Patient baseline characteristics

Between January 2020 and January 2022, a total of 18 patients (11 (61.1%) female and 7 (38.9%) male) with a median age of 74 years (range 49–92 years) were enrolled and assigned to neoadjuvant treatment with T-VEC (Fig. 1a,b). Inclusion required the presence of a difficult-to-resect BCC, defined as surgical necessity for either a skin flap or skin transplant for wound closure. All 18 patients were included for the final efficacy analysis (Fig. 1b). Detailed patient baseline and tumor characteristics are listed in Table 1 and Supplementary Table 1. T-VEC treatment was prematurely discontinued at one patient’s request, because of a grade 2 AE of fever, who underwent surgery with a local skin flap, after receiving only two cycles of T-VEC. One patient was lost to follow-up after receiving six cycles of neoadjuvant T-VEC treatment, but before undergoing surgery. As this patient did not achieve a clinical response, allowing excision with primary wound closure, he was included for efficacy, but excluded for the recurrence analysis. The data cutoff was 31 December 2022. The median follow-up period, between the first administration of T-VEC and the data cutoff, was 24 months (range 12–36 months).

Fig. 1: Clinical activity of neoadjuvant treatment with T-VEC in cutaneous basal cell carcinoma.

a, Illustrated study scheme and treatment schedule. Biomarker analysis was performed with mIF pre- and post-treatment, scRNA-seq and scTCR-seq and scBCR-seq post-treatment. Image was created in BioRender. b, Flowchart of the NeoBCC trial. In total, 18 patients started therapy and were included for the final analysis of the study. One patient requested premature termination of the study treatment due to a grade 2 AE. c, Patient response after six cycles of T-VEC administration was split into surgical outcome (with or without (w/o) skin flap/graft). d, Clinical response defined according to the WHO response criteria (CR, PR and SD). e, Pathological response classified in pCR and non-pCR. f, Exemplary clinical and dermoscopic images of two BCCs, pre- and post-treatment. One patient had a pCR with tumor regression and fibrosis post-treatment (top right) and one patient with a locally advanced BCC had a non-pCR with a tumor area reduction after T-VEC treatment (bottom right). g, Waterfall plots presenting the overall tumor area reduction (%) from baseline before first T-VEC injection until the time point of surgery in patients positive and negative for the primary end point (PE), including information about the pathological response (pCR, dark green; non-pCR, light green) on an individual patient level. Patient 20 had no change after two cycles of T-VEC and dropped out of the study due to an AE. The dashed line represents the 50% tumor area reduction threshold. h, A spider plot exemplifying the tumor area reduction (%) of each patient according to treatment time points.

Source data

Table 1 Patient baseline and tumor characteristicsPrimary end point

The primary end point of the study, defined as the proportion of patients, who after six cycles of T-VEC (13 weeks) at the time point of surgery, become resectable with primary wound closure without the need for plastic reconstructive surgery like skin flaps or skin grafts, was achieved. In the first stage of the study, 9 of 18 (50%) patients already had a positive response in the primary end point (Fig. 1c); however, only eight responses of the predefined 43 patients were required in this Simon’s two-stage design. Therefore, the null hypothesis with a rate of ≤10% could already be rejected (P < 0.0001) (lower and upper bound 29.1% and 70.9%; two-sided 90% Clopper–Pearson confidence interval) and the study was discontinued for early success.

Of the other nine patients who did not have a positive response in the primary end point, five of nine (55.6%) required a local skin flap, one of nine (11.1%) a skin transplant, one of nine (11.1%) a local skin flap in combination with a skin graft and two of nine (22.2%) patients, who would have required skin flaps/grafts, requested secondary wound healing post-excision.

