Examining the role of the surfactant family member SFTA3 in interneuron specification

The transcription factor NKX2.1, expressed at high levels in the medial ganglionic eminence (MGE), is a master regulator of cortical interneuron progenitor development. To identify gene candidates with expression profiles similar to NKX2.1, previous transcriptome analysis of human embryonic stem cell (hESC)-derived MGE-like progenitors revealed SFTA3 as the strongest candidate. Quantitative real-time PCR analysis of hESC-derived NKX2.1-positive progenitors and transcriptome data available from the Allen Institute for Brain Science revealed comparable expression patterns for NKX2.1 and SFTA3 during interneuron differentiation in vitro and demonstrated high SFTA3 expression in the human MGE. Although SFTA3 has been well studied in the lung, the possible role of this surfactant protein in the MGE during embryonic development remains unexamined. To determine if SFTA3 plays a role in MGE specification, SFTA3-/- and NKX2.1 -/- hESC lines were generated using custom designed CRISPRs. We show that NKX2.1 KOs have a significantly diminished capacity to differentiate into MGE interneuron subtypes. SFTA3 KOs also demonstrated a somewhat reduced ability to differentiate down the MGE-like lineage, although not as severe relative to NKX2.1 deficiency. These results suggest NKX2.1 and SFTA3 are co-regulated genes, and that deletion of SFTA3 does not lead to a major change in the specification of MGE derivatives.


