL puromycin on day 10 to 11 and showed reduction of other IN populations and elimination of proliferative cells. Long-term selected cultures were highly neuronal, expressing neuronal nuclear marker NeuN, dendritic marker MAP2, pre-synaptic marker Bassoon, and glutamatergic marker VGLUT2, with some cholinergic VAChT-expressing cells. Functional studies on MEAs showed that co-cultures with MNs or MNs plus V2a INs created neuronal networks with synchronized bursting. Conclusions Evx1-PAC mESCs can be used to purify V0V IN cultures for largely glutamatergic neurons that can be used in network formation studies or for rodent models requiring transplanted V0V INs. Supplementary Information The online version contains supplementary material available at 10.1186/s13287-022-02801-7.

have limited regenerative capacity. To reconnect the lost circuits, cell transplantation is a potential method of introducing cells in and around the injury site to support growth of spared neurons through release of neurotrophic factors and for cell replacement. For cell transplantation, several cell lineages have been investigated to determine their contribution to repair (see this review for more information: [1]), but the propriospinal, interneuron (IN) populations have been shown to create local connections post-injury [2][3][4]. INs are classified as dorsal or ventral, which contribute largely to sensory or motor function, respectively [5], and several studies have worked toward deconstructing the role of the different IN subtypes and the interplay between them, which drives normal function [6], as well as their potential as therapeutic populations [1]. Many of these studies utilized knockdown or knockout animal models, but such methods may obscure the functional contribution of a population through possible redundancies or compensations inherent to spinal circuits [7,8], requiring additional, alternative methods using particular IN populations in isolation. Such isolation would allow for more controlled studies but would require some method of purification for the population of interest from its source.
V0 INs are of interest as a therapeutic population. They comprise a diverse set of INs that arise from a Dbx1 + progenitor pool [9,10], and include excitatory, ventral V0 (V0 V ) INs, which express transcription factor Evx1 [11], and inhibitory, dorsal V0 (V0 D ) INs, with both ipsilateral and commissural projecting axons. V0 INs contribute to left-right alternation [12,13], with V0 D INs shown to contribute more at lower locomotor frequencies and V0 V INs contribute at higher frequencies in rodents [13]. A subset of V0 V INs expresses the transcription factor Pitx2 and includes an ipsilateral population that project onto motor neurons (MNs) [14]. These Pitx2 + cells include the cholinergic V0 C and glutamatergic V0 G populations, with V0 C involved in task-dependent excitation of MNs [14]. The contribution of V0 V INs as an excitatory population involved in locomotor circuits, including cells that project onto MNs, is therefore of great appeal for cell replacement therapies.
For an abundant source of INs, one could differentiate them from pluripotent stem cells (PSCs), including patient-derived induced PSCs (iPSCs) and embryonic stem cells (ESCs), into the desired neuronal type. Induction protocols to obtain neurons from PSCs, however, are not completely efficient-often with less than half the culture comprising the cell type of interest-thus requiring a method to purify the intended population from the heterogenous culture. For example, our recent induction protocol developed to generate V0 V INs from mouse ESCs (mESCs) yields approximately 40% of the cells expressing V0 V IN markers 2 days post-induction [15]. To purify the cell population of interest from such heterogenous inductions, our laboratory has generated transgenic mESC lines that use an IN-or MN-specific gene marker to drive expression of a puromycin resistance enzyme, puromycin-N-acetyltransferase (PAC) [16][17][18][19]. As the induced mESCs become the intended INs expressing their specific gene marker, it drives expression of PAC and thus renders those cells insensitive to puromycin exposure, while other cells in the culture do not survive. In this article, a mESC line used to select V0 V INs by insertion of PAC into the Evx1 locus is analyzed for efficiency of selection, remaining proliferative cells and other IN populations. Maturation and electrophysiological characterization of selected V0 V IN populations are also assessed.

Design and generation of vectors used to create Evx1-PAC selectable mESC line Evx1 guide RNA vectors
A list of potential guide RNA sequences to the Evx1 locus was generated by providing a partial Evx1 genomic DNA input sequence to the algorithm found on the CRISPOR website [20] (results can be found here: http:// crisp or. tefor. net/ crisp or. py? batch Id= mN1Fk RligA ThyqK v3eg3). Two guide RNA oligonucleotide sequences were selected based on their proximity to the Evx1 transcript start codon as well as low number of off-target sequences listed in CRISPOR results.
Guide RNA oligonucleotides were designed and generated according to the Joung lab gRNA cloning protocol provided on the Addgene website [21] by annealing phosphorylated single-stranded oligonucleotide sequences (ordered from Integrated DNA Technologies; see Vectors for Evx1 guide 21 and Evx1 guide 28 (see Table 1) were created by ligating the annealed oligonucleotide sequences with the digested MLM3636 vector using T4 DNA ligase for 16 h at 16 °C. Ligations were transformed in DH5α E. coli and selected on 2% agar in Lennox LB broth (MilliporeSigma, L3022) containing 50 µg/ mL ampicillin (MilliporeSigma, A0166: resuspended as a 50 mg/mL stock in water).