Secondary end points

Six of 18 (33.3%) patients had a complete response (CR), 4 of 18 (22.2%) a partial response (PR) and 8 of 18 (44.4%) a stable disease (SD), resulting in an ORR of 55.6% (Fig. 1d) according to the World Health Organization (WHO) criteria, to neoadjuvant T-VEC treatment. Of note, none of the patients had an increase in tumor size during the treatment period. Six of 18 (33.3%) patients had a pathological complete response (pCR) and 12 of 18 (66.7%) had a pathological noncomplete response (non-pCR) (Fig. 1e,f). Five of nine (55.6%) patients with a histopathological non-infiltrative BCC subtype achieved a pCR, whereas one of nine (11.1%) patients with an infiltrative BCC had a pCR. In one patient, the time to treatment failure was 35 days. The median and mean tumor area reduction from screening until surgery (n = 18) were 42.7% (range 0–83.3%) and 45.4% (s.d. ± 24.5%), respectively. Patients who achieved the primary end point of resection with primary wound closure (n = 9) had a median tumor area reduction of 62.5% (range 25.0–83.3%), whereas patients negative for the primary end point (n = 9) had a median tumor area reduction of 37.7% (range 0–70.9%; Fig. 1g). All patients receiving six cycles of T-VEC showed a nominal reduction in tumor size (Fig. 1h). Of note, in some cases it was difficult to clinically differentiate between fibrosis and viable tumor cells. Therefore, a residual tumor area was calculated in all cases, although in six post-excision samples a pCR was reported.

At a median follow-up of 11 months (range 2.1–16.3 months), from the time point of resection of any residual tumor, no patient had experienced a relapse. The 6-month relapse-free survival rate was 100%. With a median follow-up of 14 months (range 4.3–19.7 months) from the time point of the first T-VEC injection no patient death occurred. The 6-month overall survival rate was 100%. During the median follow-up of 11 months (range 2.1–16.3 months), calculated from the time point of resection of any residual tumor, 2 of 18 (11.1%) patients developed new BCCs on different body sites with a median time to appearance of 7 months (range 1.2–12.1 months).

Safety

All patients who received at least one dose of T-VEC (n = 18) were included in the safety analysis. AEs were documented according to the Common Terminology Criteria for Adverse Events (CTCAE, v.5.0). Fourteen of 18 (83.3%) patients developed T-VEC-related grade 1 or 2 AEs. The most frequent T-VEC-related AEs were local injection site reaction (21.9%), including erythema, swelling and procedural pain. All T-VEC and non-T-VEC-related AEs are listed in Table 2. No serious AEs (SAEs) occurred. At one patient´s request, T-VEC therapy was discontinued after two cycles because of grade 2 pyrexia. This patient was serum IgM and IgG negative for HSV at baseline and converted to be HSV IgM positive 4 weeks after treatment initiation (Table 1). One wound infection and one hematoma were observed following surgery, both unrelated to T-VEC (Table 2). The recruitment and treatment phase of the study commenced in parallel with the COVID-19 pandemic and no safety data were available on the concomitant use of T-VEC and RNA-based COVID-19 vaccines. In our patient cohort, eight patients received an RNA-based COVID-19 vaccination concurrently with T-VEC; neither vaccine-related AEs, nor unreported T-VEC-related AEs occurred.

Table 2 Adverse events of neoadjuvant therapy with talimogene laherparepvec according to the Common Terminology Criteria of Adverse Events (v.5.0)T-VEC reprograms the tumor immune microenvironment

To evaluate the treatment-induced dynamic changes within the immune cell landscape of BCCs, we performed spatial profiling with multiplex immunofluorescence (mIF) staining before and after treatment from available paired tissue tumor samples (n = 16 from n = 18 patients). Immune cells in treatment-naive BCCs were found at significantly higher levels in the peritumoral stroma, whereas only isolated immune cells were detected within the tumor islands (CD4+ T cells, P = 0.0001; CD4+Foxp3+ Treg cells, P = 0.0013; CD8+ T cells, P = 0.0001; CD68+ myeloid cells, P = 0.0430; CD20+ B cells, P = 0.0139, two-sided Wilcoxon signed-rank test; statistical significance determined with P < 0.05) (Fig. 2a,b) (Extended Data Fig. 1a–c). After neoadjuvant T-VEC treatment, a significant increase of CD8+ T cells (P = 0.0092), CD20+ B cells (P = 0.0004) and CD68+ myeloid cells (P = 0.0042) and a decrease of CD4+ T cells (P = 0.0042) and CD4+Foxp3+ Treg cells (P = 0.0290) was observed. Notably, also the CD8+ T cell/Treg ratio increased significantly (P = 0.0039) (Fig. 2c). In addition, we discovered a significant increase of CD8+ T cells (P = 0.002), CD20+ B cells (P = 0.002) and CD68+ myeloid cells (P = 0.002) infiltrating the interface of the tumor islands (Extended Data Fig. 2a–c) and a significant reduction of tumor cell density compared to baseline samples (P = 0.0488) (Fig. 2d) in patients with a remaining tumor post-treatment.

