Abstract

   The RVd therapy, combining lenalidomide, bortezomib, and dexamethasone,
   is a mainstay treatment for multiple myeloma. A multiple myeloma
   patient developed pure red cell aplasia (PRCA) following RVd treatment,
   despite the absence of common PRCA triggers. In vitro analyses reveal
   lenalidomide as a pivotal disruptor of erythropoiesis. Single-cell
   transcriptome analysis unveils hyperactive CD8^+ T cells and impaired
   erythropoiesis in the patient’s bone marrow. Unexpectedly, the
   patient’s erythroid cells display abnormally high expression of genes
   in the antigen presentation pathway, particularly those for major
   histocompatibility class I (MHC-I) molecules. Functional assays
   demonstrate that lenalidomide treatment further augmented MHC-I
   expression in the patient’s erythroid cells. Blocking MHC-I or
   depleting T cells alleviates the defective erythropoiesis of PRCA,
   suggesting that the interaction between erythroid cells with elevated
   MHC-I and T cells in the bone marrow might contribute to PRCA. Taken
   together, our study implicates a mechanism underlying
   lenalidomide-induced PRCA in treating cancer patients.

   Subject terms: Haematological diseases, Molecular medicine, MHC class
   I, Gene regulation in immune cells, Cytotoxic T cells
     __________________________________________________________________

   On rare occasions, some multiple myeloma patients develop pure red cell
   aplasia (PRCA) following combination therapy for multiple myeloma. Here
   the authors characterise bone marrow cells from PRCA patients by single
   cell sequencing showing hyperactive CD8^+ T cells, impaired
   erythropoiesis, increase of MHC-I proteins and antigen presentation
   associated genes in erythroid cells.

Introduction

   Pure red blood cell aplasia (PRCA) is a rare hematological disorder
   characterized by near absence of erythroblasts in the bone marrow and
   impairment of erythropoiesis^[44]1,[45]2. Acquired PRCA may be
   associated with various factors, including immune disorders, B19
   parvovirus infection^[46]3,[47]4, antibodies triggered by recombinant
   human erythropoietin, thymoma, and lymphoproliferative disorders (e.g.,
   chronic lymphocytic leukemia and large granular lymphocyte
   leukemia)^[48]2,[49]5–[50]8. Furthermore, over 50 drugs and chemicals
   have been associated with the onset of PRCA^[51]9–[52]11. Nonetheless,
   the mechanisms underlying most drug-induced PRCA remain elusive.

   RVd therapy, comprising lenalidomide, the proteasome inhibitor
   bortezomib, and the glucocorticoid dexamethasone, has been approved as
   the preferred treatment regimen for the majority of multiple myeloma
   (MM) patients^[53]12–[54]15. Anemia frequently afflicts individuals
   with MM^[55]16. Previous clinical reports have consistently shown
   significant reductions in M protein levels and the number of myeloma
   cells following RVd treatment^[56]13,[57]17,[58]18, often coinciding
   with the remission of anemia. Apart from that, lenalidomide, an
   immunomodulatory drug, has also been widely used in the treatment of
   patients with 5q-myelodysplastic syndrome^[59]19. It modulates the host
   immune system to promote the immune-mediated clearance of cancer cells
   by inducing T-cell costimulation and proliferation, as well as
   increasing the production of interleukin (IL)-2 and interferon (IFN)
   γ^[60]20–[61]22 Additionally, lenalidomide was identified to bind to
   the CRBN-CRL4 E3 ubiquitin ligase. It functions in multiple myeloma by
   promoting the ubiquitination of two transcription factors, IKZF1 and
   IKZF3, and the degradation of casein kinase 1A1 (CK1α), leading to
   clinical efficacy in del(5q) MDS^[62]23,[63]24.

   A few clinical cases have documented recurrent PRCA induced by
   lenalidomide in patients with myelodysplastic syndrome^[64]25 or
   multiple myeloma^[65]26; however, the mechanisms underlying
   lenalidomide-induced PRCA have not been elucidated. Here, we present a
   case of a multiple myeloma patient who, after undergoing three courses
   of RVd treatment, developed PRCA. This condition was characterized by a
   paucity of erythroid progenitors in the bone marrow, a hemoglobin level
   of 50 g/L, and a reticulocyte percentage of 0.04%. Clinical assessment
   of this patient ruled out the common underlying conditions leading to
   PRCA, therefore indicating the association between the RVd treatment
   and the development of PRCA.

   In this study, we comprehensively employed single-cell transcriptome
   analysis, proteomic analysis and functional assays to continuously
   monitor the cellular status of the case under investigation. The
   functional assays revealed a key role of lenalidomide, as opposed to
   bortezomib or dexamethasone, in the development of PRCA in this
   patient. We not only identified hyperactive CD8^+ T cells, which likely
   inflicted the injury on erythroid progenitors, but also found that the
   patient’s erythroblasts showed elevated MHC-I expression, a phenomenon
   induced by lenalidomide. The drug-induced augmentation of MHC-I on
   erythroblasts is associated with the development of PRCA and may serve
   as a biomarker for other clinical PRCA conditions. Overall, our
   findings illuminate the potential mechanism underlying drug-induced
   PRCA and provide insights concerning the roles of aberrant erythroid
   progenitors in affected individuals.