Evx1-PAC donor vector
PCR with Klentaq LA (DNA Polymerase Technology, 110) was used to amplify both the PAC-PGK-Neo cassette from existing vectors used to create other selectable lines in our laboratory [18,19] as well as genomic DNA sequences flanking the Evx1 guide RNA sequences for 5′ and 3′ Evx1 homology arms using genomic DNA template from the RW4 mESC line (see Table 1 for primers). The homology arm primers included attB sites to allow for Gateway recombination cloning. Overlap extension PCR was used to amplify the 5′ homology arm with the PAC cassette and then the 3′ homology arm. BP clonase (Thermo Fisher Scientific, 11789020) was used to recombine the homology arm-flanked PAC overlap extension PCR product into the pDONR221 vector (Thermo Fisher Scientific, 12536017). DNA from bacterial clones selected on LB-agar plates containing 50 µg/mL Kanamycin (Thermo Fisher Scientific, BP906: resuspended as a 50 mg/mL stock in water) were isolated using QIAprep Spin Miniprep Kit (Qiagen, 27106) and sequence-verified (Sanger sequencing at UT Austin core facility), then recombined into the pWS-TK3 vector (Addgene, 20349) using LR clonase (Thermo Fisher Scientific, 11791020) according to manufacturer instructions. Bacterial clones containing the final pWS-TK3-Evx1-PAC donor vector were selected on LB-agar plates containing ampicillin and isolated using QIAprep Spin Miniprep Kit. After pulsing, cells were quickly moved into CM +LIF +BME and grown in a 0.1% gelatin-coated 10 cm nontissue culture treated dish for 2 days. From days 2 to 10, electroporated cells were grown in CM +LIF +BME with 40 µg/mL Geneticin/G418 (Thermo Fisher Scientific, 10131027) for positive selection and 150 nM fialuridine (MilliporeSigma, SML0632) for negative selection; medium was replaced every 2 days. At day 10, visible single colonies were picked and dissociated using trypsin-EDTA into 96 well 0.1% gelatin-coated cell culture plates, quenched with CM +LIF +BME, and once confluent, dissociated into two wells of a 96 well plate; one well was used to genotype clones for insertion of PAC into the Evx1 genomic locus using junction PCR; see Table 1 for primer sequences). Clones positive for PAC by jPCR were also tested by copy number assay using qPCR with Tert (Thermo Fisher Scientific, 4458368) as the reference control, GAPDH (Thermo Fisher Scientific, Mm00186825_cn) as the reference for 2 copies in RW4 mESCs, and a PAC custom assay (forward sequence: GGT GCC CGC CTT CCT; reverse sequence: CGG CGG TGA CGG TGAA) using Hb9-puro mESCs as a reference for 1 copy and analyzed using the Copy-Caller v2.1 software. Clones with single copies of PAC were expanded in different well sizes until 80% confluent in T75 flasks (Thermo Fisher Scientific, 07202000), and then, cells were dissociated in trypsin-EDTA, quenched with CM, pelleted at 300 × g, resuspended in cell freezing medium (MilliporeSigma, C6295), and frozen in a Mr. Frosty freezing container (Thermo Fisher Scientific, 51000001) at − 80 °C before storage in liquid nitrogen.

Cre excision of PGK-Neo
In one Evx1-PAC mESC clone verified by junction PCR (jPCR) and copy number assay, the PGK-Neo selection cassette flanked by loxP sites was excised by transfecting the mESCs with pTurbo-Cre (a gift from Dr. Timothy Ley) using Lipofectamine 3000 (Thermo Fisher Scientific, L3000001). The day before transfection, Evx1-PAC mESCs were dissociated using trypsin-EDTA, quenched with CM, and then 5 × 10 5 cells were plated in a 6-well plate well and grown in CM +LIF +BME. For transfection, 3.75 µL of Lipofectamine 3000, 10 µL of P3000 Reagent, 5 µg of pTurbo-Cre plasmid were mixed in Opti-MEM medium (Thermo Fisher Scientific, 31985062) and added to the mESC culture. Once colonies were visible, clones were picked and dissociated into 96-well plates. Each was evenly split among two wells to test for sensitivity to G418 added to one of the two wells to observe cell death. Those that showed cell death were chosen for jPCR to look for the presence of Neo in the Evx1 locus. Two clones that no longer contained the PGK-Neo cassette were chosen, expanded, frozen down, and one was used for all further studies.
For selection conditions, cultures were exposed to 4 µg/ mL puromycin dihydrochloride (puro; MilliporeSigma, P8833: resuspended as a 10 mg/mL stock in water) in supplemented neuronal medium. After 24 h, medium was aspirated, cells were rinsed with NB and cultured in neuronal medium with supplements replaced every two days. For longer-term studies, after day 12, cultures were fed with half-volume exchanges of long-term medium every two days until their respective end points.

V2a IN and MN induction and selection
Chx10-puro-tdTomato mESCs were induced using a 2 − /4 + protocol using 10 nM RA and 1 μM purm from days 2 to 6 and 5 μM DAPT for days 4 to 6 [22]. On day 6, V2a IN EBs were settled, induction medium aspirated, and neuronal medium with supplements was used to move EBs back to the 10 cm dish. 2 µg/mL puro was added to the culture for 24 h. Hb9-puro mESCs were induced using the MN 2 − /4 + induction protocol using 2 μM RA and 0.5 μM SAG from days 2 to 6 [23]. Day 6 EBs were cultured in neuronal medium with supplements containing 4 µg/mL puro from day 6 to 7. 2) and coating with 5 µg/mL laminin (mouse; Thermo Fisher Scientific, 23017015; in HBSS). To measure activity, CytoView plates were loaded into a Maestro Edge instrument (Axion Biosystems) for recordings using AxIS Navigator software.

