Notch-mediated expansion of cord blood progenitors: maintenance of transcriptional and epigenetic fidelity

Recent advances have been made towards successful generation of human hematopoietic stem and progenitor cells (HSPC) derived from embryonic, induced-pluripotent, or hematopoietic stem cells. However, whether these stem cell populations, which are generated after ex vivo manipulation, are suitable for clinical application depends on the faithful preservation of their transcriptional, epigenetic, and functional properties with respect to their primary cell counterparts.

To overcome the significant delay in neutrophil recovery following cord blood (CB) transplantation (CBT), we developed methods for the ex vivo expansion of CD34⁺ CB derived HSPC by culture with the Notch-ligand Delta1. When infused into patients undergoing a myeloablative CBT, these cells have been shown to provide more rapid short-term myeloid reconstitution when co-infused with a non-manipulated cord blood unit(s). However, infusion of ex vivo generated HSPC for long-term hematopoietic reconstitution remains a goal in the context of stem cell transplantation and gene therapy, thus the development of methods to assess the safety of these cultured products are essential.

Our group and others have demonstrated successful expansion of CB HSPC for both pre-clinical and clinical application. Standardized methods for assessing transcriptional quality control of pluripotent cell lines have been developed; however there has been no standard approach to assessing long-term safety of cultured progenitor cells infused in humans.¹ Here we investigated the transcriptional and epigenetic states of CB HSPC generated after ex vivo expansion on Delta1. We found that the transcriptomes and methylomes of the expanded HSPC overall recapitulate those of their primary cell counterparts.

CD34⁺ CB HSPC were cultured under conditions identical to those in our ongoing clinical trials.² We validated in vitro and in vivo growth characteristics of cells used for transcriptional and methylation studies to ensure their similarity with our previously published results. Mean CD34⁺ fold expansion was 191 +/- 62-fold as compared to 184 +/- 35 previously. In vivo repopulation in sublethally irradiated (275 rad) NOD-SCID IL-2Rγ-null mice (NSG) approved for use by the Fred Hutchinson Cancer Research Center Institutional Animal Care and Use Committee was at least as robust as our previously published results.

For transcriptional analysis, CD34⁺ CB HSPC were isolated from six individual CB units within 24 hours of collection. Freshly isolated CD34⁺ CB HSPC were then frozen for subsequent RNA extraction while remaining cells were placed in culture for 14 days as previously described.² Upon cell harvest at day 14 of culture, CD34⁺ cells were selected for RNA extraction and analysis on the Illumina HT12v3 platform (Illumina Inc, San Diego, CA, USA). All data were analyzed using the Bioconductor framework.³ Lumi package⁴ was used for reading and normalizing the raw data output by GenomeStudio software (Illumina Inc.). Limma package³ was used for identifying differentially expressed genes between freshly isolated and cultured CD34⁺ CB HSPC. Significant genes were defined as those with False Discovery Rate (FDR) less than 1% and greater than 2-fold change between non-manipulated and ex vivo expanded cells.

Of 24 912 genes analyzed, 668 genes were up-regulated in the ex vivo expanded groups and 616 genes were down-regulated as compared to non-manipulated control HSPC from the same CB units. Using DAVID Bioinformatics Resources v6.7, a web-based functional annotation tool for data analysis of microarray data (http://david.abcc.ncifcrf.gov/home.jsp), we performed Gene Ontology (GO) analysis for biological processes (GOTERM_BP_FAT) of these differentially regulated genes.^5,6 Differentially expressed genes were used as the inputs and “GOTERM_BP_FAT” was used as the GO search term to identify terms related to biological processes. FDR < 1% was used to identify GO terms that were significantly different between fresh and cultured CD34⁺ cells. Most GO terms identified were those related to cell culture conditions. Only 4 significant GO terms were identified in the down-regulated gene set and these were associated with non-replicative cellular state including nucleosome assembly, chromatin assembly, and protein-DNA complex assembly. Significant GO terms with up-regulated genes were overlapping and related to cells in a replicative state as expected for cells in culture. These GO terms represented more general cell cycle phase (e.g. mitosis, M phase, regulation of cell cycle) or specific steps of cell division (e.g. organelle fission, spindle organization, chromosome segregation) (Figure 1A). Notably absent were GO terms associated with differentiation, lineage commitment or cellular proliferation.

Figure 1. Transcriptional analysis of expanded CB HSPC.

(A) Gene ontology (GO) analysis of differentially expressed genes between primary and expanded CB HSPC. Differentially expressed genes were defined as those with FDR < 1% and LogFC ≥1. GO analysis was performed using DAVID Bioinformatics Resources v6.7 (http://david.abcc.ncifcrf.gov/home.jsp), input term GOTERM_BP_FAT. Significantly enriched GO terms are shown. (B, C) Differentially expressed genes were compared to established tumor suppressor (B) and oncogene (C) lists. Overlapping genes and enrichment are shown.

Given concern for acquisition of genomic alterations in cultured cells and the risk these pose for dysregulation of cell growth, we compared up- and down-regulated genes to known gene groups for tumor suppressor genes (TSG) and proto-oncogenes. We compared our list of down-regulated genes to those on TSGene, a tumor suppressor gene database, (http://bioinfo.mc.vanderbilt.edu/TSGene/browser.cgi),⁷ as loss of function of these genes would suggest risk of unregulated cell proliferation. Of the 718 human tumor suppressor genes in the database, 38 overlapped with down-regulated genes in cultured cells (Figure 1B). Looking more specifically at tumor suppressor genes described in leukemia and acute myeloid leukemia (40), only BEX2 overlapped with our down-regulated gene list. While better described as a TSG in malignant gliomas and breast cancer, hypomethylation and increased expression of BEX2 have been described in leukemia with MLL rearrangements. ^8,9 Down-regulation of BEX2 in the cultured cells is of unclear significance. However, our subsequent methylation studies showed no difference in BEX2 methylation between primary and cultured cells. To determine if BEX2 down-regulation was the result of altered expression of transcription factors known to regulate BEX2, we looked at expression levels of SOX2, β-catenin, p65 and ERBB2 and found no difference between primary and cultured cells.