Fig. 2: T-VEC reprograms the tumor immune microenvironment.

a, Representative images of a baseline sample with merged seven-plex immunofluorescence staining (PanCK (rose), CD4 (green), CD8 (red), Foxp3 (magenta), CD20 (blue) and CD68 (yellow) markers were targeted in addition to nuclear counterstaining with 4,6-diamidino-2-phenylindole (DAPI)) image (left) and tissue segmentation in tumor, tumor stroma and epidermis are displayed (right). Scale bar 100 µm. b, The immune cell infiltrate of patients with treatment-naive BCCs (n = 16) is presented separately in tumor and tumor stroma by box plots (√cells per mm2) (CD4+, P = 0.0041; CD4+Foxp3+, P = 0.0290; CD8+, P = 0.0092; CD68+, P = 0.0042; CD20+, P = 0.0004) (two-sided Wilcoxon signed-rank test, statistical significance determined with P < 0.05). Data are presented as mean ± s.e.m. Each box extends from the 25th (Q1, lower bound of box) to the 75th (Q3, upper bound of box) percentile, the horizontal line in the center of the box represents the median value (Q2), the lower bound of lower whiskers mark the 5th (min) and the upper bound of whiskers the 95th (max) percentiles, dots represent individual values. c, Distribution of immune cells (relative values, %) on an individual patient level pre- and post-treatment, in accordance with the surgical outcome (with or without (w/o) skin flap/graft) and pathological response (pCR and non-pCR) are displayed in bar plots. Immune cells (%) pre- versus post-treatment are displayed with bar plots (two-sided Wilcoxon signed-rank test, statistical significance determined with P < 0.05). d, Representative post-treatment histopathological picture (seven-plex immunofluorescence panel), 500 µm and 100 µm bar, from one patient (n = 1). Tumor cell density is presented in box plots pre- and post-treatment from patients with a non-pCR (n = 10). Data are presented as mean ± s.e.m. (two-sided Wilcoxon signed-rank test, statistical significance determined with P < 0.05). Each box extends from the 25th (Q1, lower bound of box) to the 75th (Q3, upper bound of box) percentile, the horizontal line in the center of the box represents the median value (Q2), the lower bound of lower whiskers mark the 5th (min) and the upper bound of whiskers the 95th (max) percentiles; dots represent individual values.

Source data

Next, we investigated the relationship between immune cell infiltration and the clinical (CR, PR and SD) and pathological response (pCR and non-pCR). According to the clinical response we discovered that patients with a tumor area reduction of ≥50%, meaning a CR or PR, had significantly increased levels of CD8+ T cells (P = 0.0488), CD20+ B cells (P = 0.0020) and CD68+ myeloid cells (P = 0.0020) and decreased CD4+ T cells (P = 0.0059) and CD4+Foxp3+ Treg cells (P = 0.0059). We observed similar trends that did not reach the level of significance in patients with SD, which had a tumor area reduction of ≤50%, with increased numbers of CD8+ T cells (P = 0.1562), CD20+ B cells (P = 0.2188) and CD68+ myeloid cells (P = 0.4375), and a decrease of CD4+ T cells (P = 0.3125) and CD4+Foxp3+ Treg cells (P = 0.9999) (Extended Data Fig. 3a). Comparing pCR to non-pCR samples, we detected similar trends with heterogeneity in significance levels. Patients with a pCR had significantly increased levels of CD20+ B cells (P = 0.0312) and myeloid cells (P = 0.0312) only, and a nonsignificant increase of CD8+ T cells (P = 0.4375) and decrease of CD4+ T cells (0.0938) and Treg cells (P = 0.0625). In contrast, patients with a non-pCR showed significantly increased CD8+ T cells (P = 0.0195), CD20+ B cells (P = 0.0137), nonsignificantly raised myeloid cells (P = 0.0645), a significant decrease of CD4+ T cells (P = 0.0273) and a nonsignificant decrease of CD4+Foxp3+ Treg cells (P = 0.3223) (Extended Data Fig. 3b).