Introduction
During early embryonic development of the mammalian telencephalon, the transcription factor NKX2.1 is highly expressed in the medial ganglionic eminence (MGE), a subpallidal structure of the ventral forebrain, and also expressed in the pre-optic area, septum, and hypothalamus [1][2][3]. The MGE and caudal ganglionic eminence (CGE) are transient embryonic structures that are the primary source of GABAergic inhibitory progenitors, which migrate tangentially to target sites in the cortex. These progenitors then differentiate into a number of diverse inhibitory interneuron subtypes that modulate the activity of excitatory projection neurons in the cerebral cortex [3][4][5][6][7][8]. Expression of the homeobox protein NKX2.1 is a requirement for specification of the MGE and its derivatives. NKX2.1 deficient mice display gross malformations of the ganglionic eminences and a complete loss of specific MGEderived subtypes such as parvalbumin (PV) and somatostatin (SST) -expressing interneurons (Butt et al., 2008;Du et al., 2008;Ohkubo et al., 2002). The experimental downregulation of NKX2.1 in the ventral subpallium results in a conversion of MGE to CGE fates. In NKX2.1 conditional loss-of-function studies in mice, an increase in the generation of vasoactive intestinal polypeptide (VIP) and calretinin (CR)-expressing interneurons derived from the CGE is produced at the expense of MGE subtypes [9][10][11]. These results indicate NKX2.1 is a master regulator that establishes the MGE and promotes specification of interneuron subtypes. In a previous study, we described the ability of hESC-derived MGE-like interneuron progenitors to differentiate into GABAergic interneurons when co-cultured long term in vitro with mouse astrocytes (Chen et al, 2016). To identify gene candidates with expression profiles similar to NKX2.1 that could also help specify the MGE lineage, we utilized previously published data from our laboratory comparing the RNA-seq-based transcriptome of FACS isolated human embryonic stem cell (hESC) -derived NKX2.1-positive progenitors to NKX2.1-negative cells. This analysis showed that the profile of surfactant associated 3 (SFTA3) expression closely followed NKX2.1, suggesting it was a novel MGE marker that could be involved in interneuron specification [12].
SFTA3, which encodes surfactant protein H, is part of the multifunctional surfactant gene family, implicated in immune host defense and regulation of alveolar surface tension during normal respiratory mechanics of the lung. To date, there are four surfactant proteins, A, B, C, and D, that have been extensively characterized in the lung. Surfactant proteins A and D are associated with the collectin gene family and are implicated in immunoregulatory host defense. These proteins contain a lectin domain to allow surfactant binding of viruses, fungi, and bacteria, facilitating opsonization for phagocytic digestion and removal [13,14]. In contrast, SP-B and SP-C are hydrophobic proteins required for stabilization of the air-liquid interface at the lung surface to prevent collapse of the alveoli [15,16]. Recently, SFTA3 was identified by bioinformatics as a novel secretory surfactant protein expressed in the human lung [17]. Interestingly, SFTA3 shares very little sequence or structural similarity when compared to other surfactants or proteins in general. The BLAST search tool algorithm in conjunction with the Uniprot protein database revealed SP-H homologs in primate species only, including humans. No matches were found when additional BLAST searches were performed comparing the SP-H sequence to 3D biochemical structures listed in the Protein Data Bank [17].
Numerous sequence-based prediction tools used to identify post-translational modification (PTM) sites in the SP-H sequence suggested a high probability of palmitoylation, glycosylation, and phosphorylation. These data suggest that SFTA3 is an amphiphilic protein that can acquire both hydrophobic and hydrophilic properties [17]. Western blot analysis using an anti-SP-H antibody in the A549 alveolar lung cell line detected a distinct band for the expressed protein at~13kDa [17]. Similar to the other proteins in the surfactant family, studies demonstrate that SFTA3 has a role in innate immune response localized to the lipid plasma membrane surface [17,18]. Computer simulations investigating the binding affinity of SP-H with dipalmitoylphosphatidylcholine (DPPC), the most prevalent lipid in pulmonary surfactant, illustrated protein stability for SP-H at the lipid surface [19]. These data suggest PTMs are responsible for the amphiphilic properties of SFTA3 in specifying either surface regulatory or immune defense function.
SFTA3 is a single copy gene immediately adjacent to NKX2.1 on chromosome 14q13.3. This location, combined with their correlated temporal and spatial patterns of expression suggest that NKX2.1 and SFTA3 may regulate common developmental pathways. For example, NKX2.1 is also expressed during early development of the lungs and is implicated in promoting the production of surfactants in alveolar cells. The production of surfactant is perturbed upon disruption of the NKX2.1 gene [20]. Moreover, patients with mutations in SFTA3 display a variety of aberrant symptomology including choreoathetosis, hypothyroidism, and neonatal respiratory disease [21]. Approximately 50% of patients with mutations in NKX2.1 develop the same clinical phenotypes of motor ataxia and respiratory distress, all part of a larger connected network of disorders known as "brain-lung-thyroid syndrome" (OMIM 610978) [22]. The mouse orthologue of SFTA3 is NKX2.1-associated noncoding intergenic RNA (NANCI). Recent studies demonstrate a regulatory role for NANCI in NKX2.1 expression in the mouse lung [23,24]. Intriguingly, whereas the human SFTA3 gene contains an apparent open reading frame that is translated, mouse NANCI encodes a long non-coding RNA with no apparent open reading frames. The potential interaction between NKX2.1 and SFTA3 is largely unexamined, and a function for SFTA3 outside of the lung has not been established.
Using quantitative PCR (qRT-PCR) analysis, we now show upregulation in SFTA3 gene expression during differentiation of hESC-derived progenitors to an MGE-like fate. The Brain-Span Atlas, an open source database using RNA-sequencing to profile cortical and subcortical structures at various time points during early embryonic development, indicates SFTA3 expression is selectively upregulated in the MGE at 8-9 weeks gestation. We generated SFTA3 and NKX2.1 knock-out (KO) hESC lines using CRISPR-Cas9 genome editing in order to examine two fundamental questions. First, to determine if NKX2.1 and SFTA3 genes are coregulated such that the deletion of one gene will modify the expression pattern of the other. Second, to determine if SFTA3 serves a functional role in the specification of MGE GABAergic progenitors and their differentiation into mature inhibitory interneuron subtypes. An NKX2.1 KO cell line, expected to be deficient in specifying MGE-derived interneuron subtypes, served as a control to determine SFTA3 function in specifying MGE lineage identity. The deletion of