Microelectrode array co-cultures of selected V0 V INs with selected mESC-derived V2a IN and MNs
V0 V IN induction cultures were selected on day 10 with 4 µg/mL puro in supplemented neuronal medium for 24 h. On day 11, selection medium was aspirated, wells were washed with NB, and medium was replaced with supplemented neuronal medium. On day 12 of V0 V IN culture, V0 V INs were lifted with Accutase solution (Mil-liporeSigma, A6964), quenched with CM, spun at 400 × g and resuspended in 1 mL neuronal medium with supplements. Day 6-7 selected cultures of V2a INs and/or MNs were dissociated with 0.25% trypsin-EDTA, quenched with CM, strained through 100 µm cell strainers, spun at 400 × g, and resuspended in 1 mL neuronal medium with supplements. All populations were then counted to achieve cell densities of 1 × 10 5 cells/10 µL drop to load onto MEA wells, which have a 22.9 mm 2 surface area; for single-population MEA cultures, cells were plated at 1 × 10 5 cells/array, for double-population cultures, each population was plated at 5 × 10 4 cells/array, and for triple-population cultures, each population was plated at 3.3 × 10 4 cells/array. After allowing attachment for 4 h, 300 µL of long-term medium was added slowly to each well; medium was replaced by half-volumes every 2 days.
To detect spontaneous firing of action potentials in MEA cultures, AxIS Navigator software parameters were set in the spike detector. Peak detection with adaptive threshold was used with a set threshold of 5× standard deviation to detect only negative inflections of spikes that crossed this threshold. Durations for pre-and post-spike detection were set to 1 ms and 2 ms, respectively, and the spike detection interval was set to 1.6 ms to enable collection of 20,000 samples for each spike.

Isolation of RNA, reverse transcription, and qPCR
To collect cultured cells for qPCR analysis, medium was aspirated and cells were detached by addition of 0.25% trypsin-EDTA, followed by quenching and dissociation in CM. Cells were pelleted at 300 × g. Medium was aspirated, and pellets were resuspended in RLT buffer as provided from the RNeasy Mini Kit (Qiagen, 74106). Pellets were either frozen at − 80 °C or immediately used with the RNeasy kit using on-column DNase (Qiagen, 79254) to isolate RNA per manufacturer instructions. 500 ng of RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, 4368813) per manufacturer instructions.
For qPCR, a solution of ultrapure water, TaqMan Fast Advanced Master Mix (Thermo Fisher Scientific, 4444963), a TaqMan probe against mouse β-actin as a reference gene (Thermo Fisher Scientific, Mm02619580_ g1, using VIC-MGB_PL dye), and the TaqMan probe against the target gene using FAM-MGB dye (PAC [custom ordered]: forward sequence: GGT GCC CGC CTT CCT; reverse sequence: CGG CGG TGA CGG TGAA; Evx1: Mm00433154_m1) was prepared and loaded into a MicroAmp Fast Optical 96-Well Reaction Plate (Thermo Fisher Scientific, 4346906) before loading each sample in triplicate. Plates were sealed, spun briefly to remove bubbles, and loaded into the QuantStudio 3 instrument for measurement. The fold changes in mRNA expression levels were calculated using 2^− ΔΔCT values with β-actin as the reference gene relative to mESCs.

LIVE/DEAD staining to determine selection window by image analysis
Cultures were grown in 48-well plates for image analysis. To test selection, 4 µg/mL puro was added on days 9, 10, or 11 for 24 h, followed by a rinse with NB before culture in neuronal medium with supplements for an additional 2 days post-selection before staining and imaging.
The LIVE/DEAD Viability/Cytotoxicity Kit for mammalian cells (Thermo Fisher Scientific, L3224) was used with Calcein AM at 0.4 µM (live stain, 1:10,000 dilution in PBS) and Ethidium homodimer-1 at 1 µM (dead stain, 1:2000 dilution in PBS) to stain cultures that included unselected cells and cells selected at different time points. Cells were incubated for 30-40 min before rinsing once with PBS, then imaging.
Images were analyzed using a CellProfiler pipeline [24,25] to determine the percentage of living cells. Puro results in early termination of protein translation to lead to cell death, resulting in cell debris that is removed during media changes, therefore the majority of dead cells were washed away post-selection. 4 images were taken of each well for each condition at each time point with N = 6. The percentage of living cells was calculated based on the count of live cells in selected wells divided by the count of live cells in unselected wells, as all wells were plated at the same density of 1 × 10 5 cells/cm 2 .

Flow cytometry
Day 11 cultures, both unselected and selected with 4 µg/ mL puro from day 10-11, were dissociated and spun at 400 × g for 5 min, and pellets were fixed in 4% paraformaldehyde (PFA; MilliporeSigma, P6148; in 0.1 M phosphate buffer) for 20 min at room temperature. Cells were then incubated in 2% normal goat serum (NGS; MilliporeSigma, G9023) with 0.1% Triton X-100 (Mil-liporeSigma, X100) in PBS to permeabilize and block for 30 min at room temperature. Primary antibodies (see Table 2) were diluted in 1% NGS in PBS and cells were stained for 40 min at room temperature. Cells were washed 2 times for 10 min per wash with PBS. Secondary antibodies (see Table 3) were diluted in 1% NGS in PBS and cells were stained for 40 min at room temperature. Cells were washed 2 times for 10 min per wash and then resuspended in PBS before running samples on Attune NxT cytometer (Thermo Fisher Scientific).

