Validation of baboon pluripotent cells as a model for translational stem cell research

Abstract

Translation of stem cell therapies to the clinic will be most successful following optimization of efficacy and safety in appropriate preclinical model systems. Among available models, nonhuman primates (NHPs) provide the most accurate recapitulation of human anatomy, physiology, genetics and epigenetics. Here, we show that baboon pluripotent cells (PSCs) recapitulate key molecular features of human PSCs with greater accuracy than that found in PSCs from non-primate species such as mice. Specifically, baboon and human PSCs exhibit greater conservation of gene expression patterns, higher sequence and structural homology among pluripotency factors, more equivalent genome-wide patterns of histone and DNA methylation modifications, and similar maintenance of bivalent programming of developmental genes than that found between human and non-primate PSCs.

1. Introduction

Stem-cell based therapies have the potential to dramatically improve the treatment and prognosis of patients suffering from injury, aging, or various degenerative diseases or debilitations by restoring or replacing cell or tissue function (Guilak et al., 2019). Yet despite their promise, clinical implementation of stem-cell based therapies has been impeded by concerns over lack of efficacy and/or potential for off-target effects associated with transplantation of cells derived from pluripotent stem cells (PSCs) (Trounson and McDonald, 2015). Typically, clinical trials of stem-cell based therapies are initiated following rigorous testing and evaluation in animal models to preclinically optimize methodology and safety (Tompkins et al., 2018). However, the most commonly used animal models – rodents, rabbits, goats, pigs and others – while very useful for providing initial insight, all fail to optimally represent humans in size, anatomy, tissue organization, immune function, genomics or epigenomics (Grow et al., 2016a).

With the appropriate exclusion of great apes as models for most animal testing, baboons and macaques present the most accurate recapitulation of human biology among species readily available for biomedical research. Both species possess genomes that show 92–93% sequence homology with the human genome (Rogers and Hixson, 1997). However, baboons are more bipedal and more similar to humans in size and anatomy than macaques, recapitulate human tissue complexity well, offer a particularly excellent model of human neuroanatomy and neurophysiology, and possess an immune system that is more similar to that of the human than are those of other commonly used models, including other nonhuman primate (NHP) species (Grow et al, 2016a). In addition, many disease models already exist in baboons for which cell-based approaches may offer particularly effective treatment options, such as cardiovascular disease, chronic lung disease, and liver disease (VandeBerg et al., 2009), and others can be phenocopied by drug-based induction, such as Parkinson’s disease (Grow et al., 2016a). These factors suggest that baboons offer unique value as a model system in which to optimize the efficacy and safety of cell-based therapies prior to clinical adoption of these protocols.

Here we provide data showing that at the levels of gene expression and epigenetic programming, baboon PSCs provide the most accurate recapitulation of human PSCs, thereby further validating the value and utility of the baboon system for preclinical studies of cell-based therapies.

2. Materials and methods

2.1. Baboon ES/iPS cell culture

Baboon (Papio anubis) ES (B4, B15) and iPS (160-2, M14939#3) cells used in this study were previously derived (Navara et al., 2013) and grown on feeder cells (mouse embryonic fibroblasts [MEFs]) as described (Navara et al., 2018). To prevent MEFs from being included during passaging, baboon PSCs were manually selected and passaged by hand using flame polished sterile glass pipettes under 10x magnification inside a sterile biohazard hood.

2.2. Isolation of baboon pachytene spermatocytes and round spermatids

Baboon testes were obtained at necropsy from animals undergoing euthanasia for other purposes at the Southwest National Primate Research Center. Euthanasia procedures were approved by the Southwest National Primate Research Center/Texas Biomedical Research Institute IACUC. Pachytene spermatocytes and round spermatids were separated on a 2–4% BSA StaPut gradient as described (Lesch et al., 2016). Populations of pachytene spermatocytes and round spermatids showed ≥90% purity based on morphology visible under phase optics.

