Human and murine erythropoiesis

Abstract

INTRODUCTION

This is a review of recent advances in our understanding of human and murine erythropoiesis gained from transcriptome-based analyses. Advances in genomic technologies, particularly next-generation ultrahigh-throughput sequencing, have transformed our understanding of gene expression and gene regulation, as well as our understanding of mechanisms of human disease [1]. Comprehensive information on the composition of cellular transcriptomes obtained by RNA sequencing (RNA-seq) has provided us with detailed, unbiased knowledge about transcript composition and abundance, including knowledge about novel transcripts, novel isoforms, alternative splicing, and allele-specific isoform expression [2–4].

Studies of fundamental mechanisms of erythropoiesis in both human and mouse have provided critical insights into inherited and acquired disorders of the erythrocyte. Analysis of human erythropoiesis has primarily utilized in-vitro systems, whereas murinemodels have provided insights from both in-vitro and in-vivo studies. Despite their widespread utilization, detailed comparisons of human and murine erythropoiesis have not been performed. Although the essential body plan and critical physiologic axes are well conserved, humans and mice have diverged significantly. At the genomic DNA level, about half of human genomic DNA can be aligned to murine genomic DNA, with only a small percentage (~5%) under purifying selection in mammals [5]. This report reviews recent insights into human and murine erythropoiesis obtained from transcriptome studies during terminal erythroid differentiation.

HUMAN AND MURINE ERYTHROPOIESIS

It has long been known that there are fundamental differences between human and murine erythrocytes. Human erythrocytes have larger mean corpuscular volume (90 vs. 52 fl), longer life span (120 vs. 42 days), higher oxygen affinity (P50 25 vs. 40mm Hg), and lower levels of 2,3-diphosphoglycerate. Known differences between human and mouse exist in membrane protein structure and function, glucose utilization, vitamin C metabolism, signaling pathways, regulation of ion content and cell size, and mechanisms of stress erythropoiesis [6–13]. The variations in globin gene composition and developmental regulation between human and mouse are well documented [13,14]. Differences between human and murine erythrocytes persist ex vivo, as murine erythrocytes exhibit a much shorter half-life during storage [15]. Limited information on transcript isoform composition generated by alternate splicing in erythroid cells, such as the differences in exon composition of the delta-aminolevulinate synthase-2 gene cDNAs between mouse and human, is available [16]. Until recently, detailed analyses of transcript type and abundance in human or murine cells during terminal erythroid differentiation have been lacking [17▪▪,18,19▪▪,20,21,22▪▪,23].

TRANSCRIPTOME ANALYSES ACROSS TERMINAL ERYTHROID DIFFERENTIATION

An et al. isolated pure populations of human and murine erythroblasts at distinct stages in erythroid development using FACS-based methods to purify morphologically and functionally discrete populations of cells, each representing specific stages of terminal erythroid differentiation [24–26]. RNA from these cells underwent RNA-seq analyses, to create differentiation stage-specific transcriptomes [17▪▪].

Bioinformatic analyses of these transcriptomes revealed that there was tight clustering of biologic replicates from the different stages of terminal erythroid differentiation. There were remarkable differences between the individual stages at the transcriptional levels in both man and mouse. There were both shared and dissimilar gene expression profiles defining each stage of terminal erythroid differentiation within transcriptional space. These temporal changes in gene expression across differentiation revealed that each stage possesses a unique transcriptome. Clustering and network analyses revealed that differing stage-specific patterns of expression observed across erythroid differentiation were transcriptionally enriched for genes of varying function.

These observations supported, at the transcriptional level, the long-held tenet that the daughter cells produced during erythroid differentiation are structurally and functionally different than the mother cell from which they are derived. There are numerous phenotypic differences across the developmental and differentiation stages of erythropoiesis, such as changes in cell size and shape, hemoglobin composition and content, membrane structure and function, metabolic programs, nuclear alterations, and, ultimately, enucleation. These have led to the working hypothesis that erythropoiesis is a unique process in which each cell division is simultaneously coupled with a stage of differentiation [24,26–28]. This is in stark contrast to most cell types, wherein each cell division generates two daughter cells almost identical to the mother cell. These observations were true in both human and murine terminal erythroid differentiation, with the major difference being murine cells undergoing one less cell division from the proerythroblast to the orthochromatic erythroblast stages.

There were several major differences between human and murine transcriptomes. The most striking observation from these comparisons was that in contrast to human, there was a near-global decrease in gene expression during murine terminal erythroid differentiation. This was true across many clusters of genes. For instance, there were a variety of patterns of expression of transcription factor genes in human cells, whereas in murine cells, transcription factor expression exhibited the global steady decrease in gene expression across murine terminal erythroid differentiation.