Immune cell heterogeneity according to treatment response

We performed single-cell RNA sequencing (scRNA-seq) (n = 102,093 cells) on available post-treatment tissue samples at the time point of surgery, from a subset of patients (n = 12). In total, 18 cell clusters were identified, including malignant cells, keratinocytes, fibroblasts, endothelial cells, melanocytes, three T cell subsets (CD4+ T cells, Treg cells and CD8+ T cells), NK cells, proliferating cells, B and plasma cells, plasmacytoid dendritic cells (pDC), dendritic cells, Langerhans cells (LCs), mast cells and monocytes and macrophages (Mono-Mac) (Fig. 3a,b and Extended Data Fig. 4a). T cells, B cells and macrophages accounted for the highest percentage of the immune cell population in the scRNA-seq data (Fig. 3a) validated by a mIF staining panel (Fig. 2c).

Fig. 3: Immune cell heterogeneity after T-VEC treatment.

a, UMAP of tumor cells and their microenvironment in post-treatment BCC samples (n = 102,093 cells) from a subset of patients (n = 12) by scRNA-seq analysis. A total of 18 cell clusters were identified and denoted by colors such as malignant cells, keratinocytes, fibroblasts, endothelial cells, melanocytes, three T cell subsets, NK cells, proliferating cells, B and plasma cells, pDCs, dendritic cells, LCs, mast cells and Mono-Mac, as well as a nondefined ‘other’ cluster. b, Bubble plot indicating the average (Avg.) expression of selected marker genes of the cell types in a.

Source data

In the subsequent analysis we focused on the pathological response, as pCR is considered a clinically beneficial biomarker in neoadjuvant clinical trials for other tumor types21,22. Patients with a non-pCR had significantly higher numbers of Treg cells (marked by Forkhead box protein 3 (FOXP3), interleukin-2 receptor subunit α (IL2RA) and tumor necrosis factor receptor superfamily member 4 (TNFRSF4) (P = 0.028; two-sided Wilcoxon rank-sum test; statistical significance determined with P < 0.05). Proliferating cells (MKI67)) (P = 0.0480) expressed high Mobility Group Box 2 (HMGB2) (Extended Data Fig. 4a), a member of the non-histone chromosomal high mobility group protein family, which has been associated with poor prognosis and invasion in lung cancer23, were significantly elevated in patients with a non-pCR (Extended Data Fig. 4b). Dimensionality reduction after cell-cycle regression analysis further showed that these proliferating cells were primarily T cells (82%; n = 621 T cells out of 754 cells) and were predominantly found in patients with a non-pCR (n = 700 cells from non-pCR versus 54 cells from pCR) and were largely enriched in CD4+ T cells and Treg cells (Extended Data Fig. 4c,d). Direct comparison of these two complementary methods (mIF and scRNA-seq) according to the pathological and clinical response showed consistent trends (Extended Data Fig. 5a–d).

T-VEC favors hyper-expanded, cytotoxic T cell clones

T cells constituted a large proportion of the immune cell population in the TME of treatment-naive and T-VEC-treated BCCs (Figs. 2b,c and 3a). In-depth analysis revealed a total of eight T cell clusters (naive CD4+ T cells, memory T cells, Treg cells, T helper cells, cytotoxic T cells, γδ T cells, proliferating cells and exhausted T cells) and one NK cell cluster (Fig. 4a) post-treatment. Patients with a non-pCR had a significantly higher number of Treg cells (FOXP3, TNFRSF4 and tumor necrosis factor receptor superfamily member 18 (TNFRSF18)) (P = 0.0081) (two-sided Wilcoxon rank-sum test, statistical significance determined with P < 0.05) with upregulation of checkpoint molecules (cytotoxic T-lymphocyte-associated protein 4 (CTLA4), T cell immune receptor with Ig and ITIM domains (TIGIT), inducible T cell co-stimulatory (ICOS)) (Fig. 4b,c and Extended Data Fig. 6a). In contrast, we detected a significantly higher cytotoxic T cell/Treg ratio (P = 0.016) in patients with a pCR (Fig. 4d) corroborating our data from mIF (Fig. 2c). Similar trends were observed when comparing CR with SD (P = 0.029; Fig. 4d).

