Unravelling homologous recombination repair deficiency and therapeutic opportunities in soft tissue and bone sarcoma

Abstract

Defects in homologous recombination repair (HRR) in tumors correlate with poor prognosis and metastases development. Determining HRR deficiency (HRD) is of major clinical relevance as it is associated with therapeutic vulnerabilities and remains poorly investigated in sarcoma. Here, we show that specific sarcoma entities exhibit high levels of genomic instability signatures and molecular alterations in HRR genes, while harboring a complex pattern of chromosomal instability. Furthermore, sarcomas carrying HRDness traits exhibit a distinct SARC-HRD transcriptional signature that predicts PARP inhibitor sensitivity in patient-derived sarcoma cells. Concomitantly, HRD^high sarcoma cells lack RAD51 nuclear foci formation upon DNA damage, further evidencing defects in HRR. We further identify the WEE1 kinase as a therapeutic vulnerability for sarcomas with HRDness and demonstrate the clinical benefit of combining DNA damaging agents and inhibitors of DNA repair pathways ex vivo and in the clinic. In summary, we provide a personalized oncological approach to treat sarcoma patients successfully.

Keywords: HRD score; HRDness; genomic instability; sarcoma.

Figure 1. A subset of sarcoma entities exhibit elevated genomic instability signatures and a high degree of alterations in HRR genes

1. Quantification of loss‐of‐heterozygosity (LOH) in soft tissue and bone sarcoma.

2. Quantification of large‐scale transitions (LST) in soft tissue and bone sarcoma.

3. Quantification of telomeric allelic imbalances (TAI) in soft tissue and bone sarcoma.

4. Quantification of homologous recombination deficiency (HRD) score in soft tissue and bone sarcoma.

5. Oncoprint depicting the molecular alterations in 70 HRR genes in soft tissue and bone sarcoma and the total number of alterations. HGSOC, high‐grade serous ovarian cancer; TNBC, triple‐negative breast cancer; CRC, colorectal cancer; UPS, undifferentiated pleomorphic sarcoma; MFS, myxofibrosarcoma; OS, osteosarcoma; ULMS, uterine leiomyosarcoma; LMS, extra‐uterine leiomyosarcoma; DDLPS, dedifferentiated liposarcoma; SS, synovial sarcoma; DT, desmoid tumor.

Data information: Datasets from TCGA‐SARC (n = 247), TCGA‐OV (n = 61), TCGA‐BRCA (n = 92), TCGA‐COAD (n = 385) and TARGET‐OS (n = 69) were used; n indicates biological replicates. Data in (A–D) are median ± third and first quartile, the whiskers are minimum and maximum values. Source data are available online for this figure.

Figure 2. A subset of sarcoma entities exhibit high levels of chromosomal instability and mutational signatures of HRDness

1. A–D

   Heatmaps with hierarchical clustering showing aneuploidies (A), gains (B), losses (C) and total CIN (D) per chromosome in HRD^high and HRD^low soft tissue and bone sarcoma, HGSOC, CRC and TNBC.

2. E–H

   Quantification of aneuploidies (E), gains (F), losses (G) and total number of alterations including gains and losses (H) in HRD^high and HRD^low soft tissue and bone sarcoma, HGSOC, CRC and TNBC.

3. I–L

   CIN of cytobands including HRR genes in HRD^high LMS (I), ULMS (J), MFS (K) and UPS (L) compared to HRD^low.

4. M

   List of the chromosomal cytobands that include HRR genes and exhibit increased CIN in HRD^high compared with HRD^low sarcoma.

5. N

   Total number of alterations in STS patients previously treated with chemotherapy, radiotherapy or both.

6. O

   Correlation matrix showing genomic instability signatures and CIN.

7. P

   Matrix dot plot of mutational signatures in STS.