Results

Clinical manifestations of drug-induced PRCA in a patient with multiple
myeloma

   This study presents the case of a rare patient with multiple myeloma
   (Table [66]1). Anemia was noted during routine examination, with a
   hemoglobin level of 82 g/L and a reticulocyte count of 0.0351 × 10^9/mL
   (reference range: 0.024–0.084 × 10^9/mL). At this juncture, the bone
   marrow exhibited active erythroid hyperplasia, characterized by a
   granular/erythroid cell ratio of 3.4:1. The patient underwent three
   cycles of RVd therapy and achieved a very good partial response (VGPR).
   Pathological examination of a bone marrow aspirate specimen revealed
   the plasma cells dropped from 32.5% to 3.5%. However, the anemia
   persisted and worsened. Subsequent blood tests showed a drop in the
   hemoglobin level to 55 g/L, with a reticulocyte count of
   0.0005 × 10^9/mL, and EPO levels exceeding 758 mIU/mL. The Coombs test
   was negative, and no paroxysmal nocturnal hemoglobinuria (PNH) clone
   was detected. Further bone marrow examination confirmed the absence of
   erythroid lineage cells, leading to the diagnosis of PRCA. We found no
   evidence of parvovirus, CMV, or EBV infection. Both bone marrow B cell
   and TCR-Vβ flow cytometry returned negative results. CT scans reveal no
   thymoma.

Table 1.

   Patient characteristics during disease progression
   Stage P1
   Multiple myeloma Stage P2
   Initial onset of PRCA Stage P3
   Partial remission Stage P4
   PRCA recurrence Stage P5
   Cyclosporine treatment failure
   M protein 18.1 g/L 0.4 g/L 0 0 1.6 g/L
   Bone marrow plasma cells 32.50% 3.50% 3.00% 2.50% 2%
   Bone marrow myeloid/erythroid ratio 3.4:1 124:0 16.22: 1 53.67:1 12:01
   Hemoglobin level 82 g/L 55 g/L 69 g/L 42 g/L 56 g/L
   Reticulocyte ratio 1.64% 0.03% 0.48% 0.32% 0.54%
   Peripheral blood CD8^+ T cell count – 801/µL - 487/µL 467/µL
   Peripheral blood CD4^+ T cell count – 639/µL - 378/µL 361/µL
   [67]Open in a new tab

   Initially, this patient responded well to immunosuppressant. After five
   months of treatment with cyclosporine A (CsA), the level of Hb was
   restored to 109 g/L, achieving partial remission. However, three months
   after suspension, the level of Hb dropped again to 40 g/L, with
   reticulocyte count of 0.0039 × 10^9/mL, and a ratio of 0.34%, all while
   erythroid lineage cells remained absent in the bone marrow.
   Consequently, allogeneic hematopoietic stem cell transplantation
   (allo-HSCT) was scheduled for administration.

Impaired erythropoiesis in PRCA patients is associated with lenalidomide
treatment

   To investigate the mechanism underpinning the onset of PRCA, we
   collected bone marrow samples from the patient during partial remission
   after CsA treatment (P3), at recurrence (P4), and during treatment
   failure (P5) (Fig. [68]1A). Notably, in contrast to the MM stage, bone
   marrow smears during the PRCA stage exhibited a virtual absence of
   erythroid precursors, consistent with the patient’s clinical
   presentation (Fig. [69]1B). We observed a dramatic reduction in the
   percentage of CD71^+CD235a^+ erythroid precursors in the PRCA patient’s
   bone marrow compared to that of healthy donors (HDs) (Fig. [70]1C, D).
   However, there were no significant differences in the percentages of
   KIT^+ hematopoietic stem and progenitor cells (HSPCs) (Fig. [71]1E;
   Supplemental Fig. [72]1A). Therefore, the impaired erythropoiesis was
   not attributable to a shortage of HSPCs. Although bone marrow
   mononuclear cells (BMMCs) from the PRCA patient could undergo erythroid
   differentiation, they developed substantially slower than cells from
   HDs, and the enucleation rate was also reduced (Fig. [73]1F, G;
   Supplemental Fig. [74]1B). Furthermore, the patient’s erythroblasts
   displayed a severe decrease in their proliferative capacity
   (Fig. [75]1H). Colony-forming assays of the PRCA patient’s BMMCs showed
   a substantial reduction in both the number and size of burst-forming
   units-erythroid (BFU-E) colonies, whereas the numbers of other myeloid
   colonies (CFU-GEMM, CFU-GM) were unaffected (Fig. [76]1I–K;
   Supplemental Fig. [77]1C). Collectively, our results indicated that the
   deficiency in erythropoiesis was attributable to a diminished number
   and impaired differentiation capacity of erythroid cells.

Fig. 1. Clinical assessment and in vitro analysis of erythroid development in
the PRCA patient.

   [78]Fig. 1
   [79]Open in a new tab

   A Disease course and therapeutic regimen. B Morphology of cells in the
   patient’s bone marrow at the MM stage (left) and the PRCA stage
   (right). Similar results were obtained based on 30 fields of view under
   the microscope for the patient’s bone marrow smears at each stage. C
   Representative flow cytometry results showing percentages of
   erythroblasts in the bone marrow of healthy donors (HD, upper) and PRCA
   patients (lower). D, E Graphs quantifying percentages of erythroid
   precursors (D) and HSPCs (E) in HD (n = 5; samples from five
   individuals as biological replicates) and PRCA (n = 3; samples from
   three PRCA timepoints of the patient as biological replicates). Data
   are presented as mean values ± SEM. Two-tailed paired Student’s t-test.
   F Flow cytometry analysis of BMMCs from the PRCA patient and HD
   undergoing ex vivo erythroid culture. Erythroid differentiation markers
   CD71 and CD235a were assessed on days 6 and 10. G Erythroid cell
   enucleation was analyzed by flow cytometry on day 14 (n = 4; biological
   replicates). Data are presented as mean values ± SEM. Two-tailed paired
   Student’s t-test. H BMMCs from PRCA and HD were cultured in a
   serum-free erythroid medium. Cell numbers were counted every 3 days
   from three independent biological replicates. Data are presented as
   mean values ± SEM. Two-tailed paired Student’s t-test. I–K
   Colony-forming assays were performed using BMMCs from HD (left) or PRCA
   (right). Representative micrographs of BFU-E colonies are shown in
   panel (I). Area of BFU-E colonies (J) and colony numbers of BFU-E,
   CFU-GM, and colony-forming unit-granulocyte, erythroid, macrophage, and
   megakaryocyte (CFU-GEMM) (K) were quantified (n = 3; biological
   replicates). Data are presented as mean values ± SEM. Two-tailed paired
   Student’s t-test. Source data are provided as a Source Data file.