2.3. mRNA sequencing and protein structure prediction

Total RNA was harvested from baboon PSCs immersed in TRIzol using the Directzol RNA Miniprep kit (R2050, Zymo Research, Irvine, CA) and converted to cDNA with SuperScript™ III First-Strand Synthesis SuperMix (18080400, ThermoFisher) according to the manufacturer’s instructions. Primers were designed for cDNA sequences of P. anubis NANOG, POU5F1, KLF4, SOX2, C-MYC, and LIN28 predicted from the Panu_3.0 genome assembly (GCF_000264685.3). Amplified fragments were purified by agarose gel electrophoresis and extracted using the Qiaquick Gel Extraction Kit (28506, Qiagen, Hilden, Germany), or clonally selected, amplified and purified using the TOPO TA Cloning Kit (450071, ThermoFisher). Sequencing was performed by the University of Texas at Austin DNA Sequencing facility. Resulting fragments were aligned to predicted Panu_3.0 mRNA sequences using NCBI BLAST. AA sequences deduced from coding regions of confirmed P. anubis cDNA sequences were used to predict protein structures via homology modelling with SWISS-MODEL (https://swissmodel.expasy.org).

2.4. Chromatin immunoprecipitation–quantitative PCR (ChIP-qPCR)

2.4.1. Sample preparation

ChIP assays were performed using the ChIP-it-express kit (53008, Active-Motif, Carlsbad, CA) with modifications as described (Brind’Amour et al., 2015). Pre-cleared chromatin was incubated with rotation overnight with 2 μg of H3K4me3 (ab8580, Abcam, Cambridge, MA), H3K27me3 (ab6002, Abcam), or rabbit IgG (ab171870, Abcam) antibody, or no antibody at 4 °C. DNA from precipitated chromatin was recovered and purified as described (Cheng et al., 2020).

2.4.2. qPCR conditions

Real-time PCR was performed using a Power SYBR green PCR mix (Applied Biosystems) in an Applied Biosystems 7900HT 384-well Real-time thermocycler system, and fold enrichment was determined by the 2-ΔCt method based on IgG Ct value as described (Rao et al., 2013).

2.5. qPCR validation with Fluidigm IFCs

Results of ChIP assays described above were validated by assaying immunoprecipitated DNA that was diluted and pre-amplified for 15 cycles, then analyzed by Real-time PCR using a Biomark 12X12 IFC chip in a Fluidigm Biomark HD system following the manufacturer’s instructions. The results were computationally assessed using Biomark Data Collection and Real-time PCR analysis software.

2.6. RNA-seq, ChIP-seq, and MeDIP-seq

2.6.1. Sample preparation from live cells

Baboon ES and iPS cell lines were expanded in culture as described above, and harvested at a single passage. 1000 live cells per line were used for generation of cDNA for RNA-Seq libraries, while 100,000 cells each were used for preparation of MeDIP-Seq and ChIP-Seq libraries.

2.6.2. Library preparation

RNA-Seq libraries were prepared from live cells with the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (634888, Clontech Laboratories, Moutain View, CA) and the Nextera XT DNA Library Prep Kit (FC-131–1024, Illumina, San Diego, CA) according to the manufacturers’ specifications. ChIP-Seq libraries were prepared from each cell line with antibodies for H3K4me3 (abcam8580), H3K9me3 (abcam8898), H3K27me3 (abcam6002), and H3K27ac (39133, Active Motif, Carlsbad, CA) as described (Brind’Amour et al., 2013). Each library was prepared from 50,000 cells with inclusion of PEG in samples during MNase digestion, adapter ligation with NEB adapters (E7535S, New England Biolabs, Ipswich, MA) and adapter cleavage by incubation with USER enzyme (M5505S, New England Biolabs) at 37 °C for 15 min. MeDIP-Seq libraries were prepared as described (Taiwo et al., 2012).

2.7. Quality control and library sequencing

All libraries were validated for targeted fragment size and distribution using an Agilent 2100 Bioanalyzer. Library concentrations were determined by qPCR prior to pooling using the NEBNext Library Quant Kit for Illumina (NEB #E7630). Pooled samples were sequenced on an Illumina HiSeq 3000 system at UT Health San Antonio.