Differences in both individual genes, as suggested by prior studies, for example, ALAS2, GLUT1, (Fig. 1) and entire groups of clusters of genes were observed between human and mouse (Fig. 2). Differences in signaling pathways, for example, the mitogen-activated protein kinase pathway, and ubiquitination, for example, the E3 ubiquitin ligase pathway, showed major differences between species. These altered patterns mostly mirrored the global decrease in gene expression observed in murine terminal erythroid differentiation, with genes upregulated in human but downregulated in mouse. Biologic studies have identified similar differences. RHEX, a recently identified regulator of erythroid cell expansion and erythroblast development, is present in human, but not mouse, rat, and lower vertebrates [29▪].

TRANSCRIPTOMES, CONSERVATION, AND DIVERGENCE

Pishesha et al. [19▪▪] performed global comparative gene expression analyses of terminal erythroid differentiation using morphologically identical stage-matched populations of human and murine erythroid cells, from early to late erythroblasts. Although the induction and repression of major erythroid transcription factors were mostly conserved between human and mouse, at a global level, there was substantial divergence between species at comparable stages during erythroid differentiation. Exceptions were SCL/TAL1 and MYB genes encoding critical transcriptional regulatory proteins. In humans, SCL/TAL1 is upregulated only in the polychromatophilic and orthochromatic erythroblast stages, whereas mouse Scl/Tal1 is upregulated earlier in differentiation at the basophilic erythroblast stage. The authors hypothesized that these two differences could significantly alter patterns of gene expression regulated by these two factors and their co-regulatory factors. Computational analyses predicted regulation by similar key transcription factors, GATA1, NF-E2, and KLF1/EKLF, in the promoters of expressed genes at similar stages of erythroid differentiation in both species. These data suggested that key erythroid transcription factors direct groups of developmental and differentiation-stage specific genes in patterns that have evolved throughout evolution.

Investigation of major membrane protein gene expression revealed similar patterns of gene expression in both human and mouse. Genes encoding SPTA1, SPTB, EPB42, EPB41, TMOD1, and EPB49 increased during terminal erythroid differentiation. However, divergence between species did exist, for example, the human ACTB and TNFRSF1A genes showed no significant changes across differentiation stages, but their mouse orthologs decreased in late-stage erythroblasts, whereas human ANK1 and ADD1 gene expression showed minimal changes during differentiation but their orthologs increased during differentiation in mice. ADD3 gene expression increased in humans, but Add3 decreased in mice. The interpretation of these findings was that the composition of erythrocyte membranes is globally conserved in mammals; differences are present in certain proteins necessary for species-specific membrane maintenance.

A caveat of both of these studies is that no correlation between gene expression and protein composition is provided. For instance, although gamma adducin mRNA increased during erythroid differentiation in human cells and decreased during differentiation in murine cells, minimal to no gamma adducin protein is present in mature human erythrocytes, whereas it is readily detected in murine erythrocytes [30,31]. In addition, much of the data from these studies were derived from cultured erythroid cells. Obtaining unbiased transcriptomes from uncultured primary erythroid cells at varying developmental and differentiation stages is a priority for our understanding of erythropoiesis in both human and mouse.

ALTERNATE SPLICING

Studying patterns of alternative premessenger RNA splicing across the stages of human terminal erythroid differentiation, Pimentel et al. [32▪▪] discovered a complex program of alternative splicing enriched in genes controlling cell cycle, chromatin function, organelle organization, and RNA processing, with numerous alternative splicing events controlled in a differentiation stage-specific manner. Alterations in exon inclusion/exclusion efficiency were prominent in late-stage erythroblasts, paralleling the pancellular alterations associated with the final stages of terminal erythroid differentiation. Interestingly, a subset of splicing alterations was observed that introduce premature translation termination codons thereby decreasing the proportion of full-length coding mRNAs and downregulate gene expression via nonsense-mediated decay (Fig. 3) [32▪▪]. Utilization of posttranscriptional RNA processing pathways, particularly via premature termination codon usage/nonsense mediated decay, appears to be a signification mechanism regulating gene expression at the late stages of terminal erythroid differentiation. These data support the conclusion that a complex, dynamic, stage-specific program of alternative splicing is essential for human terminal erythroid differentiation, playing a critical role in cellular proliferation and differentiation in human erythropoiesis. It will be interesting to examine whether such a program also exists in murine erythropoiesis.

