Peripheral blood mononuclear cells show prominent gene expression by erythroid progenitors in diseases characterized by heightened erythropoiesis

Using peripheral blood mononuclear cells (PBMCs) to develop diagnostic or prognostic biomarkers is attractive in diseases where biopsy of the affected tissue is difficult to obtain. The rationale for this approach is that peripheral immune cells are involved in a broad array of diseases. For example, peripheral blood monocytes, a component of PBMCs, mature to macrophages at the location of disease such as the lung perivascular area in pulmonary hypertension.¹ Genomic profiles derived from PBMCs have been used in biomarker development for a number of conditions, including cancer,² chronic obstructive pulmonary disease,³ heart disease,⁴ idiopathic pulmonary fibrosis,⁵ pulmonary arterial hypertension,⁶ metabolic syndrome,⁷ rheumatoid arthritis⁸ and sickle cell disease.⁹ PBMCs are peripheral blood cells with a round nucleus, which form a distinct layer during density gradient centrifugation of peripheral blood samples. PBMCs are comprised of lymphocytes (70–90%), monocytes (10–20%), dendritic cells (1–2%),¹⁰ and a trace amount of circulating stem cells including erythroid progenitors.¹¹ As a matter of fact, PBMCs are a source of erythroid progenitors allowing evaluation of their functional characteristics.¹² The cell types comprising PBMCs have distinct expression profiles¹³ and the cell composition of PBMCs can also depend on disease conditions.¹⁴ We observed that some of the most differentially expressed genes in PBMCs from patients with accentuated erythropoiesis are erythroid genes, which led us to test the hypothesis that this observation is related to an elevation of peripheral erythroid progenitors in the PBMC fraction in these diseases.

Chuvash erythrocytosis (CE) is a congenital disorder with elevated hypoxia-inducible factor (HIF) signalling at normoxia due to the homozygous VHL^R200W mutation. It is characterized by heightened erythropoiesis as manifested by elevated erythropoietin levels and red blood cell counts. In a previous microarray gene expression study using PBMCs, we found that many of the 29 genes whose expression levels increased by ≥1·5-fold in CE relative to healthy individuals appeared to have erythrocyte- rather than PBMC-related functions.¹⁵ In the present study, we performed RNA sequencing in purified cell fractions of reticulocytes, platelets and granulocytes from CE patients and assessed the expression patterns of the genes that were up-regulated in PBMCs by ≥1·5-fold. As the RNA preparation from reticulocytes includes a haemoglobin RNA depletion step, the HBB and HBD genes, encoding the beta and delta globin chain, respectively, were removed from further analysis, resulting in 27 genes. Based on expression correlation among cell fractions of reticulocytes, platelets and granulocytes, the 27 genes were separated into two distinct hierarchical clusters (Fig 1A). The 16 genes in Cluster I were in general related to erythroid function (Table SI) and were enriched in the Reactome pathway, ‘Erythrocytes take up carbon dioxide and release oxygen’ (Benjamini–Hochberg-adjusted P = 0·068). The 11 genes in Cluster II appeared to have a role in immune responses (Table SI) and were enriched in the Reactome pathway, ‘Interferon alpha/beta signalling’ (adjusted P = 0·025).

Fig 1.

Erythroid gene expression in PBMCs from Chuvash erythrocytosis (CE) and sickle cell anaemia (SCA) patients. (A) Hierarchical clustering of gene expression levels obtained by RNA-seq among reticulocytes (n = 5), platelets (n = 3) and granulocytes (n = 8) in CE patients. (B) Mean log2 expression level (open circles) and standard error (vertical lines) of the Cluster I (orange) and Cluster II (blue) genes in reticulocytes, platelets and granulocytes in CE patients. (C) Correlation of PC1 of Cluster I gene expression in PBMCs with plasma erythropoietin concentration in CE patients. (D) Correlation of log2 fold change of PBMC expression in CE with that in SCA (red) and idiopathic pulmonary fibrosis (IPF; black), for Cluster I genes. (E) Correlation of PC1 of Cluster I gene expression in PBMCs with plasma erythropoietin concentration in SCA patients.

We then looked at the expression of Cluster I genes in reticulocytes, as a proxy of erythroid progenitors, and in platelets and granulocytes, as proxies of non-erythroid cells. The three cell fractions are excluded from the PBMC fraction during density gradient centrifugation. High expression levels of the Cluster I genes were observed specifically in reticulocytes, but not in platelets or granulocytes (Fig 1B). Reticulocyte expression of the majority of Cluster I genes was not significantly different between CE patients and healthy individuals (adjusted P > 0·05). This suggests that Cluster I gene expression observed in PBMCs of CE patients likely originates from erythroid progenitors. In keeping with this conclusion, the first principal component (PC1) of Cluster I gene expression in PBMCs, which explained 95% of total expression variance, correlated with plasma erythropoietin concentration in 43 CE patients (r = 0·72, P = 5×10⁻⁸; Fig 1C), suggesting that the amount of erythroid progenitors in PBMCs is proportional to HIF-induced erythropoiesis.

For replication of these observations, we examined Cluster I gene expression in PBMCs of patients with sickle cell anaemia (SCA),¹⁶ a condition with an elevated hypoxic response due to haemolytic anaemia. SCA is characterized by markedly heightened erythropoiesis. The PBMC separation protocol for the SCA patients was similar to that for the CE patients. The change of Cluster I gene expression in PBMCs of SCA patients¹⁶ versus controls was compared to that of CE¹⁵ versus controls. The expression pattern in SCA was similar to the expression pattern in CE (Fig 1D). Similar to the findings in CE patients, the PC1 of Cluster I gene expression in PBMCs from 113 sickle cell disease patients¹⁷ also strongly correlated with plasma erythropoietin concentration (r = 0·49, P = 3×10⁻⁸; Fig 1E).

As a final test of our hypothesis, we examined PBMC Cluster I gene expression in patients with diffuse idiopathic pulmonary fibrosis (IPF),⁵ characterized by pulmonary impairment that is associated with systemic hypoxia but lacks an erythropoietic response to hypoxaemia despite augmented erythropoietin levels.¹⁸ In contrast to SCA, the expression of Cluster I genes compared to controls in IPF did not correlate with the expression pattern in CE PBMCs (Fig 1D); this provides further support that Cluster I gene expression in PBMCs from CE and SCA patients reflects erythroid progenitor gene expression.

In conclusion, our analyses demonstrate that heightened erythropoiesis contributes to erythroid-specific gene expression in PBMCs in CE and SCA, genes that show marked differential expression compared to controls in these conditions. In keeping with our findings, a gene expression signature of erythrocyte maturation was reported in PBMCs from patients with idiopathic pulmonary arterial hypertension where hypoxia may trigger the expansion of immature erythrocyte precursors.⁶ Although heightened erythropoiesis itself can reflect the severity of the disease,⁹ the PBMC expression profile is heterogeneous due to the expected variation in both cell composition and cell-specific gene expression. Our findings highlight the pitfalls of using PBMCs to define disease-specific expression profiles and point to the need to perform such analyses in purified fractions of peripheral blood cells, such as granulocytes, platelets and reticulocytes.

Supplementary Material

Table SI. Genes in Cluster I and Cluster II. Function is based on UniProtKB/Swiss‐Prot and the provided reference. Within‐cluster genes ordered by fold changes in PBMCs of CE patients.

Table SI. Genes in Cluster I and Cluster II. Function is based on UniProtKB/Swiss-Prot and the provided reference. Within-cluster genes ordered by fold changes in PBMCs of CE patients.