Generation of KO cell lines
A dual sgRNA-directed gene knockout approach using CRISPR-Cas9 was used to remove a portion of the SFTA3 gene. One guide sequence was designed to cut upstream of exon 2 and a second guide sequence was designed to cut downstream of exon 2. Deletion of exon 2 resulted in an early termination and a severely truncated protein. For deletion of NKX2.1, two sgRNAdirected CAS9 nuclease were targeted to flank the entire gene to remove all isoforms of NKX2.1, completely removing all exons from the genome. Guide RNA sequences were cloned into addgene plasmid #62988. The dual sgRNA vector was electroporated into H9 hESC cells using a Gene Pulser X (250 V, 500 uF). Cells were plated and after 24 hours, puromycin was supplemented into the medium at 1 ug/ml. Puromycin was added for a total of 48 hours to select for the transient expression of the dual sgRNA vectors. After 14 days, clones were isolated, expanded and tested by PCR. Clones were screened for deletion, inversion, and zygosity by PCR. Double KO clones were further expanded and sequenced across the junction to confirm genomic deletion. The same procedures described above were performed to derive control lines with the exception that no gRNA sequence was used during vector electroporation.

Culture of ESCs and MGE-like cells
hESC SFTA3 and NKX2.1 control and KO cell lines were maintained and passaged as previously described [25]. Neural differentiation of ESCs was initiated using the ALK2/3 inhibitor LDN-193189 (Stemgent, 100 nM) and progenitors ventralized to an MGE-like identity by a combination of sonic hedgehog (R&D Systems, 125 ng/mL) and its agonist purmorphamine (Calbiochem, 1 μM) as previously described [26].

In vitro maturation of hESNPs
The addition of the ROCK inhibitor Y27632 (1 μM, Calbiochem) was supplemented into every medium change of SFTA3 control and KO progenitors beginning at differentiation day 21 until day of fixation.

Quantitative Real Time PCR
Quantitative measurements of mRNA expression were performed using the 7300 Real Time PCR System (Applied Biosystems) as previously described [12].

RNA sequencing and bioinformatics analyses
Characterization of the transcriptomes comparing NKX2.1-positive and NKX2.1-negative populations using shotgun mRNA sequencing (RNA-seq) was performed as previously described [12]. RNA-Seq data from the BrainSpan Atlas of the Developing Human Brain (http://brainspan.org) was assessed as previously described [12].

Data and statistical analyses
RT-PCR data was analyzed using a two-tailed student t-test comparing the average relative quantification values between the reference and control samples among 3 independent rounds of differentiation. For cell counts, 4-6 fields were imaged, total nuclei and the percent of cells expressing the markers of interest were counted. Following quantification, immunocytochemistry data was analyzed using a one-way ANOVA comparing protein expression values between knockout and control cell lines."

Comparative gene expression analysis of SFTA3 and NKX2.1 in MGE-like interneuron progenitors and fetal brain tissue
We used RNA-sequencing analysis to compare the transcriptome of NKX2.1-positive and NKX2.1-negative neural progenitors to identify genes whose expression was enriched in the NKX2.1-positive MGE-like population. SFTA3 expression was consistently enriched in the NKX2.1-positive population, and to the same extent observed for NKX2.1 (>8 fold gene enrichment) (Chen et al., 2016; Fig 1A). The expression values for SFTA3 are comparable to NKX2.1, with a correlative-coefficient R value >0.98; indicating strong statistical significance in fold change enrichment between these two genes (S1 Table).
We used RT-PCR analysis to examine expression levels of these two genes at specific time points during the differentiation of hESC-derived interneurons. Both genes were upregulated in a similar time course, consistent with shared regulatory elements (Fig 1B). Recently, an enhancer sequence region that may confer ventral forebrain expression was identified on . This shared enhancer region may be responsible for upregulating the transcription of both SFTA3 and NKX2.1. Additionally, immunocytochemistry revealed a large percentage of neural progenitors at day 26 of differentiation co-expressed NKX2.1 and SP-H (Fig 1C).
To investigate whether SFTA3 is expressed during the development of the human fetal brain, we examined transcriptome data from the BrainSpan Atlas. The Atlas indicates enrichment of SFTA3 in the MGE at 8 post-conception weeks (pcw) (Fig 1D). Furthermore, when comparing the expression levels of SFTA3 to all brain structures at 8 pcw, the MGE is the only structure to demonstrate a significant elevation in expression of the gene (Fig 1D). These results suggest that SFTA3 is a novel biomarker specific to the MGE during early ventral forebrain development.

