Publisher Correction: Single-cell and bulk transcriptomics uncovers PRKD2-driven tumor stemness and progression in multiple myeloma

Correction to: Scientific Reports https://doi.org/10.1038/s41598-025-20615-4, published online 21 October 2025

The original version of this Article contained an error in the order of the Figures. Figure 1 was published as Figure 3, Figure 2 was published as Figure 1, Figure 3 was published as Figure 4, Figure 4 was published as Figure 6, Figure 5 was published as Figure 2, and Figure 6 was published as Figure 5.

The original Figures 1, 2, 3, 4, 5, 6 and accompanying legends appear below.

The original Article has been corrected.

Fig. 1

Single-cell atlas of myeloma progression pinpoints PRKD2 as a stage-linked hub. (A) UMAP coloured by 20 clusters; split panels show stages. Dashed circle highlights the plasma-cell region where an “MM-malignant plasma” cluster emerges in MGUS and expands through SMM to dominate MM. HD (n = 5), MGUS (n = 6), SMM (n = 4) and MM (n = 4). (B) Bar chart of mean cell-type frequencies per stage. (*P < 0.05, compared with HD) (C) Heat-map of normal plasma-cell DEGs; PRKD2 (red) rises stepwise HD→MM. Heat-map was drawn in R [v4.3.3] using ComplexHeatmap [v2.1] (Bioconductor; https://bioconductor.org/packages/ComplexHeatmap) from vst-normalized expression with per-gene z-scores. (D) Heat-map for MM-malignant plasma cells, showing further PRKD2 up-regulation plus stress/immune-evasion genes. Generated in R [v4.3.3] with ComplexHeatmap [v2.1] as in (C). (E) Box-scatter plots of mRNAsi and EREG_mRNAsi in bulk plasma cells (****P < 0.0001, Wilcoxon). (F) Same indices within the MM-malignant subset. (G) Feature-density UMAPs: PRKD2 signal (yellow) intensifies from HD to MM within plasma cells. (H–K) PRKD2 expression correlates with mRNAsi and EREG_mRNAsi; Points are coloured by clinical group (HD, MGUS, SMM, MM). Separate solid lines represent group-specific linear regressions with 95% confidence ribbons. An ANCOVA comparing the four slopes yields a significant group × PRKD2 interaction (F = 4.89, p = 0.003), indicating that the relationship between PRKD2 expression and stemness indices differs across disease stages. (L) iTALK network linking CD4⁺, CD8⁺, NK/NKT, normal and malignant plasma cells in MM; arrow width ∝ ligand–receptor counts. (M) Chord diagram of dominant MM-plasma/NK→CD4⁺ T signals; VCAN, HSP90B1, RNASE2 and HRAS target TLR2. PRKD2 expression in sending plasma cells scales with total TLR2 engagement.

Fig. 2

Pseudotime analysis uncovers a PRKD2-centred malignant branch. (A,B) Monocle2 principal graph of 2137 plasma cells (red = normal, cyan = MM-malignant). Branch nodes ① and ② split a quiescent upper arm from a malignant lower arm. (C) Same graph by clinical stage. Cells shift from root (HD) through MGUS and SMM to populate the malignant arm in MM; grey curves show stage-specific density along pseudotime. (D) Gene-smooths along pseudotime. Early regulators (BTG2, FOS/FOSB, CD79A) wane, ribosome/secretory genes (IGHG, IGHD/M, RPL29, RPS5/A) rise. PRKD2 climbs steadily, peaking where malignant cells accumulate. (E) BEAM heat-map of 278 branch genes. Module 1—up-regulated in malignant fate—contains PRKD2 with ribosome/ER genes (RPS5, RPL29); modules 2–4 house stress, Ig heavy- and light-chain signatures, respectively. BEAM analysis was performed with Monocle 2 [v2.3] (http://cole-trapnell-lab.github.io/monocle-release/) and the heat-map rendered in R [v4.3.3] using ComplexHeatmap [v2.1] (https://bioconductor.org/packages/ComplexHeatmap).

Fig. 3

Stemness indices, clinical stratification, survival impact. mRNAsi and EREG-mRNAsi across patient stratifications. Box-plots show log-scaled indices after propensity-score matching (PSM). Older patients (≥ 65 year) harbour higher stemness than younger cases. (A) Scores further increase from complete/partial remission (CR/PR) through newly diagnosed (NDMM) to relapsed/refractory (RRMM) myeloma, with lowest values in healthy donors (HD). (B) A step-wise rise is also evident across ISS stages I→III, whereas MGUS/SMM (“Other”) remain low. (C) Kruskal–Wallis tests: **P < 0.01; ***P < 0.001. (D,E) Kaplan–Meier curves dichotomised at the cohort median. High-stemness myeloma exhibits inferior 5-year overall survival (mRNAsi: 38% vs. 73%, log-rank P = 0.012; EREG-mRNAsi: 36% vs. 70%, P = 0.063).