   This patient had not been diagnosed with any common triggers of
   PRCA^[80]1. However, PRCA manifested after three courses of RVd
   treatment, implying a potential contribution of RVd drugs to its
   development. With the addition of RVd in the erythroid culture medium,
   both cells from HD and the PRCA patient exhibited a delay in erythroid
   differentiation compared to the control (DMSO treated). However, HD
   cells eventually underwent erythroid differentiation, achieving a
   significantly higher number of erythroid cells compared to the control.
   In contrast, PRCA cells showed severely impaired cell proliferation and
   differentiation (Supplemental Fig. [81]2A–C). To identify the specific
   drug responsible for PRCA, we systematically tested each component of
   the RVd regimen in a primary human erythroid culture (Fig. [82]2A–C;
   Supplemental Fig. [83]2D). Dexamethasone considerably increased
   erythroid proliferative capacity, consistent with previous
   reports^[84]27. Bortezomib had no discernible effect on the patient’s
   BMMCs. Notably, lenalidomide exhibited a modest promotion of
   proliferation in cells from healthy donors but led to a substantial
   delay in erythroid differentiation and impairment of proliferation in
   the patient’s cells. Moreover, lenalidomide induced a significant
   increase in apoptosis in the patient’s cells (Fig. [85]2D, E).
   Consistently, the addition of lenalidomide profoundly reduced the
   number of BFU-E colonies but did not impact CFU-GM colonies
   (Fig. [86]2F, G; Supplemental Fig. [87]3A, B). Collectively, these
   findings strongly suggest that lenalidomide played a crucial role in
   inducing PRCA in this patient by specifically targeting erythropoiesis.

Fig. 2. Lenalidomide has an inhibitory effect on in vitro erythroid
development in cells from the PRCA patient.

   [88]Fig. 2
   [89]Open in a new tab

   A BMMCs from healthy donors (HD, left) and PRCA (right) were cultured
   in a serum-free erythroid medium with indicated treatments. Cell
   numbers were counted every 4 days from three independent biological
   replicates. DMSO was added as the untreated control group. Data are
   presented as mean values ± SEM. Two-tailed paired Student’s t-test. B
   Quantification of the impact of drugs on erythroid proliferation. The
   fold change indicates the drug’s impact on the final number of
   erythrocytes from the same donor, relative to the untreated group
   (n = 3, biological replicates). Data are presented as mean
   values ± SEM. Two-tailed paired Student’s t-test. C BMMCs from HD
   (left) and PRCA (right) were cultured in a serum-free erythroid medium,
   with or without lenalidomide (LENA). Flow cytometry analyses were
   performed on days 8 and 16. D Apoptosis assay on BMMCs cultured for 9
   days in erythroid medium, using 4’,6-diamidino-2-phenylindole (DAPI)
   and Annexin-V. E Quantification of cells from panel (D) undergoing
   early and late apoptosis(apop.) (n = 3, biological replicates). Data
   are presented as mean values ± SEM. Two-tailed paired Student’s t-test.
   F Colony-forming assays of BMMCs from PRCA or HD with indicated
   treatments. Quantification of BFU-E, CFU-GM, and CFU-GEMM were
   performed on day 14 (n = 3, biological replicates). Data are presented
   as mean values ± SEM. Two-tailed paired Student’s t-test. G Percentage
   of BFU-E colonies in each group (n = 3, biological replicates). Data
   are presented as mean values ± SEM. Two-tailed paired Student’s t-test.
   Source data are provided as a Source Data file.

scRNA-Seq analysis revealed significant alterations in the percentages of
multiple cell lineages in the bone marrow of PRCA

   We next performed single-cell RNA-Seq (scRNA-Seq) analysis on BMMCs
   from HDs and the PRCA patient at stages of partial remission and
   recurrence (Fig. [90]3A, B). Analysis of cell composition revealed that
   the percentages of B cell lineage and plasma cells had significantly
   decreased in the patient’s bone marrow, confirming the remission state
   of multiple myeloma (Supplemental Fig. [91]4A, B). scRNA-Seq revealed a
   relative increase in the percentage of CD8^+ T cells in the patient,
   which is consistent with previous reports (Supplemental Fig. [92]4B,
   C)^[93]28–[94]30. We further conducted clustering of T cells, which
   revealed an increase in the relative percentages of CD8^+ memory and
   cytotoxic cells in the patient, exhibiting elevated cytotoxicity
   (Supplemental Fig. [95]4D). Consistent with our analysis in
   Fig. [96]1C–E, CD71^+ erythroid cells in the patient’s bone marrow were
   significantly reduced, although there were no significant changes in
   HSPCs (Supplemental Fig. [97]4E, F). Moreover, the patient’s erythroid
   cells were primarily concentrated in the progenitor stage,
   characterized by the expression of genes such as GATA2, CD34, and KIT
   (Fig. [98]3C). Conversely, the patient had a lower percentage of
   erythroblasts, which typically express the genes such as hemoglobin
   gene HBB and GYPA (Supplemental Fig. [99]4G, H). Gene set enrichment
   analysis (GSEA) revealed that cell cycle and development-related
   pathways were downregulated in erythroid cells of PRCA, indicating
   aberrant erythroid development and proliferation. In contrast, the
   patient’s erythroid progenitors exhibited a marked activation of
   immune-related pathways, including those for MHC-I-mediated antigen
   presentation (Fig. [100]3D). This activity extended from proteasome and
   peptide transport genes to peptide-loading complex (PLC) genes and
   MHC-I heterodimer molecules (Fig. [101]3E). Notably, erythroid cells
   showed the greatest fold increase in MHC-I upregulation compared to
   other cell types (Supplemental Fig. [102]5A). These alterations in
   transcriptional profiles suggested the involvement of an abnormal
   immune activation mechanism in erythroid progenitors and precursors
   (Supplemental Fig. [103]5B, C).