Day 25 characterization of SFTA3 and NKX2.1 knock-out neural progenitors
Given the highly correlative spatiotemporal gene expression patterns of NKX2.1 and SFTA3 during in vitro differentiation of FACS-enriched NKX2.1-postive neural progenitors and in the human MGE at 8 pcw (Fig 1), we determined whether SFTA3 plays a role in specifying MGE-like neural progenitors, as observed for NKX2.1. To investigate SFTA3 gene function, NKX2.1 and SFTA3 control and knockout hESCs were generated using CRISPR-Cas9 mediated genome editing. To verify successful KO of each gene, we performed qRT-PCR at day 25 of differentiation, when both genes are expressed at significant levels. Day 25 qRT-PCR analysis of NKX2.1 control and KO progenitors indicated no detectable RNA transcripts of NKX2.1 in the KO cell line (S1 Fig). In addition, the deletion of NKX2.1 significantly reduced SFTA3 transcript levels, suggesting NKX2.1 is needed to maintain control levels of SFTA3 gene expression (S1 Fig). Both SFTA3 KO cell lines had no detectable RNA transcripts for SFTA3. Interestingly, NKX2.1 RNA expression is not significantly affected in SFTA3 KO lines (S1 Fig;  Fig 2B). We also examined levels of protein expression for these two genes. Immunocytochemistry at differentiation day 30, revealed the absence of NKX2.1 and SP-H protein in their respective KO cell lines (S1 Fig). Whereas qRT-PCR showed little change in NKX2.1 mRNA expression in SFTA3 KOs, these immunocytochemistry data suggest that both SFTA3 KO lines had somewhat lower levels of NKX2.1 protein (Fig 2C and 2D).
To investigate SFTA3 gene function in the specification of MGE cell fate, NKX2.1 and SFTA3 control and knockout hESCs were differentiated to MGE-like neural progenitors in vitro for 25 days and analyzed for expression of several neural markers using qRT-PCR. At this time point, region-specific transcription factors are expressed at sufficient levels to determine lineage specificity of the differentiating cells. Again, no NKX2.1 was detected in the KO cell line, verifying gene deletion (Fig 2A). Markers of the MGE (SFTA3, DLX1, NKX6.2, LHX6), telencephalon (FOXG1), ventral telencephalon (ASCL1), striatal medium spiny neurons (FOXP1), mediobasal hypothalamic neurons (RAX) and GABAergic cells (GAD2) displayed significantly decreased levels of expression in the NKX2.1 KO relative to the control cell line (Fig 2A). Expression of choline acetyltransferase (ChAT), an enzyme that synthesizes acetylcholine in cholinergic neurons, also exhibited a significant decrease in the NKX2.1 KO compared to the control. In addition, expression of markers of dorsal telencephalon and CGE lineage (Pax6, NR2F2, SP8) were enriched in the NKX2.1 KO relative to controls (Fig 2A). There was no significant difference in expression levels of FOXP1, a marker of hypothalamic neurons, in the NKX2.1 KO relative to its control (Fig 2A and 2B). These data are consistent with in vivo studies that demonstrate a re-specification of MGE to non-MGE cell neural fates in conditional NKX2.1 KO mutant mice [3,9,27,28] No detectable SFTA3 mRNA transcripts were observed in the two SFTA3 KO cell lines, verifying gene mutation (Fig 2B). Though the SFTA3 KOs showed a general trend towards decreased gene expression of markers of the ventral telencephalon and MGE (Ascl1, DLX1, NKX6.2, GAD2, and LHX6), the downward trend was not as severe as observed for the NKX2.1 KOs. Intriguingly, contrary to what was observed in the NKX2.1 KO cell population, there was no compensatory increase of dorsal forebrain and CGE neural lineage markers (PAX6, NR2F2, SP8) in the SFTA3 KO cells (Fig 2B). These data suggest a discrepancy in phenotype between NKX2.1 and SFTA3 KO RNA expression levels, suggesting SFTA3 is not solely responsible for conferring MGE fate downstream of NKX2.1.
We next examined the expression of neural stem cell (NSC), immature neuronal, and ventral progenitor markers in both our control and KO generated cell lines. Greater than 95% of the cells at day 25 in all cell lines expressed the NSC marker nestin, indicating that KO of NKX2.1 or SFTA3 did not interfere with neural differentiation (Fig 2D). In addition, comparable levels of the M-phase marker phosphohistone H3 were observed in all cell lines at around this stage of differentiation, suggesting mutation did not alter proliferation levels (data not shown). While 7.57%, 8.17%, and 11.74% of the NKX2.1 control cells expressed DLX2, Olig2, and FOXG1 respectively, only 1.40%, 1.51%, and 5.04% of cells expressed these markers in the NKX2.1 KO cell population, indicating a significant reduction in interneuron progenitor differentiation (Fig 2C and 2D). SFTA3 KO progenitors did not display the same decline in comparison to the control cells. Intriguingly, there was a marginal but significant decrease in cells expressing NKX2.1, OLIG2, and FOXG1 protein expression for both SFTA3 KO lines in comparison to the control (Fig 2C and 2D). Consistent with the high percentage of cells expressing immature NSC markers at day 25, there was no significant expression of the inhibitory neurotransmitter GABA at this time point (<2%; Fig 2D). Surprisingly, the percentage of cells expressing DCX, a post-mitotic neuronal migration marker and MAP2, a mature neuronal marker, was somewhat higher in the SFTA3 KO lines (8-11%; 4-5%, respectively) compared to its control (4.4%; 2.6%, respectively; Fig 2D). This suggests SFTA3 may play a role in preventing premature differentiation of neural progenitors and keeping them at the NSC stage. As observed with RNA data, declines in levels of cells expressing MGE and forebrain protein markers were not as dramatic for the SFTA3 KOs as the NKX2.1 KOs relative to controls. Overall, the mutation of SFTA3 alone was not sufficient to diminish expression of MGE and subpallidal markers to NKX2.1 KO levels, as SFTA3 KOs showed only a moderate reduction in NKX2.1 expression. Taken together, these results suggest that SFTA3 does not have a central role in specifying MGE-like progenitor cell lineage as established for NKX2.1.