Data information: Datasets from TCGA‐SARC (n = 247), TCGA‐OV (n = 61), TCGA‐BRCA (n = 92), TCGA‐COAD (n = 385) and TARGET‐OS (n = 69) were used; n indicates biological replicates. Data in (A–D) are mean‐centre and scaled, in (E–H) and (N) are median ± third and first quartile, the whiskers are minimum and maximum values. Statistical significance in (I–L) and (N) was determined using Mann–Whitney U‐test and showed statistical differences in (N): Chemo‐ & radiotherapy, P = 0.008; Chemotherapy, P = 0.047; and Radiotherapy, P = 0.0001; *P < 0.05; **P < 0.01; ***P < 0.001. Exact P‐values for (I–L) can be found in Appendix Table S3. Source data are available online for this figure.

Figure EV1. HRD score cut‐off based on genomic alterations in HRR genes (HRR‐CIN)

1. A–D

   Histograms depicting the frequency of LOH, LST, TAI, and HRD score in sarcoma.

2. E

   Bimodal distribution of HRR‐CIN in sarcoma.

3. F

   Receiver operating characteristic (ROC) curve of HRD scores from TCGA‐SARC cohort using HRR‐CIN for binary classification.

4. G

   Implementation of the Youden index on the ROC curve (F) to select the optimal cut‐off value for the HRD score 32 in soft tissue sarcoma. Datasets from TCGA‐SARC (n = 247) were used.

Source data are available online for this figure.

Figure 3. Additional sarcoma cohorts show a similar pattern of HRDness and chromosomal instability in distinct sarcoma entities

1. A

   Quantification of LOH, LST, TAI and HRD score in MFS.

2. B

   Fraction of genome with gains and losses in an MFS cohort.

3. C

   Oncoprint depicting the molecular alterations in HRR genes in MFS and the total number of alterations.

4. D–F

   Heatmaps of gains (D), losses (E), and total CIN (F) per chromosome in MFS.

5. G

   Quantification of LOH, LST, TAI and HRD score in angiosarcoma. Dotted line indicates the HRD score cut‐off of 32.

6. H

   Fraction of genome with gains and losses in an angiosarcoma cohort.

7. I

   Oncoprint depicting the molecular alterations in HRR genes in angiosarcoma and the total number of alterations.

8. J–L

   Heatmaps of gains (J), losses (K) and total CIN (L) per chromosome in angiosarcoma.

9. M

   LOH and tumor mutational burden (TMB) across several soft tissue and bone sarcoma cohorts. LOH values are reported as genome‐wide percentage of LOH events. TMB values are reported as mutations per megabase.

10. N

    Oncoprint depicting molecular alterations in HRR genes (included in the FoundationOne®HEME assay) in several sarcoma entities as well as their corresponding microsatellite (MS) status.

Data information: Datasets from patient cohorts of the University Hospital Zurich were used; n = 5 (A–F), 21 (G‐L) and 282 (M, N); n indicates biological replicates. Data in (D–F) and (J–L) were mean‐centre and scaled. Source data are available online for this figure.

Figure EV2. HRDness in soft tissue and bone sarcoma cohorts is associated with molecular alterations in HRR pathway genes

1. LOH, LST, TAI, and HRD score in OS, LMS, ULMS and UPS cohorts.

2. Fraction of genome with gains and losses in OS, LMS, ULMS and UPS cohorts.

3. Oncoprint depicting gains and losses in chromosomal regions of HRR genes (included in the Affymetrix Genome‐wide Human SNP 6.0 arrays and Oncoscan array) and the total number of alterations in OS, LMS, ULMS and UPS cohorts.

Data information: Datasets from GSE33153, GSE154591, and GSE119043 were used; n = 30 OS, 34 ULMS, 30 LMS, and 20 UPS. Source data are available online for this figure.

Figure 4. HRD^high sarcoma exhibit a distinct SARC‐HRD transcriptional signature and enrichment in DNA repair and cell cycle control pathways

1. A

   Volcano plot showing enrichment of HRR genes in HRD^high compared with HRD^low sarcoma cases.

2. B

   Enrichment score of HRR genes in HRD^high compared with HRD^low sarcoma cases. Normalized enrichment P‐value = 2.5e10⁻¹⁰.

3. C

   Heatmap with hierarchical clustering of differentially expressed genes in HRD^high compared with HRD^low sarcoma cases.