Fig. 3. MHC-I genes were increased in erythroid cells of PRCA.

   [104]Fig. 3
   [105]Open in a new tab

   A Experimental scheme for the analysis of clinical samples. B Overview
   of cell clusters in integrated single-cell transcriptomes derived from
   BMMCs of healthy donors (HD, upper) and the PRCA patient (lower).
   Clusters were named based on cluster-specific gene expression patterns.
   (HSPC hematopoietic stem and progenitor cell, GMP granulocyte–monocyte
   progenitors, ery erythroid cells, NK natural killer cells, pDC
   plasmacytoid dendritic cells, cDC conventional (classical) dendritic
   cells, ncMono[comp] C1QA^+/CD16^+ complement-expressing non-classical
   monocytes, ncMono CD14^dimCD16^+ non-classical monocytes, cMono
   CD14^++CD16^+ classical monocytes). C UMAP analysis of erythroid cells.
   HD and PRCA samples were combined and colored according to the donor.
   Expression levels of erythroid markers were used to determine cell
   stages. D GSEA pathway enrichment results for different cell types. The
   dot color represents the q-value; the dot size represents the number of
   genes related to the indicated pathway. To enable a more accurate
   comparison of transcriptional differences, we reduced the number of
   cells in the HD samples to match those in the PRCA samples (as detailed
   in “Methods”). E Expression levels of antigen processing and peptide
   loading genes, proteasome and immunoproteasome genes, and MHC-I
   heterodimer genes in erythroid cells based on scRNA-Seq. Erythroid
   lineage-related cells were divided based on specific markers into
   progenitor cells (including megakaryocyte–erythroid progenitor cells
   (MEP), BFU-E and colony-forming unit-erythroid (CFU-E)), early
   precursor cells (consisting of proerythroblasts and basophilic
   erythroblasts (Pro/Baso-E)), and late precursor cells (consisting of
   polychromatic erythroblasts and orthochromatic erythroblasts
   (Poly/Ortho-E)). F Representative flow cytometry results showing MHC-I
   levels on erythroid cells in the bone marrow from HD (upper) and the
   PRCA patient (lower). G Relative expression levels of HLA-B, detected
   by quantitative RT-PCR. BMMCs from HD and PRCA patients were cultured
   in erythroid differentiation medium and sorted into specific stages
   based on CD71 and CD235a expression by FACS: CD71^+CD235a^− (S2),
   CD71^+CD235a^+ (S3), and CD71^-CD235a^+ (S4) (n = 3, biological
   replicates). Data are presented as mean values ± SEM. Two-tailed paired
   Student’s t-test. Source data are provided as a Source Data file.

   Indeed, unlike the usual decline in MHC-I gene expression seen from
   progenitors to precursors in HDs, the patient’s erythroid precursors
   sustained a high expression of MHC-I genes. Flow cytometry analyses
   revealed that the protein level of MHC-I molecules on the surface of
   CD235a^+ erythroblasts in the patient is significantly higher compared
   to those in HDs (Fig. [106]3F). In addition, we collected peripheral
   blood samples from another patient with multiple myeloma who developed
   PRCA after lenalidomide treatment^[107]31. Consistent with our
   observation in the bone marrow, MHC-I upregulation in erythroid cells
   was also evident in the peripheral blood samples of both MM patients
   with induced PRCA after treatment. However, this phenomenon was not
   present in typical RVd-treated MM patients who did not develop PRCA
   after RVd treatment (Supplemental Fig. [108]6).

   In vitro cell cultures showed that even the terminally differentiated
   CD71^−CD235a^+ erythroblasts from the patient persistently manifested a
   pronounced up-regulation of HLA-B expression (Fig. [109]3G).

Lenalidomide induces an increase in MHC-I molecules and upregulation of
related genes in PRCA erythroblasts

   We next tested whether lenalidomide can induce the expression of MHC-I
   on erythroid cells directly. To test that, BMMCs from the patient or
   HDs were cultured under conditions that supports erythroid
   differentiation (Fig. [110]4A). The addition of lenalidomide resulted
   in further elevation in the expression of MHC-I molecules on the
   patient’s CD71^+CD235a^+ erythroblasts without any effect on those
   cells of HD’s (Fig. [111]4B). To elucidate the mechanism underlying
   defective erythropoiesis in the PRCA patient, we isolated CD71^+
   erythroid progenitors after 6 days of in vitro erythroid
   differentiation of BMMC with or without lenalidomide for bulk RNA
   sequencing (Supplemental Fig. [112]7).

Fig. 4. Lenalidomide induced elevation of MHC-I expression levels in
erythroblasts from PRCA.

   [113]Fig. 4
   [114]Open in a new tab

   A Experimental scheme for the analysis of the drug function. B
   Histogram of MHC-I level on CD71^+CD235a^+ erythroblasts after 12 days
   of culture in the groups of untreated BMMCs and BMMCs with lenalidomide
   (all BMMCs from PRCA or HD). C Volcano plot of differentially expressed
   genes (DEGs) after lenalidomide treatment. The y-axis shows DEGs in
   BMMCs from the PRCA patient; the x-axis represents DEGs after
   lenalidomide addition, compared with untreated BMMCs from HD. The slope
   of the auxiliary line is 1. PRCA-specific unregulated or downregulated
   genes are shown in red and blue, respectively. D Top GSEA pathway
   enrichment results for PRCA-specific genes. Normalized enrichment
   scores (NES) are shown, with the color denoting the FDR q-value. E
   Relative expression levels of antigen processing and presentation and
   regulatory network for MHC-I related genes, detected in bulk RNA-seq.
   Source data are provided as a Source Data file.