Day 45 characterization of SFTA3 and NKX2.1 knock-out neural cell populations
To evaluate whether the absence of SFTA3 affects the differentiation of hESC-derived GABAergic progenitors into inhibitory interneurons, neural progenitor cells were cultured for an additional 20 days and analyzed at day 45 in vitro. As hESC-derived interneurons demonstrate a protracted developmental timeline for maturation, maintaining long-term cultures for immunocytochemical analysis was difficult, with some cell lines demonstrating decreased viability at time of analysis. Due to decreased survival at this time point, analyses were limited to only a few key markers, including the interneuron marker GABA. At this time point, levels of GABA are sufficient to quantify interneuron lineage specificity, but we were unable to determine whether SST or PV interneuron subtypes were present. The loss of NKX2.1 or SFTA3 expression did not significantly affect maturation of the neural progenitors, as there was comparable expression of both DCX and MAP2 in all KO lines compared to controls (Fig 3A and  3B). Day 45 neurons generated from the NKX2.1 KO line exhibited minimal GABA expression, (< 1%) relative to control cells at this time point (> 38%), consistent with a role for NKX2.1 in conferring GABAergic inhibitory identity. Immunocytochemistry-based quantification of SFTA3 mutants indicated~10% GABA expression in both SFTA3 knockout lines compared to~18% expression in the control, demonstrating that the deletion of SFTA3 did not eliminate neural progenitor differentiation down the GABAergic lineage (Fig 3A and  3B).