Immunocytochemistry and image analysis
Long-term cultures in 48-well, laminin-coated plates were used for immunocytochemistry (ICC) analysis. After aspirating culture medium, wells were rinsed with PBS, and cells were fixed in 4% PFA for 20 min at room temperature. Cells were then incubated in 2% normal goat serum (NGS; MilliporeSigma, G9023) with 0.1% Triton X-100 (MilliporeSigma, X100) in PBS to permeabilize and block for 30 min at room temperature. Primary antibodies (see Table 2) were diluted in 1% NGS in PBS and cells were stained overnight at 4 °C. Wells were washed 3 times for 10 min per wash with PBS. Secondary antibodies (see Table 3) were diluted in 1% NGS in PBS and filtered through a 0.22 µm PVDF syringe filter (Mil-liporeSigma, SLGV033RS). To prevent fluorescence photobleaching of conjugated secondary antibodies, plates were wrapped in foil, while cells were incubated for 1 h at room temperature. Secondary antibody solutions were removed, and cells were washed 3 times for 10 min per wash with PBS. Cells were then stained in 1:1000 Hoechst 33258 (Thermo Fisher Scientific, H3569) in PBS for 10 min, washed once with PBS, then imaged on a DMi8 inverted widefield microscope (Leica).

Statistics
GraphPad Prism version 7 and Microsoft Excel were used for statistical analyses. In Prism, ROUT method with Q = 1% was used to remove outliers-this method initially generates a Lorenzian distribution and uses a robust curve fit and Q as a false discovery rate to determine outliers from the residuals of the fit, and then for the remaining data, it uses least-squares regression to determine any additional outliers. Values are reported as means and error bars are standard error of the mean (S.E.M.) unless otherwise stated. One-way analysis of variance (ANOVA) using Scheffe's multiple comparison method with 95% confidence was used to determine significance, which is indicated in figures as follows: * for p < 0.05, **

Inserting PAC into a single Evx1 gene locus
V0 V IN marker-expressing cells and other propriospinal IN populations can be induced from mESCs, but the resulting culture is heterogenous and can include proliferative cell types that will dilute the post-mitotic neurons. To better characterize the derived V0 V IN population for use in controlled in vitro studies and potentially as a transplantable population in rodent SCI models, a method of purification is desirable. To this end, a gene for an enzyme that confers protection against puromycin (puro), PAC, was inserted into the locus of Evx1, the distinguishing marker for V0 V INs. The goal was to use CRISPR/Cas9-mediated homology-directed repair (HDR) to insert the PAC selection marker gene into a single locus to confer puro resistance while maintaining Evx1 expression to enable V0 V IN induction (Fig. 1). Three plasmids were electroporated into wildtype RW4 mESCs to promote HDR: one with a guide RNA (gRNA) to direct Cas9 to an appropriate site in the Evx1 locus to create a double-stranded break (DSB), one encoding the SpCas9 enzyme, and a vector to act as donor DNA to insert PAC into the locus during HDR (Fig. 1A). The donor vector included homology arms-with sequences homologous to the genomic locus of Evx1 adjacent to the guide RNA sequences in the genome-flanking PAC and floxed PGK-Neo, a positive selection marker; the vector backbone also included a negative selection marker, thymidine kinase 3 (TK3) to select against random, non-HDR-mediated insertion. After Cas9 generated a DSB at the site specified by the Evx1 gRNA, the donor vector provided the donor DNA for DNA repair mechanisms to perform HDR (Fig. 1B). Junction PCR (jPCR) using a primer aligned to Evx1 genomic sequence upstream of the insertion and a primer aligned to the PAC sequence was used to confirm insertion into the Evx1 locus (Fig. 1C). To ensure that the primary transcript being expressed was PAC in the generated Evx1-PAC mESCs, the floxed PGK-Neo sequence was excised through Cremediated recombination; Cre was introduced by transfecting positive Evx1-PAC mESC clones screened by jPCR and determining their sensitivity to G418 (Neomycin). G418-sensitive clones were screened by jPCR using a primer aligned to Neo and a primer aligned to Evx1 genomic sequence downstream of the cassette insertion to confirm removal of PGK-Neo (Fig. 1C). Using screened clones of Evx1-PAC and Neo-excised Evx1-PAC (Evx1-PAC n.e.) mESCs from jPCR, a copy number assay was performed to ensure that only one Evx1 locus was modified (Fig. 1D). A single verified clone was used in all further experiments and is hereafter referenced as Evx1-PAC in all data and discussion.