4. D, E

   GSEA showing hallmark (D) and KEGG (E) pathways enriched in HRD^high compared with HRD^low sarcoma cases.

Data information: Datasets from TCGA‐SARC (n = 247) were used; n indicates biological replicates. Source data are available online for this figure.

Figure EV3. Differential expression of HRR genes and gene set enrichment analysis in HRD^high compared with HRD^low sarcoma across sarcoma histotypes

1. A–F

   Volcano plot showing HRR genes, enrichment score of HRR genes, GSEA showing hallmark and KEGG pathways enriched in HRD^high compared with HRD^low UPS (A), MFS (B), ULMS (C), MPNST (D), LMS (E) and DDLPS (F). Normalized enrichment P‐values < 0.01 for UPS, MFS, MPNST, LMS, and DDLPS. Datasets from TCGA‐SARC (n = 247) were used.

Source data are available online for this figure.

Figure 5. HRD^high sarcomas show sensitivity to PARP and WEE1 inhibition and synergy with chemotherapy drugs

1. A

   Quantification of LOH, LST, TAI and HRD score in patient‐derived sarcoma cell models.

2. B

   Fraction of genome with gains and losses in patient‐derived sarcoma cell models.

3. C

   Oncoprint depicting the molecular alterations in HRR genes in patient‐derived sarcoma cell models and the total number of alterations.

4. D

   Heatmap with hierarchical clustering of the SARC‐HRD gene signature in HRD^high compared with HRD^low sarcoma cell models. Data was mean‐centre and scaled.

5. E–G

   Ex vivo treatment of HRD^high (NMFH‐1, OH931, PM197, USZ‐21_MFS2 and USZ‐21_UPS1) sarcoma models compared with HRD^low (USZ‐20_REA1, USZ‐21_LG1, USZ‐22_EMC2, USZ‐20_SFT1 and USZ‐21_CIC1) sarcoma models for 4 days with six doses of the chemotherapy agents oxaliplatin, doxorubicin, and trabectedin.

6. H

   Heatmap of IC₅₀ showing drug sensitivity responses in all cell models.

7. I, J

   HRD^high sarcoma cell models show sensitivity to the PARPi olaparib (I) and niraparib (J) when treated for 12 days.

8. K

   HRD^high sarcoma cell models show sensitivity to the WEE1 inhibitor adavosertib when treated for 12 days.

9. L

   Heatmap of IC₅₀ showing sensitivity to PARPi and WEE1i in HRD^high but not HRD^low sarcoma cell models. The ovarian carcinoma cell line UWB1.289 with BRCA1 mutations was used as positive control for PARPi response and HRDness.

10. M–O

    HRD^high sarcoma cell models and UWB1.289 treated for 3 days with five doses of olaparib alone and in combination with 1 nM trabectedin. Note that no sensitivity to olaparib in monotherapy was observed at 3 days but from 8 days on (see Figs EV5A and B).

11. P

    Heatmap of the synergy scores ZIP, Loewe, Bliss and HSA showing synergy in the combinatorial modality.

12. Q–S

    HRD^high sarcoma cell models and UWB1.289 treated for 3 days with five doses adavosertib alone and in combination with 100 nM doxorubicin.

13. T

    Heatmap of the synergy scores ZIP, Loewe, Bliss and HSA showing synergy in the combinatorial modality.

Data information: n = 5 HRD^high and 5 HRD^low sarcoma cell models (A–L), 2–3 repetitions in technical triplicates (A–S); n indicates biological replicates. Data in (E–G), (I–K), (M–O) and (Q–S) are mean ± s.d. Source data are available online for this figure.

Figure EV4. Molecular characterization of patient‐derived sarcoma cell models

1. A–C

   Heatmaps of gains (A), losses (B) and total CIN (C) per chromosome in sarcoma cell models.

2. D

   Volcano plot showing enrichment of HRR genes in HRD^high compared with HRD^low sarcoma cell models.

3. E

   Enrichment score of HRR genes in HRD^high compared with HRD^low sarcoma cell models. Normalized enrichment P‐value < 0.01.