2.8. Computational analyses

Human and mouse RNA-, ChIP-, and MeDIP-Seq archived data originating from different ES and iPS cell lines from multiple laboratories and sourced from GEO (Table S2) was used for comparison to baboon data generated in this study. All sequenced baboon ESC or iPSC libraries were aligned to the Panu_3.0 genome (GCF_000264685.3) with bowtie 2.3.2. Archived human and mouse datasets were aligned to the Homo sapiens GRCh38.p12 (GCF_000001405.38) or Mus musculus GRCm38.p6 (GCF_000001635.26) genomes with bowtie 2.3.2, respectively. RNA-Seq read counting and gene identification was performed in R with the GenomicAlignments, GenomicRanges, and GenomicFeatures packages and the corresponding species-specific genomic annotations. Homologous genes in human, baboon and mouse were identified with the biomaRt package and differential expression was calculated with edgR. Differentially expressed genes were identified by ranking log fold change, p value, and FDR. Associated GO terms were analyzed with clusterProfiler. ChIP-Seq and MeDIP-Seq alignment files were analyzed with MACS2 to identify peak regions of histone modifications or DNA methylation. ChIP-Seq coverage maps and binding profiles were generated with the deepTools package. Peak annotation and feature graphs were generated in R with the Chipseeker package. Genome coordinates and genome annotation files of baboon and mouse were liftovered to human using the CrossMap. Human, baboon and mouse genomes were visualized using the USCS genomic browser.

3. Data and code availability

Human and mouse data were drawn from previously published data sets (Table S2). Baboon RNA-seq, ChIP-seq and MeDIP-seq data have been deposited in the Gene Expression Omnibus (accession ID: GSE165191). Only genes found in all three species were included in this study. Any species-specific genes were excluded. Codes were deposited on github https://github.com/chengkeren/baboon. A UCSC genome browser session was deposited http://genome.ucsc.edu/s/kerencheng/papAnu4.

4. Results

4.1. Baboon pluripotency factors are closely related to human pluripotency factors

We first confirmed the sequences of transcripts encoding five baboon transcription factors and one RNA-binding protein commonly expressed in pluripotent cells, SOX2, POU5F1, NANOG, MYC, KLF4, and LIN28A, and then compared those and the encoded amino acid sequences with those associated with the homologous factors in macaques, mice and humans (Fig. 1A). Nucleic acid (NA) identity was calculated for all transcripts using the ORF to reduce confounding effects introduced by incompatibility within intronic or non-translated regions of the transcript. The baboon (Papio anubis) and rhesus monkey (Macaca mulatta) factors both showed high sequence similarity to the human pluripotency factors at both the NA (95–99%) and amino acid (AA) (94–99%) levels, whereas the equivalent similarities between mouse and human factors were lower at 75–94% at the NA and 58–97% at the AA levels. For all three model species, NANOG showed greater sequence divergence from its human counterpart than did the other five factors. NANOG also showed the unusual characteristic that among the different species compared, the AA identity detected for the NANOG protein is lower than the NA identity for the NANOG gene (Fig. 1A). We attribute this to positive selection having promoted the divergence in the NANOG AA sequence among these species such that a minimum number of changes in the NA gene sequence (e.g. as few as one base/codon = a 33% change) appears to have resulted in a maximum difference in the encoded AA protein sequence (e.g. an amino acid replacement = a 100% change).

Fig. 1.

Comparison of key pluripotency-related transcription factors in human, baboon and mouse PSCs. (A) Nucleic acid (NA) and amino acid (AA) sequence similarity among H. sapiens, P. anubis, M. musculus and M. mulatta homologues of key pluripotency-related transcription factors. (B) A comparison of 3D structures of regions of pluripotency factor homologues confirmed by crystal structure resolution in humans and predicted from AA sequences in baboons and mice. Arrows indicate predicted species-specific structural differences in baboon or mouse homologues relative to the corresponding human homologues. Numbers correspond to AA sequence differences (shown in C) that contribute to these predicted structural differences. (C-F) Species-specific variation in AA sequences in known functional domains or regions that contribute to predicted structural differences between human and either baboon or mouse homologues of pluripotency factors. The extent of AA sequence consensus among all species is indicated by bar height above each amino acid residue. Potential functionally significant mismatches are indicated by red numbers. Red lines beneath domain maps indicate protein regions with available structural models. (G-H) Pluripotency factors that do not exhibit species-specific AA variations in known functional domains or regions for which predicted structures are available. Asterisks indicate areas where AA changes do occur. ZF = Zinc Finger; LZ = Leucine zipper; bHLH = basic helix-loop-helix.