Determining puromycin selection window
Previously we found mESC-derived V0 V IN inductions have the highest fold increase in Evx1 mRNA expression over uninduced cultures at day 8, and the greatest percentage of cells as Evx1 + /Lim1 + /βIII tubulin + neurons occurs at day 10 [15]. As PAC is inserted into the Evx1 locus, Evx1 gene regulatory elements drive expression of PAC, and Evx1 mRNA expression is possibly altered. Both Evx1 and PAC mRNA expression levels were measured to determine whether insertion of PAC resulted in its expression while maintaining expression of Evx1 (Fig. 2). Evx1 mRNA was found to be significantly decreased in Evx1-PAC-derived V0 V IN cultures relative to RW4derived V0 V IN cultures ( Fig. 2A). However, image analysis comparing RW4-and Evx1-PAC-derived V0 V IN cultures on day 11 showed similar levels of Evx1 + /Lim1 + / βIII tubulin + cells (Fig. 2B), with a trend toward a greater percentage of these cells present in Evx1-PAC-derived inductions (25.12 ± 3.58% vs 38.53 ± 8.16% for RW4 vs Evx1-PAC, respectively; p = 0.14) demonstrating that even with a reduction in Evx1 mRNA expression, production of V0 V IN marker-expressing cells did not have a corresponding reduction. Selection of mESC-derived V0 V IN cultures should kill mESCs and other cell types while sparing V0 V INs. This requires that PAC be expressed at a high enough level to achieve selection. In alignment with the window of peak Evx1 mRNA and protein expression, PAC mRNA was examined on days 8, 9, 10, and 11 in V0 V IN induction cultures (Fig. 2C). PAC mRNA expression was found to have the highest fold increase, around 20-fold relative to mESCs, at day 8. To ensure that the achieved PAC mRNA expression level was sufficient for selection, Evx1-PAC mESC cultures and V0 V IN cultures were examined for sensitivity to puro (Fig. 2D). The mESCs died in the presence of 4 µg/mL puro. In selected V0 V IN cultures, some living cells remained at a lower cell density relative to Cre-mediated recombination removes the Neo cassette so that only PAC remains in the Evx1 locus. C jPCR images showing the wildtype RW4 mESCs as a control, with PAC insertion in both Evx1-PAC and Neo-excised (n.e.) Evx1-PAC, as well as presence of Neo in Evx1-PAC and absence in Evx1-PAC n.e. mESCs. D Copy number assay using the previously established Hb9-puro mESC line as a control for one copy of PAC and RW4 mESCs as a control for two copies of GAPDH as a reference. Evx1-PAC and Evx1-PAC n.e. mESCs show one copy of PAC. Error bars show the range of possible copies, not S.E.M unselected cultures, showing cell death of some of the heterogenous cultured cells.
After verifying that puro killed mESCs and spared some of the induced cells, the time period for selection was established. 4 µg/mL puro was added to cultures for 24 h on days 9, 10, or 11. After a further 2 days of culture, selected cultures were compared to unselected control cultures plated at the same density at the same time by LIVE/DEAD staining and imaging. A CellProfiler image analysis pipeline was used to determine counts of living cells-only expressing LIVE stain-and the percentage of cells surviving selection was determined by using the  (Fig. 2E). Selection on day 10 to 11 or on day 11 to 12 yielded approximately the same survival percentage, around 15%. Based on this survival percentage and the coinciding peak number of V0 V IN marker-expressing cells occurring at day 10, cultures were exposed to 4 µg/mL puro on day 10-11 as "selected" cultures for subsequent experiments.