Fig. 4: T-VEC favors the expansion of hyper-expanded, cytotoxic T cell clones.

a, UMAP presenting post-treatment T cells (eight clusters) and NK cells (one cluster) b, Bubble plot indicating the average (Avg.) expression of selected marker genes in each cell cluster, including a subclassification of T helper cells (13 clusters). c, Bar plots presenting the composition of T cells. The colors reflect the ones of the immune cell clusters of a, split by pathological (pCR versus non-pCR) and clinical (CR, PR and SD) response (two-sided Wilcoxon rank-sum test; statistical significance determined with P < 0.05). d, Violin plot demonstrating the ratio of cytotoxic T cells and Treg cells. Cytotoxic T cell/Treg ratio of patients according to pathological response (pCR (n = 4) and non-pCR (n = 8)) and clinical response (CR (n = 4), PR (n = 4) and SD (n = 4)) (two-sided Wilcoxon rank-sum test; statistical significance determined with P < 0.05). e, Bar plots presenting terms from the MSigDB Hallmarks database that were enriched in hyper-expanded T cell clones. Terms are ranked by the total number of genes overlapping differentially expressed genes and adjusted P values are indicated (enricher one-sided hypergeometric test with false discovery rate correction; Benjamini–Hochberg)77. The universe is defined by all detected genes in the T cell subset. f, Bar plots displaying the number of T cell clones aligned to the MacPAS-TCR database according to human HSV- and cancer-associated TCRs.

Source data

Next, we paired our scRNA-seq data with single-cell T cell receptor sequencing (scTCR-seq) assigning a T cell receptor (TCR) clonotype with exactly one α and one β CDR3 chain to 23,099 of 43,661 T cells (53%). We detected small, medium, large and hyper-expanded T cell clones (Extended Data Fig. 6b) but did not observe significant differences between patients with pCR and non-pCR, regarding the absolute number of TCR clones and repertoire clonality (Extended Data Fig. 6c). Cytotoxic T cells had the highest clonal density and made up the majority of hyper-expanded T cell clones (Extended Data Fig. 6d). Large T cell clones showed characteristics of proliferative, dysfunctional CD8+ T cells24,25, whereas small T cell clones expressed naive T cell markers C–C motif chemokine receptor 7 (CCR7) and lymphoid enhancer binding factor 1 (LEF1)26 (Extended Data Fig. 6e), which were found to a higher extent in patients with a non-pCR. Notably, enrichment analysis of marker genes of hyper-expanded T cell clones demonstrated high inflammatory activity and IFN-γ response, suggesting a fully functional activation (Fig. 4e). We further aligned our scTCR-seq dataset with a curated catalog of TCR sequences, McPAS-TCR27. The 435 α-chains of the BCC TCR repertoire were listed in this McPAS-TCR database in 10 of 12 patient samples. We were able to detect human HSV-associated TCRs in two of ten samples and cancer-associated TCRs in ten of ten samples (Fig. 4f).

Macrophage diversity in the TME of T-VEC-treated BCCs

We detected a significant increase of CD68+ macrophages upon T-VEC treatment (P = 0.0430) (two-sided Wilcoxon signed-rank test; statistical significance determined with P < 0.05) (Fig. 2c) and scRNA-seq analysis further unveiled distinct myeloid cell populations post-treatment, such as ficolin 1 (FCN1+), complement C1q C chain (C1QC+) and secreted phosphoprotein 1 (SPP1+)-positive macrophages (Extended Data Fig. 7a,b). The majority were FCN1+ and C1QC+ (74%, n = 10,898 FCN1 and C1QC macrophages out of 14,738 total macrophages) and their distribution was comparable between patients with a pCR and non-pCR, and patients with CR, PR and SD (Extended Data Fig. 7c).