   We compared the fold changes of DEGs after lenalidomide addition
   between cells from HDs and those from the patient (Fig. [115]4C). Genes
   that exhibited greater up-regulation after lenalidomide addition in the
   patient’s cells were regarded as “PRCA-specific upregulated genes”,
   while genes that displayed greater downregulation after lenalidomide
   addition in the patient’s cells were regarded as “PRCA-specific
   downregulated genes.” In total, there were 1629 PRCA-specific
   upregulated genes and 1276 PRCA-specific downregulated genes. In
   consistency with scRNA-seq results, GSEA analysis showed that
   upregulated genes of the PRCA patient treated with lenalidomide are
   involved in antigen presentation, immune pathways as well as cell
   death, whereas down-regulated genes are related to cell cycle and DNA
   replication (Fig. [116]4D). Genes encoding MHC proteins and factors
   involved in antigen presenting pathways, are upregulated significantly
   in erythroid cells from the patient with the addition of lenalidomide
   (Fig. [117]4E). Lenalidomide might augment IL-2 receptor levels in the
   patient’s erythroid progenitors, thereby bolstering the cellular
   responses to IL-2 and IFNs, leading to an upregulation of transcription
   factor genes such as NLRC5 and IRF1^[118]32 These results confirmed the
   disruptive role of lenalidomide in erythropoiesis, which might be
   through upregulation of MHC molecules and antigen-presenting pathways.

Proteomic analysis reveals the activation of MHC-I and ubiquitination
pathways in PRCA patients

   Lenalidomide was identified to promote the ubiquitination and
   degradation of specific substrates^[119]23,[120]24. To further examine
   the role of lenalidomide in PRCA, we performed micro-proteomics
   analysis on CD71^+ erythroid progenitor cells sorted by FACS after 6
   days of erythroid culture. The inter-sample analysis was conducted
   based on the results of mass spectrometry (Fig. [121]4A and
   Supplemental Data [122]1). Differential gene clustering analysis of the
   detected proteins yielded four clusters: Cluster 1 and Cluster 4
   contained proteins that were significantly more abundant in PRCA
   patients and healthy donors, respectively. Cluster 2 and Cluster 3
   contained proteins that were significantly more abundant after in vitro
   lenalidomide treatment in healthy donor and PRCA patient cells,
   respectively (Fig. [123]5A).

Fig. 5. Proteomics analysis reveals the activation of MHC-I and
ubiquitination pathways in PRCA.

   [124]Fig. 5
   [125]Open in a new tab

   A Heatmap of proteins detected by micro-proteomics analysis. Clusters
   1–4 were identified by hierarchical clustering. B KEGG pathway
   enrichment results for different clusters. Fisher’s exact test was used
   for data analysis, with adjustments made for multiple comparisons. The
   dot color represents the −Log10 q-value; the dot size represents the
   number of genes related to the indicated pathway. C Relative protein
   levels of ubiquitin-related genes, detected in micro-proteomics
   analysis. D Relative protein levels of antigen processing and
   presentation and the regulatory network for MHC-I related genes,
   detected in micro-proteomics analysis. Source data are provided as a
   Source Data file.

   Further comparison of the differentially enriched proteins in these
   clusters revealed that the ubiquitin-mediated proteolysis pathway is
   upregulated in PRCA, while the antigen processing and presentation
   pathway is distinctly upregulated in PRCA after lenalidomide treatment
   (Fig. [126]5B). The upregulated proteins in the ubiquitin-related
   pathway in PRCA include various E1 activating enzyme, E2
   ubiquitin-conjugating enzymes, E3 ubiquitin-protein ligases (including
   the Cullin ubiquitin ligases), and E4 ubiquitin conjugation factors
   (Fig. [127]5C).

   The MHC-I pathway enriched in Cluster 3 includes proteins involved in
   the peptide-loading complex (PLC), MHC-I heterodimer, and proteasome
   required for peptide degradation, with a notable enrichment of proteins
   related to the immunoproteasome (i.e., PSMB8, PSMB9, and PSMB10). The
   levels of these proteins were further upregulated in PRCA cells after
   the addition of lenalidomide, whereas no significant changes were
   observed between lenalidomide-treated and untreated BMMCs from healthy
   donors (Fig. [128]5D).

   Collectively, the proteomic analysis reveals higher levels of
   ubiquitination-related pathway protein expression in PRCA cells
   compared to HD. Moreover, the analysis shows upregulation of MHC-I
   pathway protein expression in PRCA cells, which is further induced by
   lenalidomide treatment, consistent with our previous findings from
   single-cell transcriptome analysis.

Blocking of MHC-I-TCR interaction rescues erythropoiesis from the PRCA
patient

   Interestingly, we also observed enhanced interactions between CD8^+ T
   cells and erythroid cells in the PRCA patient. Notably, CD8^+ T cells
   from the patient displayed elevated levels of FAS ligand (FASLG)–FAS
   interactions, which can directly induce apoptosis of erythroid cells
   (Fig. [129]6A). Single-cell RNA-seq analysis revealed that, compared to
   HDs, T cells, particularly CD8^+ T cells, were hyperactivated in the
   patient (Fig. [130]6B, C).

Fig. 6. Blocking MHC-I or T cell depletion rescued erythropoiesis of PRCA.

   [131]Fig. 6
   [132]Open in a new tab

   A Circle plot showing changes in the interaction intensity of
   ligand–receptor pairs between CD8^+ T cells and erythroid cells in PRCA
   vs. HD samples. B UMAP plot of annotated clusters from T cells
   (n = 16,000) in PRCA and HD BMMC samples. C Cytotoxicity scores in T
   cells from the PRCA patient and HD. D BMMCs from the PRCA patient
   (left) and HD (right) were cultured in a serum-free erythroid medium,
   with or without T-cell depletion. Flow cytometry analyses were
   performed on days 7 and 11. E Cell proliferation curve demonstrating
   effects of T cells on erythropoiesis. Cell numbers were counted every 4
   days, from three independent biological replicates. Data are presented
   as mean values ± SEM. Two-tailed paired Student’s t-test. (*P < 0.05).
   F, G BMMCs from HD and PRCA were cultured in the coculture medium, with
   or without anti-HLA class I antibody (Ab). CD235a^+ erythroid cells (F)
   and cells undergoing apoptosis (G) were measured on day 9 (n = 3,
   biological replicates). Data are presented as mean values ± SEM.
   Two-tailed paired Student’s t-test. Source data are provided as a
   Source Data file.