Sequence-based homology modeling of protein structures revealed that in the mouse, two key pluripotency factors, OCT4 and NANOG, showed multiple predicted structural differences from their human equivalents, whereas we found only one potential structural difference between baboon and human pluipotency factors which was specific to LIN28A (Fig. 1B). We then compared AA sequences specific to known functional DNA-binding and/or effector domains within each factor from each species and found 12 AA sequence differences between mouse and human OCT4, but only a single AA difference between baboon and human OCT4 in the DNA-binding and homeobox domains (Fig. 1C). Relative to human NANOG, baboon and mouse NANOG showed nine and 23 AA substitutions in their DNA-binding domains, respectively, and another nine and 73 differences in their transactivating domains (Fig. 1D). Insertion of a serine at position 78 and functionally distinct AAs at positions 75, 77, 79–81, and 85–87 likely contribute to the distinct predicted structure of mouse NANOG shown in Fig. 1B.

Many fewer species-specific AA sequence differences were found within functional domains of the other four pluripotency-related factors (LIN28A, C-MYC, SOX2, KLF4) (Fig. 1E-1H). Among baboon factors, only LIN28A is predicted to show any structural variation from the corresponding human factor, specifically in the linker region (Fig. 1B and 1E), though whether this change is functionally significant remains unknown. The AA sequence of mouse C-MYC differs from human C-MYC at multiple AA residues in the leucine zipper region, but neither the baboon nor mouse C-MYC protein structures are predicted to be distinct from human C-MYC (Fig. 1B and 1F). Finally, none of the AA sequence differences between human, baboon and mouse KLF4 or SOX2 occur in functional domains or in regions for which a homology-based structure can be predicted (Fig. 1B, 1G, and 1H).

4.2. Gene expression in baboon pluripotent cells mimics that in human pluripotent cells

We next evaluated transcriptomes of human PSCs in comparison to those of baboon and mouse PSCs. A comparison of the annotated genome databases for each species revealed 19,200 genes for which homologues could be identified in all three species (Table S1) which formed a basis for comparison of PSC transcriptomes. Within each species, transcriptomes in ESCs and iPSCs were strikingly similar. However, principle component analyses of the top 2000 variably expressed genes revealed that PSCs from each species clustered distinctly (Fig. 2A). Nevertheless, hierarchical clustering of this data showed that patterns of gene expression in baboon PSCs were more similar to those in human PSCs than were those in mouse PSCs (Fig. 2B). A GO enrichment analysis of differentially expressed genes (DEGs) in PSCs from each species is shown in Fig. 2C, including GO terms for genes most up- (top row) or down- (bottom row) regulated in baboon iPSCs versus human iPSCs (first column), mouse iPSCs versus human iPSCs (second column), baboon ESCs versus human ESCs (third column) and mouse ESCs versus human ESCs (fourth column). Among the GO terms associated with these DEGs were those related to stem cell function and/or pluripotency, including regulation of stem cell population maintenance, stem cell differentiation, stem cell population maintenance, positive regulation of stem cell population maintenance, and stem cell proliferation.

Fig. 2.

Gene expression patterns in human, baboon and mouse PSCs. (A) A principal component analysis of PSC transcriptomes from each species (human, baboon, mouse) shows species-specific clustering of PSCs in general, plus a high degree of similarity between iPSCs and ESCs within each species. (B) Hierarchical clustering of expression patterns of genes related to stemness. The color bar indicates the log2 normalized FPKM (Fragments Per Kilobase of transcript per Million mapped reads) (See Table S3).(C) GO analysis of DEGs among baboon, mouse and human PSCs. Bar length indicates the significance, color indicates fold of enrichment in GO term (See Table S4).