Examining purity of selected V0 V IN cultures
Selection should ideally spare only V0 V INs in the induction cultures, but selection of transgenic cell line-derived neurons by puro does not completely purify the desired population [18,19]. Evx1 is transiently expressed in vivo and in vitro, and in mESC-derived cultures, the number of cells co-expressing Evx1 and Lim1 that are also identified as neurons from positive βIII tubulin staining decreases dramatically by day 12 [15]. Therefore, although some cell death still occurs over the next couple of days (data observed but not shown; readers are referred to [23]), V0 V IN induction cultures were examined by ICC immediately after selection to look at V0 V IN markers on day 11 (Fig. 3A, B) to capture the presence of Evx1 + cells. During selection, neuronal cells were observed as having fewer and/or shorter neurites, which is not unexpected, as many of the surrounding cells in contact with the surviving cells are dying from exposure to puro. This morphological effect could persist for a day or two before cultures recovered post-selection, and connections remained sparse among the resultant less densely populated surviving cells. At day 11, there were still some Evx1 + /Lim1 + /βIII tubulin + co-expressing cells in both unselected and selected cultures as measured by ICC. Selected cultures appeared to have cells expressing higher levels of Evx1 protein based on observed intensity of staining. Additional file 1: Figure S1 shows quantification for day 11 unselected and selected cultures stained with Evx1, Lim1, and βIII tubulin, suggesting that selection may increase the proportion of Evx1 + cells in these cultures. However, the percentage of V0 V IN-marker expressing cells was also measured by flow cytometry for unselected and selected cultures immediately after selection on day 11 (Fig. 3E), and there was a trend toward a decrease in the percentage of Evx1 + /Lim1 + cells in selected versus unselected cultures (18.04 ± 5.04% vs. 32.04 ± 5.04% for selected vs unselected, respectively; p = 0.062).
These heterogenous V0 V IN induction cultures were previously found to have large proportions of non-V0 V INs belonging to the populations arising adjacent to V0 V INs during development [15]. Markers for these populations were measured by flow cytometry to examine whether they were reduced or eliminated after selection: Dmrt3 for dI6 INs, En1 for V1 INs, Chx10 for V2a INs, Hb9 for MNs, and Olig2 for pMNs and oligodendroglial precursors were used (Fig. 3F). Most of these markers were expressed in very few cells, possibly due to a high percentage of these cells having already become postmitotic. As such, they may have already downregulated their transiently-expressed definitive transcription factor markers, as induction protocols for other mESC-derived propriospinal neurons yield their respective transcription factor marker-expressing post-mitotic populations after 6 to 8 days [22,[26][27][28]. For the most adjacent populations of dI6 and V1 INs, Dmrt3 + cells significantly decreased from 24.16 ± 6.33% to 8.02 ± 2.57% (p = 0.035), while En1 + cells decreased from 1.65 ± 0.78% to 0.61 ± 0.25% (p = 0.161). V0 V INs include a cholinergic Pitx2 + subpopulation, V0 C INs, which form monosynaptic connections with MNs. Therefore, co-expression of Pitx2 and vesicular acetylcholine transporter (VAChT) was also measured by flow cytometry in selected cultures to compare against unselected cultures (Fig. 3E). There was no difference in the percentage of cells expressing these markers (6.19 ± 2.20% in unselected vs. 6.36 ± 1.54% in selected, p = 0.952), even with the measured decrease in cells expressing V0 V IN markers.
Proliferative cell types can dilute the post-mitotic neuronal population and result in reduced purity over time and possibly result in teratoma formation after transplantation. Ki67 was used to mark proliferative cells. ICC on day 16 showed few Ki67 + cells in unselected cultures (Fig. 3C), while little to no stained cells were shown after selection (Fig. 3D). βIII tubulin was co-stained with Ki67, which showed that neuronal cells did not co-express Ki67, and after selection, neurons persisted, while proliferative cells were removed. Flow cytometry data of day 11 cultures showed that Ki67 + cells significantly decreased from 7.06 ± 1.80% to 0.61 ± 0.22% ( Fig. 3E; p = 0.019). βIII tubulin + cells showed similar or slight increases (80.59 ± 4.94% vs. 85.54 ± 3.80%, p = 0.337). Interestingly, the percentage of neuronal cells increased in Evx1-PAC-derived inductions compared to RW4-derived cultures, which had 61.72 ± 7.89% βIII tubulin + cells (data not shown; N = 3, p = 0.095 vs unselected Evx1-PAC mESC-derived βIII tubulin + cells), perhaps due to altered neuronal specification related to reduced Evx1 expression.

Observing maturation of long-term selected V0 V IN cultures
To determine whether selected V0 V IN cultures showed maturation over time, cells were examined by ICC for mature neuronal markers and synaptic markers on days 16 and 22 (Fig. 4). Fox3 + post-mitotic neuronal nuclei marked by NeuN, microtubule-associated protein 2 (MAP2) as a dendritic marker of maturing neurons, and vesicular glutamate transporter 2 (VGLUT2) for glutamatergic neurons were used to stain cultures (Fig. 4A,  B). All three markers were found to be co-expressed in a large proportion of the cells that had survived selection and long-term culture, with clearer processes marked by MAP2 by day 22. Additional file 1: Figure S2A shows quantification of the proportion of cells expressing VGLUT2, NeuN, and MAP2 in day 22 selected cultures. Image analysis shows around 40% of the cells co-expressing VGLUT2 with neuronal markers MAP2 or NeuN.
Potential V0 C INs were detected in selected day 16 and day 22 cultures by expression of VAChT; these cells also showed co-expression with the pre-synaptic marker Bassoon and dendritic marker MAP2 (Fig. 4C, D). As flow cytometry data suggest that there are very few or no remaining MNs in selected V0 V IN cultures, while Pitx2 + /VAChT + cells remained, the observed incidence of VAChT + cells additionally supports that there may be V0 C INs present. Based on image analysis of ICC data, as seen in Additional file 1: Fig. S2B, around 2% of the selected cells co-express MAP2 and VAChT.

Measuring selected V0 V IN culture activity
Combinations of single population, double population, or triple population co-cultures of mESC-derived selected cultures of V0 V INs, MNs, and V2a INs were examined for electrophysiological activity by MEA recordings. Two trials were conducted out to day 30, with one plate having obvious glial presence (+glia), as medium color changed quickly at the later days of culture. Glia were not intentionally included in these selected cultures, but as with any experiment, variability occurs and may have led to selection conditions being more permissive to glial survival. Glial presence was determined by medium color change by rapid acidification and visual assessment of cell morphology, which did not occur in cultures with only neurons present. Acidification did not allow for culture beyond 30 days without requiring more frequent media changes to avoid over-acidification of the cultures. Although not intended, cultures with glia were maintained because of the observed differences in firing activity.
Some but not all populations showed bursting activity. Bursting patterns at day 26 for those populations that had  (Fig. 5A); this pattern is also observed in the MN: V2a IN co-cultures (Fig. 5B). In MN:V0 V IN cultures, there is a longer train of bursting before any distinct, shorter bursts are observed (Fig. 5C). When MN:V2a IN:V0 V INs are cultured together, this burst train still occurs but is shorter and is followed by the distinct short bursts seen in V2a IN cultures (Fig. 5D). These distinct patterns support the idea that different excitatory populations-V2a and V0 V INs-are present and communicating differently with MNs, and when all are present, both IN populations contribute to the network circuitry. This sort of synchronized bursting was observed less frequently, if at all, in the MEA cultures with little glial presence (not shown). Although cultures without glia had less synchronized bursting, patterns reminiscent of alternation occurred in cultures that contained V0 V INs (Fig.  5E-H). These "alternations" were observed from around day 24 to around day 28. Figure 6 shows activity, network, and synchrony metrics from day 22, 26, and 30 MEA recordings. Activity and synchrony metrics are shown for all population combinations, while bursting metrics are only shown for cultures that displayed bursting, which did not include V0 V INs, MNs, or V2a:V0 V IN co-cultures. To highlight the differences in firing for cultures with no obvious glia versus those with surmised glial presence, data are presented for each replicate rather than as means. The mean firing rate (MFR) of either MNs or V0 V INs alone or V2a IN:V0 V IN co-cultures did not increase above what occurred at the initial plating (on day 12) over time (Fig. 6A), which was sporadic and therefore could be considered background just above the set threshold for spike detection. Otherwise, the general trend was that MFR increased over time, and MFR for cultures in the presence of glia were higher than without glia. MN (Fig. 6B). Spikes per burst for V2a IN cultures +glia and co-cultures with MNs +glia remained level over time. Burst durations for cultures +glia also decreased or remained level (Fig. 6C), coinciding with the reduced number of spikes per burst over time. These data also align with the burst frequency remaining level over time (Fig. 6D). Of note are that by day 30, the V2a IN culture burst frequency increases, while the MN:V0 V IN cocultures have an increased burst duration and spikes per burst; these data may allude to the IN populations' roles in the network.