   To confirm the role of T cells in PRCA, BMMCs from HDs and the PRCA
   patient were depleted of T cells before being cultured in the erythroid
   medium. Our results showed that while T cell depletion did not yield
   any effect on cell proliferation from HDs, it markedly alleviated
   defective erythroid proliferation and differentiation in PRCA cells
   (Fig. [133]6D, E).

   In addition, a marked enhancement of interactions between MHC-I and CD3
   was observed (Fig. [134]6A). MHC-I serves an essential ligand that
   engages with TCR on the surface of T cells^[135]33,[136]34. The
   increased expression of MHC-I on erythroid cells enables us to
   hypothesize that targeting MHC-I-TCR interaction might rescue defective
   erythropoiesis in PRCA. To test that, we added antibodies specific for
   HLA-A, B, and C of MHC-I into the erythroid culture of BMMCs^[137]35.
   After a 9-day culture period, it was evident that the addition of HLA
   antibody considerably promoted the differentiation of the erythroid
   lineage in PRCA but not in HDs (Fig. [138]6F, G; Supplemental
   Fig. [139]8). We also observed a significant decrease in the percentage
   of apoptotic erythroid cells from the patient. Therefore, blocking the
   MHC-I-TCR interaction alleviates erythropoiesis, indicating that the
   MHC-I-TCR interaction is essential for impaired erythropoiesis in the
   PRCA patient.

Discussion

   PRCA is a hematopoietic disorder characterized by abnormal development
   of the erythroid lineage. While more than 50 drugs have been reportedly
   associated with PRCA^[140]8,[141]9, the mechanism underlying
   drug-induced PRCA remains unclear^[142]13. Here, we present a case of a
   patient with multiple myeloma who developed PRCA symptoms after RVd
   therapy. Functional in vitro experiments revealed that lenalidomide was
   a key factor contributing to erythropoiesis impairment in this patient.
   Remarkably, there is an abnormal overexpression of MHC-I on PRCA
   erythroblasts, and lenalidomide exacerbated the already elevated
   expression of MHC-I on the surface of erythroblasts, which may further
   promote their clearance by CD8^+ T cells. T cells have been the focus
   of research in autoimmune diseases, while the role of erythroblasts in
   these diseases is often overlooked^[143]28,[144]36,[145]37. In this
   study, we elucidated the mechanism of drug-induced PRCA in a rare case,
   thus accentuating the role of aberrant erythroblasts expressing immune
   markers in the pathogenesis of PRCA.

   Consistent with previous studies in aplastic anemia and
   PRCA^[146]1,[147]2,[148]30,[149]38, we observed a relative enrichment
   of CD8^+ T cells. Since this patient was not diagnosed with thymoma,
   and flow cytometry did not reveal the clonal expansion of any TCR
   subtype, we speculate that the patient’s abnormal transcriptional
   status may suggest genomic abnormalities in the CD8^+ T cells,
   potentially contributing to drug-induced proliferation and
   hyperactivation in the bone marrow. Further genomic analyses and
   functional assays will be needed to address this possibility.
   Cyclosporin A, despite its known T-cell suppressive
   effects^[150]39–[151]41, was ineffective in treating this PRCA
   patient’s relapse, suggesting the involvement of non-T-cell factors in
   this pathological process. Elevated IFN-γ levels, detected through
   transcriptomic and serum analyses, are recognized to hinder erythroid
   progenitor differentiation^[152]42–[153]44. Thus, the increased IFN-γ
   concentration in the patient’s serum likely acted as an obstacle to
   erythropoiesis.

   Interestingly, we uncovered an unexpected upregulation of MHC-I-related
   genes in the patient’s erythroid cells, as revealed by both the
   transcriptome and proteomics analyses. MHC-I molecules are normally
   absent from the surface of mature erythrocytes. During erythroid
   development, MHC-I-related genes are progressively
   downregulated^[154]45,[155]46. Nonetheless, erythrocytes have the
   capacity to express antigen-presenting molecules under specific
   pathological conditions^[156]47. Such a response is noted in
   erythrocytes when infected with the malaria parasite, which can
   activate MHC molecule-mediated CD8^+ T cell recognition. This
   activation leads to the externalization of phosphatidylserine and
   subsequent phagocytosis of the implicated erythrocytes^[157]48.
   Additionally, enhanced MHC-I expression on erythrocytes has been
   observed in patients suffering from systemic lupus erythematosus (SLE),
   with a correlation to the severity of the disease^[158]49. These
   findings imply that erythrocytes may play a supporting role in immune
   regulation, yet the full extent of their impact on immune cell activity
   is still to be determined.

   In our study, we found elevated expression of HLA genes in both
   erythroid progenitors and precursors in the PRCA patient’s bone marrow.
   Even the terminal differentiated CD71^-CD235a^+ erythroid cells from
   the patient, a high level of HLA-B was maintained. Moreover, the
   addition of lenalidomide appeared to promote the sustained high
   expression levels of MHC-I-related genes in the PRCA patient’s cells, a
   phenomenon not observed in cells from healthy donors. The unique
   transcriptional status of MHC-I genes, especially HLA-B, in the
   patient’s erythroid progenitors, along with the differential response
   to drugs, likely play a major role in the disrupted erythropoiesis.
   This is supported by our finding that MHC-I-specific antibodies could
   substantially restore the defective erythroid differentiation of PRCA
   cells. Additionally, adverse reactions to drugs in populations with
   specific MHC-I subtypes have been reported in various
   diseases^[159]50–[160]52. Whether the adverse reactions of lenalidomide
   are associated with this HLA-B subtype still requires exploration with
   more clinical case data.