Fig. 4

PRKD2 is a shared DEG and ER-stress effector in multiple myeloma. (A) Venn diagram of DESeq2 analyses (NDMM vs. HD, RRMM vs. HD, GSE24080 dead vs. alive) showing 28 common DEGs centred on PRKD2. (B) Heat-map of the 28 genes across control, CR/PR, NDMM and RRMM marrows; PRKD2 is up-regulated in every MM group. Bottom annotation bars carry numbered colour codes to aid cross-reference: 1 Group, 2 Age, 3 Gender, 4 MM type, 5 Durie–Salmon stage, 6 ISS stage. PRKD2 heads the list and shows strongest up-regulation in the RRMM samples. Heat-map generated in R [v4.3.3] with ComplexHeatmap [v2.1] (https://bioconductor.org/packages/ComplexHeatmap) from vst-normalized expression with per-gene z-scores. (C,D) Western blots validating PRKD2 knock-down (sh1641/2074/1965) and knock-up (cDNA-Homo) in H929, 8226, U266 cells; β-actin, loading control. (E) CCK-8 proliferation assay in 8226 cells: PRKD2 knock-down suppresses growth (0–72 h), over-expression has no consistent effect (mean ± SD, n = 9). (F) GO enrichment of ID DEGs: top terms cluster around unfolded-protein response, ER stress and chaperone activity. (G) KEGG enrichment of the same DEGs highlights ‘Protein processing in endoplasmic reticulum’ and ‘Protein export’, linking PRKD2 to secretory-stress and pro-survival pathways (RRMM vs. CR). KEGG resource cited per guidelines: KEGG—Kyoto Encyclopedia of Genes and Genomes (https://www.kegg.jp/kegg/kegg1.html).

Fig. 5

PRKD2 expression dictates immune-pathway activity and leukocyte composition. (A–D) ssGSEA enrichment plots. In PRKD2-low marrows, B-cell receptor signalling, antigen presentation and other immune pathways rank among the top positively enriched sets; the same pathways are depleted in PRKD2-high samples. (E) CIBERSORT-derived leukocyte fractions for each sample, stacked by cell type. PRKD2-high cases (red) show a relative increase in ​memory B-cell compartments and a reduction in CD4 memory T cells, monocytes and mast cells compared with PRKD2-low cases (blue). (F) Box-scatter comparison of the most significantly altered immune subsets between PRKD2 groups (Wilcoxon; P < 0.05 marked by asterisks). (G) Correlation matrix (colour) and network (edges > |0.2|) linking PRKD family members to 22 immune-cell subsets. PRKD2 is uniquely and negatively correlated with mature plasma cells (blue edge), whereas PRKD1/3 show weaker, non-specific associations.

Fig. 6

High PRKD2 suppresses antigen-presentation pathways yet confers axitinib sensitivity. (A) ssGSEA comparing PRKD2-high vs. -low marrows: APC signalling, HLA machinery, CCR axis and type-II IFN response are all down-regulated in the high group. (B,C) HLA gene expression under PRKD2 manipulation: over-expression has little impact, whereas PRKD2 knock-down markedly elevates HLA-E. (D) RNA-seq of PRKD2-knockout myeloma cells. Loss of PRKD2 in RPMI-8226 cells lowers CD80 and FCGR1A while raising VEGFA (log₂ FC ± SD, n = 3 replicates). (E) RNA-seq of PRKD2-over-expressing myeloma cells. Conversely, PRKD2 over-expression increases IL-6 and MERTK in the same cell lines (log₂ FC ± SD, n = 3). (F) PD-L1 mRNA shows no difference between PRKD2 strata. (G) GDSC mining links rising PRKD2 to greater in-vitro sensitivity to the VEGFR/PDGFR inhibitor axitinib (Pearson r = 0.68). (H) Boxplot confirms higher axitinib sensitivity (lower IC₅₀ index) in the PRKD2-high cohort. (I) Representative Annexin V/7-AAD density plots of vector control (LV3-vector, LV5-vector), PRKD2 knock-down (1965-PRKD2-KD) and PRKD2 over-expression (HOMO-PRKD2-OE) cells after 48 h exposure to DMSO or axitinib (1.25, 2.5, 5, 10 µM). Quadrants: Q1, necrotic; Q2, late apoptotic; Q3, early apoptotic; Q4, viable. (J) Quantification of total apoptosis (Q2 + Q3) from three independent experiments (mean ± SD). PRKD2-OE cells show significantly higher apoptosis than the LV5 control at every dose, whereas PRKD2-KD cells are intrinsically pro-apoptotic and further sensitised by axitinib. (K) Relative cell proliferation measured by CCK-8 after the same 48 h treatments. PRKD2 loss reduces basal growth and slightly lowers viability at 10 µM, while axitinib abolishes the proliferative advantage of PRKD2-OE cells. Statistical analysis: two-way ANOVA with Tukey’s multiple-comparison test; P < 0.05 versus corresponding vector control at the same concentration; *P < 0.05 versus DMSO within the same cell line. Error bars denote SD (n = 3).