Discussion
CRISPR/Cas9 technology was used to generate a DSB for HDR-mediated insertion of PAC into a single Evx1 locus. Two Evx1 gRNAs, guide 21 and guide 28, were designed and delivered to electroporated mESCs, but the clones derived from guide 28 resulted in much higher levels of PAC mRNA expression in the mESCs (data not shown), which made it difficult to detect a fold change increase in PAC mRNA expression in V0 V IN inductions relative to mESCs. Likely, this would have resulted in poor selection of V0 V INs from cultures and/or remaining mESCs post-selection. Therefore, clones chosen for further screening were all from electroporation with guide 21, which produced little PAC mRNA in mESCs.
After insertion of PAC into Evx1, there was a significant decrease in the amount of Evx1 mRNA expression in V0 V IN inductions derived from the Evx1-PAC mESC line compared to RW4-derived inductions. However, there was a considerable, although not significant, increase in the proportion of induced cells co-expressing the V0 V IN markers Evx1, Lim1, and pan-neuronal marker βIII tubulin (Fig. 2B). There was also an increase in the percent of βIII tubulin + cells in Evx1-PAC mESC-derived cultures compared to RW4 mESC-derived cultures, as measured by flow cytometry (data not shown). This suggests that disruption of one Evx1 allele results in altered neuronal specification, but the mechanism behind this change is unclear and requires further study. Previous work using mutated Evx1 showed it has involvement in anteroposterior patterning through associated BMP/Wnt pathway signaling effects [29]. Evx1 mutation resulted in downregulated expression of BMP and Wnt pathway genes; as the proteins of these genes are known to be involved in dorsal IN specification [30], perhaps Evx1-PAC mESCs can more easily generate ventral IN subtypes, resulting in the observed increase of V0 V IN marker-expressing cells.
The puro selection window was determined to be 24 h from day 10 to 11 or day 11 to 12. All analyses for selected culture composition, maturation, and functional recordings were completed with day 10 to 11 selected cultures. It is possible the results of these analyses would differ if day 11 to 12 cultures were examined instead, such as by allowing more time for V0 C IN specification before selecting, for example. Future work is needed to determine if the alternate selection window changes the measured outcomes for these cultures.
Evx1-expressing cells appeared to be enriched in selected versus unselected cultures as observed by ICC (Fig. 3A, B, and Fig. S1). However, flow cytometry data showed that there was an obvious, although not significant, decrease in Evx1 + /Lim1 + cells in selected cultures. It is possible that the robust Evx1 expression observed by ICC is due to those Evx1 + surviving cells expressing relatively higher levels of Evx1 compared to others that succumbed to puro, and that the discrepancy between ICC and flow data is due in part to this relatively higher expression as well as the transient expression of Evx1 known to occur in these cells [11].
To further illuminate, consider the following potential explanation. On day 11, Evx1 is expressed in about 30% of the induced cells, some with higher expression levels than others (this percentage would be slightly higher at day 10 and lower at day 12 due to transient Evx1 expression). This is likely due to the use of EB-based inductions resulting in "waves" of V0 V IN specification, the foremost wave starting around day 8, followed by successive waves that build up to yield the peak of cells expressing V0 V IN markers at day 10, and continue until about day 12 when marker-expressing cells wane. Of the 30% Evx1 + day 11 cells, those from earlier waves might already be turning Evx1 expression off, thus having a concomitant decrease in PAC expression. These cells expressing lower levels succumb to puro, while those with higher Evx1/ PAC levels from waves closer to the time of selection survive. Some of the V0 V INs that survived selection might, at the time of assaying, have low Evx1 expression or have already turned off Evx1 expression, while the ones retaining Evx1 expression have high levels of expression. Therefore, flow cytometry measurements might not be a true representation of the surviving percentage of V0 V INs these cultures. Further work is needed to verify that the surviving cells are truly V0 V INs, perhaps by finding targets downstream of Evx1 and examining their expression in these cells or by generating an alternative, lineagetraceable mESC line.
Although verification of selection yielding purification of cells expressing definitive V0 V IN markers was not achieved, ICC staining showed that surviving cells were largely glutamatergic, as are V0 V INs, and included some VAChT + cells as putative V0 C INs. Previous work shows the largest proportion of non-V0 V IN marker-expressing cells in these cultures are Dmrt3 + or En1 + neurons [15]. These data are indicative that surviving neurons are V0 V INs, as dI6 INs and V1 INs are both inhibitory populations [31,32].