Functional enrichment analyses of differentially upregulated genes highlighted the migratory and inflammatory phenotype of FCN1+ macrophages with several processes related to chemotaxis and defense processes enriched (Clusterprofiler:enricher with false discovery rate correction28 adjusted P < 0.005; Extended Data Fig. 7d–f). Further characterization of the FCN1+ subset using recently published tumor-associated macrophages (TAM) gene signatures29 revealed their contribution to immunoregulatory TAMs (Reg_TAM1) and classical tumor-infiltrating monocytes (Extended Data Fig. 7g), in line with what has been previously described in colorectal cancer for this subset29. C1QC+ macrophages displayed enrichment in genes associated with IFN-γ response, antigen presentation and humoral immune response (Extended Data Fig. 7d–f), and showed similarity to a resident tissue macrophage-like phenotype (Extended Data Fig. 7g)30. At the same time, genes upregulated in SPP1+ macrophages were enriched for enhanced lipid metabolism and hypoxia-inducible genes (Extended Data Fig. 7d–f) and showed similarity to inflammatory TAMs (Inf-TAMs) with additional pro-angiogenic (Angio) signatures (Extended Data Fig. 7g)29. In addition, macrophages expressed Fc γ receptor 3A (FCGR3A), which is associated with antibody-dependent cellular phagocytosis and elimination of cancer cells31 (Extended Data Fig. 7h).

Spatial proximity analysis with mIF, revealed a trend toward an increased colocalization of CD68+ myeloid cells with CD4+ T cells upon T-VEC treatment (P = 0.1754) (CD68+ myeloid cells within 10 μm of CD4+ T cells; mean 0.2059 (s.d. ± 0.1979) pre-treatment versus 0.2976 (s.d. ± 0.1459) post-treatment; two-sided Wilcoxon signed-rank test, statistical significance determined with P < 0.05) and a significantly increased colocalization with CD20+ B cells (P = 0.0052) (CD68+ myeloid cells within 10 μm of CD20+ B cells; mean 0.0083 (s.d. ± 0.1377) pre-treatment versus 0.2277 (s.d. ± 0.1283) post-treatment; Extended Data Fig. 7i).

T-VEC induces hyper-expanded IGHG1 plasma cells

Investigation of B cell subsets after treatment uncovered inter- and intrapatient heterogeneity and revealed a total of nine B and plasma cell clusters (Fig. 5a,b); however, no significant differences between the cluster gene expression and pathological and clinical response were observed (Fig. 5c and Extended Data Fig. 8a). We identified a subpopulation of B cells expressing high levels of major histocompatibility complex, class II, DR α (HLA-DRA) in clusters 34, 7, 21 and 38 (Extended Data Fig. 8b), which suggests those tumor-infiltrating B cells were capable of antigen presentation32. Furthermore, we detected cytotoxic and antiviral features (granzyme K (GZMK) and perforin 1 (PRF1)) in B cells in cluster 38 (ref. 33) and memory features (CD27) in B and plasma cells (cluster 24, 4, 3, 22 and 38)34 (Extended Data Fig. 8b,c). Plasma cells (cluster 15, 22, 3, 4 and 24) showed a high expression of the transcription factor X-box binding protein 1 (XBP1) (Extended Data Fig. 8b), which is necessary for terminal differentiation of B lymphocytes to plasma cells35. We did not detect immunosuppressive phenotypes (interleukin-10 (IL10) and transforming growth factor β1 (TGFB1); Extended Data Fig. 8c).