Discussion
The transcription factor NKX2.1 is a master regulator of MGE-specific lineages, promoting the expression of genes involved in specifying cortical inhibitory interneuron progenitors and helping to maintain proper ratios of mature interneuron subtypes in the adult brain [3,29,30]. This report highlights the identification of a surfactant gene that is expressed in NKX2.1-positive cells and may support inhibitory interneuron differentiation, though not as robustly as NKX2.1. Based upon transcriptome and qRT-PCR analysis, we identify SFTA3 as a novel MGE marker (Fig 1) and examine SFTA3's role in interneuron differentiation by generating NKX2.1 and SFTA3 KO hESC lines and determining their ability to produce MGE-subtypes and derivatives. We show that MGE-lineage specific markers and interneuron subtypes are only marginally decreased when SFTA3 is mutated, whereas the NKX2.1 KO showed significant depletion in both GABAergic neural progenitor and cortical inhibitory interneurons. By day 25 of differentiation, NKX2.1 KO progenitors showed significant levels of enrichment in expression of dorsal and CGE-associated genes at the expense of MGE marker expression ( Fig  2A). Surprisingly, at day 45, despite our observation that non-MGE like interneuron progenitors were present at day 25, no cells expressing GABA were detected in the NKX2.1 KO. This suggests the non-MGE alternative fated cells specified at day 25 could not mature under our in vitro conditions. (Fig 3A and 3B; Butt et al., 2008;Sussel et al., 1999).
Analysis of RNA expression levels for ASCL1, a proneural gene associated with promoting cell cycle exit and neuronal differentiation was reduced in both SFTA3 KO lines in comparison to the control. Additionally, immunocytochemistry data interestingly showed a significant increase in the percentage of cells that expressed DCX and MAP2 in the SFTA3 KOs relative to the control at day 25 of differentiation. These results highlight a possible function of SFTA3 as a cell-cycle regulator of MGE-like progenitors during neurogenesis.
The orthologue of the human SFTA3 gene in rodents is NANCI, which appears to encode a long noncoding RNA (lncRNA). A comparison of coding sequences using nucleotide BLAST algorithms was performed in order to understand why human SFTA3 is a protein encoding gene and mouse NANCI a lncRNA (data not shown). The NANCI coding sequence lacks long open reading frames due to the presence of numerous translational stop codons interspersed throughout the sequence. In contrast, the human SFTA3 coding sequence contains long uninterrupted open reading frames that do not contain stop codons, consistent with protein coding regions. Recently, Herriges and colleagues observed nearly identical expression patterns during development for both NANCI and NKX2.1 in the mouse lung epithelium and forebrain [23]. Heterozygous NANCI mouse mutants had decreased levels of NKX2.1 expression but did not display significant morphological defects in the lungs. Furthermore, heterozygous NKX2.1 mutants showed a compensatory increase in NANCI expression, resulting in an upregulation of NKX2.1 expression back to wild-type levels [24]. These data suggest that NANCI may function as a regulator of NKX2.1 expression. We observed similar tissue specific NKX2.1 and SFTA3 gene expression patterns in in vitro derived MGE-like progenitor cell population and in transcriptome data of human 8 pcw MGE fetal tissue. Despite similarities in the tissue specificity of expression of SFTA3 and NANCI, our data do not support a role for SFTA3 in regulating NKX2.1 mRNA expression. However, our observation that levels of NKX2.1 protein are somewhat decreased in the SFTA3 mutant lines, suggests the possibility that SFTA may act at a posttranscriptional level to modulate NKX2.1 protein levels. Quantitative Western blot analysis would be an essential next step in determining if this is the case. Further investigation to verify the presence and potential function of an SFTA3 long noncoding RNA in human cells is needed.
In summary, our data demonstrate that the expression of SFTA3 and NKX2.1 are coordinated during differentiation of hESCs in vitro to an MGE-like fate. Deletion of SFTA3 only marginally affects the expression of NKX2.1, and does not lead to any major changes in the expression profiles of MGE markers and interneuron derivatives. Further research is needed to fully assess the functional differences between SFTA3/NANCI and NKX2.1 in rodents and humans.