In MEA functional studies, of the single population cultures, V2a INs were the only population that achieved bursting. This could possibly be due to survival of other propriospinal cell types, such as MNs, V2b INs, or even V0 V INs, which arise close to V2a INs and have been shown to be part of the heterogenous resultant mESC-derived V2a IN culture, even after selection with 2 µg/mL puro, albeit at very low percentages [18,22]. However, V2a INs purposely co-cultured with either V0 V INs or MNs did not achieve bursting, except for with MNs when glia were present. Co-culture with MNs +glia also did not achieve as high a level of activity as V2a INs cultured alone, suggesting that inhibitory populations such as V2b INs are not likely to be present, and possibly, if there are MNs in the V2a IN only cultures, the proportion of MNs must remain low relative to V2a INs to achieve higher bursting activity. Also, previous work in our laboratory showed that these mESC-derived selected V2a IN cultures could synapse with each other [18]. Studies examining the electrophysiological characteristics of V2a INs in vivo described three electrophysiological classes of V2a INs, and those of the same class can undergo electrical coupling and fire rhythmically during fictive locomotion [33]. This also aligns with the observed bursting pattern of V2a IN single population cultures having distinct, rhythmic short bursts at the late stage of the network burst (Fig. 5A), as well as contributing a rhythmic bursting pattern to MN co-cultures (Fig. 5B). Examining recordings of the different combinations of selected V2a INs, MNs, and V0 V INs allowed some light to be shed on their role in the generated networks. Even with variable data within and among groups, a general idea of the activity of the different populations can be gleaned. The presence of glia appears to support network formation, as cultures with glia showed improved maturation based on MFR, burst metrics, and greater network synchrony; these results could be anticipated based on previous work showing glial support of network formation and maturation [34]. The difference in bursting for cultures containing V0 V INs with (Fig. 5C, D) versus without glia (Fig. 5E-H) could suggest that some population of V0 V IN contributing to alternation is not as reliant on glial support for maturation, while another population that contributes to burst trains observed in cultures +glia requires glial support for survival and/or maturation. It is exciting to think these could be V0 C INs, which are involved in MN output modulation [14], as the burst trains suggest some level of burst modulation through the extended burst duration. ICC shows VAChT + cells are present; however, further work is needed to verify these are the cells that contribute to this bursting pattern in these cultures. More replicates, both with and without glia, might help further elucidate MN and IN functions in MEA networks, but a recent analysis of MEA-based data suggests that many-perhaps even an impractical number, depending on the complexity of the culture and activity level achieved-replicates would be needed to obtain statistically relevant data due to batch-to-batch variability and other factors influencing MEA measurements [35].
In studies examining CPG circuitry using spinal cord preparations, the native circuitry is organized, resulting in clear pairs of alternators, whether flexor-extensor or left-right pairs. However, in these in vitro cultures, there are not necessarily distinct pairs but, as appears to be the case observed in Fig. 5F, H, there could be three or more alternators communicating with each other. This may seem counterproductive if these cells were to be transplanted into injured spinal cords with the aim of improving motor function, since motor functions require pairing and additional nodes may be introduced by these cells. However, it might be possible that the remaining organized tissue after injury would incorporate and lend organization to any transplanted cells, especially since supportive glia and other circuitry components like inhibitory INs may contribute to and guide their activity post-transplantation.
In this paper we have shown the successful creation of a Evx1-PAC transgenic mESC line with a single modified locus to use as a tool in selected IN circuit studies and for transplantation studies in rodent models. V0 V IN markers were detected, although not at enriched levels after selection-likely due to transient Evx1 expression. The surviving INs were largely neurons expressing glutamatergic marker VGLUT2 and showing maturation and synaptic markers in long-term cultures. A proportion of maturing cells also showed expression of VAChT, a marker for the V0 V subpopulation, cholinergic V0 C INs, which normally modulate MN output. In terms of functionality, the selected Evx1-PAC mESC-derived V0 V cultures formed connections with MNs or V2a INs and MNs that resulted in synchronized network bursting in the presence of glia and ostensible alternating activity in V0 V IN co-cultures without glia. This transgenic mESC line is a valuable tool for controlled experiments exploring locomotor CPG circuitry and potentially as a therapeutic option in animal studies.

Conclusions
This work has shown the successful generation of a transgenic Evx1-PAC mESC line to allow for purification of a glutamatergic neuronal population. After selection with puromycin from day 10-11, cultures show removal of other IN populations and proliferative cells, thus enriching cultures for V0 V INs. Long-term selected cultures express mature neuronal markers, and when co-cultured with MNs and V2a INs, these selected cultures show functional network activity. These data recommend this cell line as a useful tool for in vitro network studies and in vivo rodent models investigating recovery.