Fig. 5: T-VEC fosters a humoral immune response.

a, UMAP of plasma and B cells after treatment (nine clusters). b, Bubble plot indicating the average (Avg.) expression levels of selected marker genes in each cell cluster. c, Bar plots presenting the composition of B and plasma cells. The colors reflect the ones of the immune cell clusters of a, split by pathological response classified in pCR and non-pCR and clinical response categorized in CR, PR and SD. d, Bar plots showing immunoglobulin (Ig) types in B and plasma cells according to their clonality (left) and cell type (right). e, Exemplary merged mIF staining (IgG1+, yellow; CD138+, green; PanCK+, pink) of a pre-treatment BCC, and corresponding single stainings (IgG1+, yellow; CD138+, green; PanCK+, pink). Scale bar, 50 µm, ×40 magnification. CD138+ plasma cells pre- versus post-treatment (P = 0.0182) and IgG1+CD138+ plasma cells pre- versus post-treatment (P = 0.0155) from patients with paired tumor tissue samples (n = 16) are separately represented by box plots (cells per mm2), and according to pathological response (pCR (n = 6) (P = 0.013)/(P = 0.0312) and non-pCR (n = 10) (P = 0.275)/(P = 0.1602)) (two-sided Wilcoxon signed-rank test; statistical significance determined with P < 0.05). Data are presented as mean ± s.e.m. Each box extends from the 25th (Q1, lower bound of the box) to the 75th (Q3, upper bound of the box) percentile, the horizontal line in the center of the box represents the median value (Q2), the lower bound of lower whiskers mark the 5th (min) percentile and the upper bound of whiskers the 95th (max) percentile. Dots represent individual values. f, Bar plots displaying terms from the MSigDB Hallmarks database. Terms are ranked by P value and the total number of genes overlapping differentially expressed genes and adjusted P value are indicated (enricher one-sided hypergeometric test with false discovery rate correction; Benjamini–Hochberg)77. The universe is defined by all detected genes in the B cell/plasma cell subsets.

Source data

The integrated single-cell B cell receptor sequencing (scBCR-seq) and scRNA-seq of 13,219 of 25,166 B cells (53%) revealed that half of the B cell receptor (BCR) repertoire was composed of large or hyper-expanded clones. Hyper-expanded clones were observed in plasma cells only (cluster 3, 4, 24 and 22), whereas B cell clones presented a medium and small expansion (cluster 21, 7 and 38) (Extended Data Fig. 9a,b). We did not identify significant differences between patients with a pCR and non-pCR, regarding the absolute number of BCR clones and repertoire clonality (Extended Data Fig. 9c). Characterization of marker genes of hyper-expanded, large, medium and small clones further confirmed transcriptional heterogeneity of B and plasma cells (Extended Data Fig. 9d). Relative frequency of immunoglobulin heavy constant (IGH) isotypes in B and plasma cell clusters revealed that plasma cells were mainly enriched within IGHγ (IGHG) classes, whereas B cells were enriched within IGHμ (IGHM) classes, and to a smaller extend IGHG, IGHδ (IGHD) and IGHα (IGHA). Hyper-expanded plasma clones were nearly exclusively IGHG1-positive plasma cells (Fig. 5d), which were previously reported to be associated with a proinflammatory and cytotoxic antitumor response36. Four-plex IF staining validated the presence of CD138+ and IgG1+CD138+ plasma cells and we detected a significant increase upon T-VEC treatment (P = 0.0182; two-sided Wilcoxon signed-rank test, statistical significance determined with P < 0.05) (137.0 versus 678.6 cells per mm2) and (P = 0.0155) (45.6 versus 388.1 cells per mm2), respectively. Of note, we observed that patients with a pCR had a significantly increased number of CD138+ plasma cells (P = 0.013) (mean 1.0 (s.d. ± 2.103) versus mean 482.0 (s.d. ± 385.7) cells per mm2) and IgG1+CD138+ plasma cells (P = 0.0312) (mean 0.1 (s.d. ± 0.1830) versus 235.9 (s.d. ± 241.5) cells per mm2), whereas non-pCR patients had a nonsignificant increase of CD138+ plasma cells (P = 0.2754) (mean 218.5 (s.d. ± 289.1) versus 796.6 (s.d. ± 798.1) cells per mm2) and IgG1+CD138+ plasma cells (P = 0.1602) (mean 73.0 (s.d. ± 98.24) versus 479.5 (s.d. ± 599.2) cells per mm2) (Fig. 5e). In addition, hyper-expanded clones showed an enrichment in oxidative phosphorylation signatures associated with antibody producing plasma cells37 (Fig. 5f).