4.3. Epigenetic programming patterns are similar in baboon and human pluripotent cells

Cell fate is driven by cell-type specific differential gene expresion which, in turn, is regulated by differential epigenetic programming (Roadmap Epigenomics Consortium et al., 2015). Directed differentiation of PSCs in vitro involves controlled transitions in epigenetic programming from that associated with the pluripotent state to that associated with each particular differentiated state (Meissner, 2010). To investigate the extent of similarity in epigenetic programming among human, baboon and mouse PSCs, we examined genome-wide patterns of four chromatin modifications (H3K4me3, H3K9me3, H3K27me3, H3K27ac) as well as DNA methylation in baboon PSCs. A majority of peaks of each histone modification or enriched DNA methylation region were found in intronic or intergenic regions or promoters (Fig. 3A and 3B). In general, when compared to data from human and mouse PSCs (Creyghton et al., 2010; Nicetto and Zaret, 2019), genome-wide patterns of two histone modifications – H3K9me3 and H3K27me3 –in baboon PSCs more closely recapitulate those in human PSCs than do those in mouse PSCs, while patterns of a third histone modification – H3K4me3 – and DNA methylation appear similar among all three species (Fig. 3C).

Fig. 3.

Epigenomic profiling of human, baboon and mouse PSCs. (A) Narrow peaks of H3K4me3, H3K27ac, and DNA methylation along with broader peaks of H3K27me3 and H3K9me3 were detected throughout the genome by ChIP-seq in baboon PSCs. (B) The genomic distribution of each type of peak shows a majority occurring in distal intergenic or intronic regions, with high concordance between patterns detected in baboon ESCs and iPSCs. (C) Similarities are shown among patterns of each histone modification in human, baboon and mouse PSCs based on Spearman correlation analysis. Red = high correlation, Blue = low correlation. Correlation coefficients are labeled. (D) A bar graph summarizes the occurrence of poised genes in baboon PSCs that did (23 genes) or did not (3 genes and 1 gene) match the pattern of poised genes in human PSCs. 20 of the 23 genes poised in baboon PSCs were also poised in baboon pachytene spermatocytes (PS) and round spermatids (RS) (see Fig. S1).

In humans and mice, promoters of a specific set of “poised genes” involved in early development have been shown to be bivalently marked by both the active H3K4me3 and repressive H3K27me3 marks in pluripotent cells (Bernstein et al., 2006) and in the germ line (Lesch et al., 2016). We found that 23 of 27 genes reported to be poised in human and mouse PSCs were similarly poised in both baboon iPSCs and ESCs, with the remaining four genes being poised in either baboon iPSCs or ESCs. Twenty of these genes were also poised in baboon spermatocytes and spermatids (Fig. 3D and S1).

4.4. Integration of genic region epigenetic programming profiles

We next sought to assess integrated multiparametric epigenetic programming within genic regions in PSCs from all three species. We analyzed aliquots of the same samples of each baboon PSC type by 1) ChIP-seq to detect four histone modifications (H3K4me3, H3K27me3, H3K27ac, H3K9me3), MeDIP-seq to assess DNA methylation, and bulk RNA-seq to catalogue gene expression, and compared the results to similar transcriptomic and epigenomic profiling data from human and mouse PSCs (Table S2). Heatmaps of genic region epigenetic programming in PSCs from all three species reveal peaks of H3K4me3 and H3K27ac colocalized with a dearth of H3K27me3, H3K9me3 and DNA methylation at promoters of actively expressed genes, as well as a noticible enrichment of H3K27me3, relatively reduced enrichment of H3K4me3 and H3K27ac, and a less pronounced lack of DNA methylation was found at promoters of genes that are not expressed in PSCs of any of the three species (Fig. 4A-C). Finally, to validate comparable patterns of epigenetic programming within the context of an individual pluripotency gene (NANOG/Nanog) in PSCs of each species, we compiled genome browser tracings of peaks representing these same epigenomic parameters (Fig. 4D). These stacked tracings confirm epigenomic profiles indicated by the heatmap representations in Fig. 4A-4C, including peaks of H3K4me3 and H3K27ac colocalizing with an absence of H3K27me3, H3K9me3 and DNA methylation at the Nanog gene promoter in PSCs from each species.