   Regarding the erythroid-selective cytotoxic effects of hyperactive
   CD8^+ T cells in PRCA, our scRNA-seq analysis revealed a plethora of T
   cell inhibitory genes, such as SERPINB9, LILRB4, and
   LGALS1^[161]53–[162]56, which are highly expressed in dendritic cells
   and monocytes, but not in erythroid cells of PRCA. These observations
   may explain the targeted cytotoxicity of hyperactive CD8^+ T cells
   towards erythroid cells in PRCA.

   In conclusion, we have comprehensively interrogated the mechanism of
   lenalidomide-induced PRCA, focusing on the perspective of erythroid
   cells. While we have identified extrinsic factors, including CD8^+ T
   cells and serum factors like IFN-γ play disruptive roles in
   erythropoiesis, intrinsic defects of erythroid progenitors also
   contribute to PRCA development. Moreover, we unveiled a mechanism
   involving elevated HLA expression in PRCA erythroblasts. Our findings
   suggest that the MHC-I molecule may serve as a crucial clinical marker
   for other PRCA cases. Detecting abnormal MHC-I levels in erythroblasts
   or other cells should prompt caution when considering the use of
   immunomodulatory drugs like lenalidomide to prevent potential adverse
   reactions.

Methods

Human sample collection and ethics statement

   BM and peripheral blood samples from healthy donors (3 males and 1
   female), multiple myeloma patients (3 males and 1 female), and PRCA
   patients (2 males) were collected at Peking University People’s
   Hospital and Peking Union Medical College Hospital. Gender was not
   considered in the study design due to the limitation of the sample
   size. All samples were collected after informed consent had been
   obtained, in accordance with the Declaration of Helsinki, under a
   protocol approved by the Research Ethics Board at Peking University
   Health Science Center and People’s Hospital. Bone marrow cells and
   peripheral blood cells were diluted 1:1 in ice-cold phosphate-buffered
   saline (PBS). Mononuclear cell separation was performed using density
   centrifugation media (Ficoll–Paque; GE Healthcare Life Sciences, Cat. #
   07851) at a 1:1 ratio with diluted blood or marrow cells. After
   centrifugation (1200×g, 10 min), mononuclear cells were carefully
   aspirated to lyse red blood cells.

   Human cord blood cells were obtained from the Cord Blood Bank of
   Beijing. After monocytes had been extracted from cord blood, CD34^+
   cells were isolated via magnetic-activated cell sorting technology
   using a MicroBead kit (Miltenyi, Cat. # 130-100-453).

Ex vivo BMMC erythroid culture

   BMMCs were cultured in Serum-Free Expansion Medium II (SFEM II) based
   erythroid differentiation medium. The medium also contained 3 IU/ml
   erythropoietin (Amgen, Cat. #55513-144-10), 10 ng/ml hIL-3 (StemCell
   Technologies, Cat. #78042), and 50 ng/ml human stem cell factor
   (StemCell Technologies, Cat. #78062). When testing the effects of drugs
   on cells,10 nM bortezomib, 250 nM lenalidomide, or 250 nM dexamethasone
   was added to the medium. Cell differentiation was determined by flow
   cytometry analyses of erythroid markers, including CD71, CD235a, and
   Hoechst 33342.

Colony-forming assays

   For colony-forming assays involving BMMCs, 20,000 cells were plated in
   each well of a 6-well plate using MethoCult H4435 Optimum (StemCell
   Technologies). For colony-forming assays involving human CD34^+ cells,
   200 cells in MethoCult H4435 Optimum (StemCell Technologies) were
   plated in each well of a 6-well plate. In experiments to determine the
   effects of serum and drugs, drugs and 10% serum were added to the
   medium on day 0. Cells were cultured for 14 days; subsequently, BFU-E,
   CFU-GM, and CFU-GEMM colonies were counted and scored.

T cell isolation

   T cells were isolated from BMMCs of the patient or healthy donors. The
   mouse anti-human CD3-BV650 (BD, clone SK7) and mouse anti-human
   CD8a-FITC (BD, clone HIT8a) were used to identified T cells. CD3^+
   cells were isolated using a FACS AriaIII flow cytometer (BD
   Biosciences).

Inhibition of MHC-I interaction in total BMMC cultures

   BMMCs were cultured in SFEM II medium formulated for erythroblast and T
   cell coculture. This medium was supplemented with 3 IU/ml of
   erythropoietin (Amgen, Cat. #55513-144-10), 10 ng/ml human IL-3
   (StemCell Technologies, Cat. #78042), 50 ng/ml human stem cell factor
   (StemCell Technologies, Cat. #78062), 10 ng/ml recombinant IL-2
   (Peprotech, 200-02), and 2 μg/ml anti-CD28 (BD Biosciences, Cat.
   #555725). An anti-human HLA-A, B, C antibody (BioLegend, Cat. #311402)
   at a concentration of 1 μg/ml was introduced to inhibit the interaction
   between the HLA class complex and CD8^+ T cells. Post 16 h of
   incubation, cells were intracellularly labeled with
   fluorophore-conjugated antibodies against CD4, CD8, Granzyme B (GzmB),
   and IFNγ and analyzed with a BD LSR Fortessa flow cytometer. The
   culture medium was refreshed every 3 days. On day 7, cell apoptosis
   rates were quantified by flow cytometry after staining with
   fluorophore-conjugated anti-CD4, anti-CD8, and anti-CD235a antibodies.

scRNA-Seq experiment

   Single-cell capture was achieved by the BD Rhapsody system. Whole
   transcriptome libraries were prepared according to the BD Rhapsody
   single-cell whole-transcriptome amplification workflow and sequenced
   using HiSeq Xten (Illumina, San Diego, California, USA) on a 150 bp
   paired-end run. Raw data were processed using fastQ to filter adaptor
   sequences and remove low-quality reads. The cell barcode whitelist was
   identified by UMI tools. The UMI-based clean data were mapped to the
   human reference genome (Ensemble V.91) using the STAR algorithm. UMI
   count matrices were generated for each sample, and imported into the
   Seurat R toolkit (V. 4.1.3). Cluster analysis was performed by Seurat.