We compared genes up-regulated in the cultured cells to an established list of proto-oncogenes (www.uniprot.org/uniprot, search terms “proto-oncogene”, “human”) as up-regulation of these genes might be expected to result in unintended cell proliferation. Among the list of 232 proto-oncogenes, 11 were up-regulated in cultured cells (Figure 1C). While many of these genes have been implicated in hematologic malignancies, they also play important roles in normal hematopoiesis and thus may be up-regulated in this setting. Because the CB HSPC were cultured in the presence of Delta1, we also determined whether there was persistent altered expression of Notch downstream target genes as some of these have known roles in cell cycle entry and proliferation. We found no difference in expression of known Notch targets SKP2, HES1, HEY1, CCND1, or BCL2. As shown in Figure 1C, MYC expression was upregulated following culture on Delta1 ligand, however, Notch is one of many regulators of MYC expression.

Whereas transcriptional activity represents cellular function, we expected that DNA methylation states would offer a more detailed picture of the fidelity of the expanded populations with respect to cell identity and thus safety for clinical use. We assessed the epigenetic landscape of the cultured cells through genome-wide DNA methylation studies. CD34⁺ selected CB HSPC were cultured for 14 days and again CD34-selected for comparison to primary cells. DNA methylation analysis was performed by high-throughput reduced representation bisulfite sequencing (RRBS) profiling of CD34-selected primary and cultured cells.¹⁰ The original dataset contained 24 981 genes. M-value statistics were used to measure methylation levels (M-value=log₂[(methy +1)/(unmethy+1)]). Genes with M-values of 0 across all samples were filtered out leaving 17 346 genes for analysis. Analysis was performed on primary and cultured cell populations from 6 individual cord blood units using Limma package³ with an FDR threshold of 1%, as previously described. Overall, DNA methylation patterns were similar between primary and cultured cells (Figure 2A) with only 21 genes identified with increased methylation following culture, 14 of which have known function. This is in contrast to many cultured cell lines which show hypermethylation including at tumor suppressor genes.^11-13 None of these genes overlapped with tumor suppressor or proto-oncogenes evaluated in our transcriptional analysis. GO terms associated with these genes include cytoskeleton, translation, proteolysis, and cell activation (Figure 2B).

Figure 2. Methylation analysis of expanded CB HSPC.

(A) Violin plot depicting methylation density (M-value) of fresh (red) and cultured (blue) CB HSPC. (B) Gene ontology (GO) analysis of hypermethylated genes in cultured CB HSPC as compared to fresh cells. Differentially methylated genes were defined as those with adjusted p-value (FDR) <0.01. GO analysis was performed using DAVID Bioinformatics Resources v6.7 (http://david.abcc.ncifcrf.gov/home.jsp), input term GOTERM_BP_FAT.

To better determine whether small numbers of cells were generated that had functionally significant alterations able to lead to clonal dominance, we transplanted freshly isolated or cultured CB HSPC into NSG mice to determine if in vivo growth selected for and subsequently allowed detection of clones of cultured cells with altered DNA methylation. CD34⁺ CB HSPC were sorted from mice following 3-22 weeks of in vivo growth and global DNA methylation analysis was performed. Using the Limma package, we compared the methylomes of transplanted primary CD34⁺ cells and transplanted cultured CD34⁺ cells and their progeny and found no differential methylation at an FDR level of 1%. This included the 21 genes identified when comparing fresh and cultured cells further suggesting that this differential methylation in vitro did not provide in vivo growth advantage. We additionally compared the methylomes of transplanted cultured CD34⁺ cells to primary CD34⁺ cells to determine if any differentially methylated genes overlapped with those identified when fresh and cultured cells were compared. Two genes were identified with differential methylation (DRG2 and SERPINA1D) but these did not overlap with those shown in Figure 2A, providing further support that Notch signaling does not induce aberrant epigenetic changes. Overall, these data demonstrate that culture of CB HSPC on Delta-ligand does not result in significant alterations in DNA methylation of the cultured cells. Finally, we compared methylation levels of genes shown to be up or down-regulated in our transcriptional analysis to see if altered methylation could contribute to differential gene expression and found no difference in methylation at these genes. Looking specifically at methylation status of the differentially expressed TSG and proto-oncogenes from our transcriptional analysis, we again found no significant differences in methylation.

Taken together, our studies indicate that, in contrast to transcriptional differences between primary and cultured cells including at some TSG and proto-oncogenes reflecting differences in cellular proliferation, DNA methylation showed strong fidelity between the primary and cultured cells. Methods for assessing the safety of cultured progenitor cell products for clinical intent will become increasingly important as these products become more widely utilized and intended for sustained, long-term engraftment. This will be particularly true for ex vivo derived cells, either from induced pluripotent stem cells (iPSC) or embryonic stem cells (ESC), or by direct conversion to hematopoietic stem cells. These data provide a baseline for future critical assessment of expanded populations of HSPC including iPSC and ESC derived HSPC.