Fig. 4.

Multiparametric analysis of epigenomic profiles in human, baboon and mouse PSCs. (A-C) Multiparametric epigenetic + transcriptomic profiles within genic regions (TSS-TES + 3 kb upstream and 3 kb downstream) of the annotated genes common to all three species are shown for human (A), baboon (B) and mouse (C) PSCs. Peaks of H3K4me3 and H3K27ac, plus hypomethylation of DNA can be seen at promoters of expressed genes, whereas either promoter peaks or complete genic region coverage by H3K27me3 is associated with non-expressed genes in PSCs from all three species. (D) Aligned browser tracks showing the distribution of these same epigenomic parameters within the NANOG gene in PSCs from all three species confirm enrichment of H3K4me3 and H3K27ac and the absence of H3K27me3, H3K9me3 or DNA methylation in the promoter region of this gene in PSCs from all three species.

5. Discussion

Translation of basic research from initial findings in model species less closely related to humans to refined protocols in model species more closely related to humans and then to optimized procedures in the clinic propels biomedical progress. Stem cells, especially PSCs, have the potential to facilitate a paradigm shift in the way many debilitating diseases or conditions are treated by implementing a cell-based restorative approach in lieu of, or in combination with, surgical or drug-based methodologies. In particular, iPSC technology has significantly enhanced the potential for implementation of precision, patient-specific cell based therapies (Wiegand and Banerjee, 2019). However, this approach is laden with the potential for suboptimal outcomes and/or the danger of deleterious off-target effects, mandating the need to maximize its efficacy and safety prior to clinical implementation.

To this end, final optimization in a model species that most accurately recapitulates human biology is preferable, and the baboon in particular provides maximum available similarity to humans in terms of anatomical, physiological, neurological, behavioral, immunological, genomic and epigenomic characteristics (Grow et al., 2016a). A critical aspect of cell-based therapeutic approaches will be the ability of transplanted cells to function properly in the organ, tissue or cell system into which they are transplanted. Cell-type fate and function are manifest on the basis of differential gene expression (Cameron and Davidson, 1991) which is regulated by cell-type specific transcription factor networks functioning within unique epigenetic landscapes (Roadmap Epigenomics Consortium et al., 2015). Our analyses of global patterns of gene expression and epigenetic programming in human, baboon and mouse PSCs demonstrate higher similarity between human and baboon PSCs than between human and non-primate PSCs.

Several specific aspects of modeling human pluripotent cells in preclinical studies designed to optimize the efficacy and safety of cell-based therapeutic approaches will potentially benefit from use of cells from a species as closely related to humans as possible. For instance, concerns have been raised regarding the potential for immunogenicity of transplanted cells that could affect long-term graft survival and/or the effect reprogramming may have on HLA expression profiles (Okita, 2011; Liu, 2013; Mizukami 2014, Xian 2015). This is an area in which the baboon model offers particular advantages based on the greater similarity of the organization of its immune response genes to that in humans (Grow et al., 2016a). Similarly, caution has been advised with respect to specifics of differentiation protocols that might be applied to derivation of specific cell types from human PSCs and the extent to which “epigenetic memory” may skew the precision with which specific cell types derived in vitro from PSCs may accurately recapitulate the corresponding endogenous cell types in vivo (refs). Once again, use of PSCs that show overall epigenetic programming patterns that are as similar to those found in human PSCs as possible should facilitate development of maximally useful preclinical protocols to then be translated to use for clinical purposes (refs). Finally, it is also worth noting that like human ESCs, NHP ESCs are more likely to be captured in the naïve state, whereas rodent ESCs are more likely to be captured in the primed state.

The results from this study complement those we have previously reported regarding the utility of baboon PSCs (Grow et al., 2016a; Navara et al., 2013; Simerly et al., 2009), which, taken together, establish the baboon as one of the most accurate model of human biology available for preclinical optimization of PSC-based therapies.