NONCODING RNAs

Utilization of high-throughput sequencing methodology for transcriptome analyses led to the observation that many noncoding RNAs (ncRNAs) are produced within the cell [33]. These ncRNAs are classified as housekeeping RNAs, microRNAs, siRNAs, PIWI-interacting RNAs, small ncRNAs (<200 nt in length), and long ncRNAs (lncRNAs, >200 nt in length).

Several studies have addressed the role of lncRNAs in erythropoiesis. Alvarez-Dominguez et al. [22▪▪] identified erythroid-specific lncRNAs in BFU-E, CFU-E, and erythroblasts and demonstrated that these lncRNAs demonstrated dynamic patterns of expression throughout erythropoiesis. Alterations in lncRNA expression during erythropoiesis were reflected at the chromatin level, as patterns of gene expression correlated well with patterns of histone architecture reflecting active and repressive chromatin, and erythroid transcription factors bound the promoter regions of erythroid cell expressed lncRNAs.

Paralkar et al. [21] studied lncRNAs obtained from transcriptome analyses of murine fetal liver erythroblasts, megakaryocytes, and megakaryocyte– erythroid progenitor cells. Remarkable lineage specificity was demonstrated for lncRNAs, with ~30% of high stringency lncRNAs detectable in only one of the three cell types examined. When compared with human, murine erythroid lncRNAs were very poorly conserved in both nucleotide sequence and expression, yet they were conserved across eight different mouse strains.

Together, these studies fit well with observations that lncRNA transcription is tightly regulated by cellular developmental or differentiationstage, external environment, and so on [34], and that there is poor conservation of lncRNA sequence across species [35].

TRANSCRIPTIONAL AND REGULATORY NETWORKS

The ENCODE Consortium mapped transcription, replication domains, DNAse I hypersensitivity, chromatin architecture, transcription factor occupancy, and replication-timing domains, and compared these in human and mouse [36–38,39▪▪]. Comparison of features obtained from mouse compared with human cells revealed several key results [36]. Although there is significant conservation of patterns of gene expression between species, expression profiles of many murine genes in specific biological pathways show significant divergence from their human orthologues. Much of the cis-regulatory landscape has diverged between mouse and human, but this divergence varies widely between different classes of regulatory elements in different cellular contexts. In addition, mouse and human transcription factor networks are much more conserved than cis-regulatory DNA. Chromatin domains, as studied via genome-wide analysis of DNA replication timing, are developmentally stable and evolutionarily conserved. A significant caveat to these studies is that they were completed on immortalized cell lines.

IMPLICATIONS

The availability of these data sets is essential for interpreting the transcriptional architecture of erythropoiesis. They can be leveraged to better understand steady state and stress erythropoiesis. These data can be used to create transcriptional circuits during hematopoietic cell development.

These data sets can be utilized to better understand disorders of erythropoiesis, including the bone marrow failure syndromes, aplastic anemia, the myelodysplasia syndromes, and erythroid disorders with prominent ineffective erythropoiesis. Comparative analyses between wild type and variant cells may provide insight into disease pathobiology, allowing understanding of mechanisms of abnormal erythropoiesis over time in specific diseases, and may provide insights into identification of potential therapeutic targets.

As noted by Pishesha et al. [19▪▪], global comparisons allow integration of data between human and mouse erythropoiesis, providing insight into why some human hematologic disorders are not recapitulated in mouse models and highlight the problems in translating therapeutic observations from mice to human.

CONCLUSION

Genome-wide based studies are rapidly advancing our knowledge of the transcriptomes of human and murine cells during hematopoietic development and differentiation. Understanding the features that regulate conserved and divergent features of human and murine erythropoiesis is critical. Comparing and contrasting the similarities and differences between human and murine erythropoiesis will provide novel insights into these processes while expanding our understanding of both. Together, these data can be leveraged to improve disease detection and diagnosis, understand and predict disease-specific complications, and develop novel therapeutic strategies for patients with hematologic disease.

KEY POINTS.

• There are many differences in programs regulating human and murine erythropoiesis.

• There are significant functional differences in human and murine mature erythrocytes.

• Genome-wide transcriptome profiling during erythropoiesis shows species-specific profiles.

• LncRNAs are poorly conserved between human and murine erythroid cells.

Acknowledgements

None.

Financial support and sponsorship

None.

This work was supported by NIH grants DK26263, HL65448, HL106184, DK100810 and DK104046.

Footnotes

REFERENCES AND RECOMMENDED READING

Papers of particular interest, published within the annual period of review, have been highlighted as:

▪ of special interest

▪▪ of outstanding interest