Cell–cell communication analysis using iTALK

   The R package iTALK^[163]57 was used to identify and visualize possible
   cell-cell interactions and to quantify differences between the PRCA
   patient and healthy donors. Three hundred and twenty cytokine/chemokine
   ligand–receptor pairs from the database were analyzed. The Wilcoxon
   rank-sum test was employed to identify differentially expressed genes
   (DEGs) between the PRCA patient and healthy donors for each cell type.
   Subsequently, the ligand–receptor database was paired and matched to
   construct a putative cell–cell communication network. An interaction
   score was defined as the product of the log[2] (fold change) of the
   ligand and receptor.

Unsupervised clustering, marker identification, and cell type annotation

   The Seurat package (with default settings) was used for the
   normalization and scaling of the expression matrix. Mitochondrial
   contamination was regressed by adjusting the “vars.to.regress”
   parameter. To reduce expression matrix dimensionality, principal
   component analysis was performed based on 2000 highly variable genes.
   Unsupervised cell clusters were acquired using a graph-based clustering
   approach (i.e., the top 20 principal components were selected with a
   resolution of 0.75), then visualized by uniform manifold approximation
   and projection (UMAP) or t-Distributed Stochastic Neighbor Embedding
   (t-SNE) dimensionality reduction. Clusters were annotated to known
   biological cell types according to the expression patterns of canonical
   markers. Marker genes in each cluster were identified using the
   FindAllMarkers function with the following criteria: log (fold
   change) > 0.25, min. pct > 0.25, and adjusted P-value < 0.05. Pathway
   analysis was used to identify significant pathways of DEGs according to
   the Kyoto Encyclopedia of Genes and Genomes (KEGG) and GSEA database.

   To enable a more accurate comparison of transcriptional differences
   between HD and PRCA samples, we used Seurat’s ‘downsample’ function to
   randomly reduce the HD sample to 300 HSC cells, 200 erythroid
   progenitor cells, and 60 erythroid precursor cells, matching the cell
   numbers in the PRCA samples (Supplemental Fig. [164]5B, C). The
   differential gene analysis (Supplemental Fig. [165]5B, C) informed the
   GSEA pathway enrichment analysis in Fig. [166]3D.

Bulk RNA-seq

   After treatment with indicated drugs in an erythroid differentiation
   medium for 6 days. Cells were spun down and resuspended in the staining
   buffer (2% FBS in PBS); then the 5 μl anti-CD235a (APC, clone HIR2,
   eBioscience, Cat.# 17-9987-42), 5 μl anti-CD71 (FITC, clone OKT9,
   eBioscience, Cat.#11-0719-42) were added as per 1,000,000 cells. FACS
   sorting was performed on BD FACSAria III. The CD71^+ erythroid
   progenitor cells were sorted by fluorescence-activated cell sorting
   (FACS). The cell populations of CD71^+ differentiation stages were
   isolated. RNA was extracted with TRIzol and cleaned with an RNeasy
   MinElute Cleanup Kit.

   RNA was extracted using the RNA isolation Total RNA Extraction Reagent
   (Vazyme, Cat.# R401-01). Sequencing libraries were prepared by
   Novogene, and 150 bp paired-end sequencing of each condition with two
   biological replicates was obtained in a HiSeq X-Ten (PE150, Illumina)
   by Novogene (Beijing). FASTQ files were generated by Novogene. Hisat2
   (version 2.1.0) was used to align the paired-end raw data human
   reference genome, HTSeq was used to calculate the reads counts, and
   DESeq2 was used to identify the differential expression genes.

Proteomics analysis

   As described above, CD71^+ erythroid cells were sorted by FACS after
   six days of in vitro culture. The sample was sonicated three times on
   ice using a high-intensity ultrasonic processor (Scientz) in lysis
   buffer (8 M urea, 1% protease inhibitor cocktail). The remaining debris
   was removed by centrifugation. The protein concentration was determined
   using a bicinchoninic acid (BCA) protein assay kit. For digestion, the
   protein solution was reduced with 5 mM dithiothreitol for 30 min at
   56 °C and alkylated with 11 mM iodoacetamide for 15 min at room
   temperature in darkness. The protein sample was then diluted by adding
   200 mM TEAB to reduce the urea concentration to less than 2 M. Trypsin
   was added at a 1:50 trypsin-to-protein mass ratio for the first
   digestion overnight, followed by a second 4 h-digestion. Finally, the
   peptides were desalted by the Strata X SPE column. LC–MS/MS Analysis
   was conducted at Jingjie PTM Biolab (Hangzhou).

Statistics

   Statistical analyses were performed using the Prime 10 software
   (Graphpad, La Jolla, CA). We considered P values of less than 0.05 to
   be statistically significant. Data are expressed as the mean ± standard
   deviation. A two-tailed unpaired/paired Student’s t-test was performed
   for the comparison between two groups (ns, P > 0.05; *, P ≤ 0.05; **,
   P ≤ 0.01; ***, P ≤ 0.001).

Reporting summary

   Further information on research design is available in the [167]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [168]Supplementary Information^ (2.5MB, pdf)
   [169]41467_2024_54571_MOESM2_ESM.pdf^ (78.7KB, pdf)

   Description of Additional Supplementary Files
   [170]Supplementary Data 1^ (4MB, xlsx)
   [171]Reporting Summary^ (1.6MB, pdf)
   [172]Transparent Peer Review file^ (4.1MB, pdf)

Source data

   [173]Source Data^ (55.5KB, xls)

Acknowledgements