To understand the final stages of RBC maturation, we studied the transcriptomic landscapes of RBCs and their EVs using traditional RNAseq and single-cell RNAseq. Based on our single-cell analysis, there are significant differences in the transcriptomes of three RBC subpopulations. Two distinct subpopulations are defined by the expression of globin genes, and one is defined by the expression of the lncRNA, MALAT1, also expressed in bulk RNAseq data and verified by qPCR.
Indeed, lncRNAs in general and MALAT1 are dynamically expressed and regulated during erythropoiesis (13,24). Their depletion severely impaired erythrocyte maturation by inhibiting enucleation (25). Only some cells in the MALAT1 population contained globin transcripts. Instead, this population was enriched in transcripts coding for ribosomal and mitochondrial structural proteins. We assume that these subpopulations reflect the different stages of reticulocyte maturation and that other transcripts in the MALAT1 -population can only be recovered in the late stage of reticulocyte differentiation when not masked by the bulk globin transcripts and not being bound to organelles due to ceased biogenesis of ribosomes and mitochondria.
During their terminal differentiation in peripheral blood, anucleated reticulocytes adjust their protein expression profiles post-transcriptionally. Although not yet studied in RBCs, there is rapidly accumulating evidence for translational regulation executed by ribosome-associated lncRNAs (26,27). Our data support this, as we found co-expression of MALAT1 transcripts and a plethora of mRNAs coding for ribosomal structural proteins (Fig. 5A) in the same subpopulation of RBCs. Recent footprinting studies have shown that the ribosome is the primary destination for the majority of cytoplasmic lncRNAs (28). In addition, MALAT1 has also been shown to act as a miRNA sponge and as a potent autophagy inducer (29). Reticulocytes are known to employ miRNA-based mechanisms that could have a role in both protein translation control and degradation of mRNA during maturation (30–32). Intriguingly, MALAT1 hits from our single cell data all mapped to the same region, showing predictions of a short, cytosolic previously unknown transcript forms of MALAT1, instead of nuclear-retained, known long form of MALAT1. Since it is poorly polyadenylated (33), the full length MALAT1 transcript is unlikely captured in our poly-A-based single-cell assay in RBCs.
Obvious decrease in the mRNA quantity was observed across the subpopulations in single cell RNAseq data. To study support these findings, we employed imaging cell cytometry to correlate cellular RNA content to the expression of reticulocyte marker CD71. Decreasing trend in CD71 expression and intensity of TO staining was observed in CD71-enriched reticulocytes, which is known to indicate reticulocyte maturation (34,35). However, TO intensity and thus cellular RNA amount, was surprisingly heterogeneous, especially within CD71medium and CD71low populations. Furthermore, we observed varying morphologies across the populations, i.e., also concaved cells having high CD71 and TO intensity. This is partially in contrast to the study by Malleret et al. who concluded that CD71high and CD71medium reticulocytes share a similar gross morphology (large and multilobular) when compared to the smaller and increasingly concave reticulocytes seen in the CD71low and CD71neg populations (35), however, their imaging method was different from ours.
In our bulk RBC transcriptomes, both the number of expressed genes and the selection of highly expressed genes match well to the previously published ones (9,11,12) despite the different experimental set up used. The RBC and EV transcriptomes reflect the processes needed both in the reticulocyte maturation as well as for mature erythrocyte functions. During the two days in circulation, reticulocytes need to balance the production of Hb and final termination of protein translation processes. The enrichment of both protein synthesis -category and the mTOR pathway in our data reflect the uniquely high demand for Hb protein translation, supported by the finding that inhibition of mTOR resulted in reduced maturation of reticulocytes (36). Additionally, the enrichment of eIF2 signaling indicates active protein translation. In contrast, the enriched ubiquitin proteasome pathway and ubiquitination -category, also found to be enriched by Doss et al. (12), points to the targeted degradation of unnecessary proteins in reticulocytes (37) and ubiquitination is an essential step both in enucleation (38) and in mitochondrial autophagy (39). Interestingly, a recent proteomics study by Chu et al. comparing CD71-positive and -negative RBCs revealed the exactly same set of canonical pathways upregulated during reticulocyte maturation than our transcriptomic level data (40).
Autophagy was highly enriched process in RBC transcriptome and is crucial for the removal of mitochondria and ribosomes at the final maturation stage (41). Mitophagy can occur in reticulocytes through the autophagy-related gene (ATG) protein pathway (41) and through ATG-independent mechanism involving BCL2 interacting protein 3 like (BNIP3L)(42,43), which was one of the highly expressed transcripts in RBCs (bulk and single-cell data) and EVs, supporting the finding of Doss et al. (12). Players from both of these pathways were present in our data, except for the lysosomal compartment. According to the theory of Griffits and co-workers, autophagosomes could be combined with glycophorin A-coated vesicles, instead of lysosomes, and subsequently be removed by exocytosis (44). Our observations support this theory of organelle removal in EVs, as the RBC transcriptome was observed in EVs and especially enriched in mitochondrial transcripts and functions (oxidative phosphorylation, mitochondrial dysfunction), as well as transcripts for ribosomal proteins and translational machinery (eIF2 signaling). Whether RNA molecules are passively captured to EVs along with removal of, e.g., unwanted proteins and organelles or if there might be some active selection, remains to be explored.
EV transcriptome contained a group of expressed genes not found in RBCs. They most likely originate from EVs of residual plasma in the RBC product resulting from the manufacturing process. Consequently, there is inherent biological variation between EV transcriptomes arising from small but variable amounts of plasma EVs. However, the overall EV transcriptome is similar to that of RBCs with RBC function -related transcripts expressed at high level compared to contaminating ones indicating that vesiculation is truly a way to remove cellular components also from mature RBCs.
The most highly expressed transcript found in EVs but not in bulk RBC transcriptome, TMSB4X, was also found in some cells of MALAT1 -subpopulation and is previously associated with platelets (45). However, it is unlikely that we have platelet contamination in single cell data, since the same cells occasionally contained Hb and other RBC markers. In addition, cells were collected and purified in the same way for the bulk data, where platelet contamination was not detected. These transcripts could either have a very low expression in RBCs, hidden in the bulk data and only become detectable in single-cell data in the cells that have lower Hb expression. If TMSB4X and TMSB10 transcripts are truly of platelet origin, one hypothesis is that platelet-EVs may have adhered to or even fused with (46) RBCs, thereby becoming a part of the RBC transcriptome. The transfer of transcripts via EVs is a well-known phenomenon (18).