4. F, G

   GSEA showing hallmark (F) and KEGG (G) pathways enriched in HRD^high compared with HRD^low sarcoma cell models.

Data information: n = 5 HRD^high and 5 HRD^low sarcoma cell models. Data in (A–C) are mean‐centre and scaled. Source data are available online for this figure.

Figure EV5. PARPi and WEE1i monotherapy and combination therapy in patient‐derived sarcoma cell models

1. A, B

   HRD^high sarcoma cell models show sensitivity to the PARPi olaparib (A) and niraparib (B) when treated for 8 days.

2. C

   HRD^high sarcoma cell models show sensitivity to the WEE1 inhibitor adavosertib when treated for 8 days.

3. D

   Heatmap of IC₅₀ showing sensitivity to PARPi and WEE1i in HRD^high but not HRD^low sarcoma cell models. The ovarian carcinoma cell line UWB1.289 with BRCA1 mutations was used as positive control for PARPi response and HRDness.

4. E

   Area under the curve (AUC) for oxaliplatin, doxorubicin and trabectedin upon 4 days treatment; corresponding dose–response curves depicted in Fig 5E–G.

5. F

   AUC for olaparib, niraparib and adavosertib upon 8 days treatment; corresponding dose–response curves depicted in Fig EV5A–C.

6. G

   AUC for olaparib, niraparib and adavosertib upon 12 days treatment; corresponding dose–response curves depicted in Fig 5I–K.

7. H, I

   HRD^low sarcoma cell models treated for 3 days with five doses olaparib alone and in combination with 1 nM trabectedin.

8. J, K

   HRD^low sarcoma cell models treated for 3 days with five doses adavosertib alone and in combination with 100 nM doxorubicin.

Data information: n = 5 HRD^high and 5 HRD^low sarcoma cell models (A–G), n indicates biological replicates. Data are mean ± s.d. One‐tailed unpaired t‐test comparing AUC of HRD^high and HRD^low sarcoma cell models showed significant differences in olaparib, niraparib and adavosertib response in HRD^high and HRD^low cell models; *P < 0.05; **P < 0.01; ***P < 0.0001; ns, not significant (P > 0.05); P = 2.75 × 10⁻⁵ (Ola, F), P = 0.039 (Nir, F), P = 1.49 × 10⁻⁴ (Ada, F), P = 1.17 × 10⁻⁵ (Ola, G), P = 0.028 (Nir, G), P = 6.82 × 10⁻⁴ (Ada, G). Source data are available online for this figure.

Figure 6. Absence of RAD51 nuclear foci in patient‐derived HRD^high sarcoma cells upon trabectedin and olaparib‐induced DNA damage

1. A

   Immunofluorescence showing nuclear expression of the DNA damage marker γH2A.X (magenta) upon 6 h treatment with 10 nM trabectedin and 100 nM olaparib in combination in HRD^high and HRD^low sarcoma cell models as well as ovarian carcinoma UWB1.289 cells.

2. B

   Immunofluorescence showing RAD51 nuclear foci (magenta) upon 6 h treatment with 10 nM trabectedin and 100 nM olaparib in combination only in the HRD^low sarcoma cell model (USZ‐21_LG1).

3. C, D

   Quantification of γH2A.X (C) or RAD51 (D) nuclear intensity per cell compared to untreated control cells.

4. E

   Immunohistochemistry (IHC) in human tissue sections showing RAD51 (brown) expression in HRD^low but not HRD^high sarcoma patients. Small‐nucleated lymphocytes and/or proliferating cells appeared RAD51‐positive in HRD^high models.

5. F

   Immunofluorescence in human tissue sections showing RAD51 (magenta) nuclear foci in an HRD^low but not an HRD^high sarcoma patient.

6. G

   Timeline of LMS patient diagnosis and treatment.

7. H

   Magnetic resonance imaging (MRI) of LMS patient since primary diagnosis. Open arrowheads point at metastatic lesions. PDX, primary diagnosis; PD, progressive disease; AMI, acute myocardial infarction; PR, partial response; SD, stable disease; CT, computer tomography; MRI, magnetic resonance imaging.

8. I

   Genomic profiling of a gastric metastasis showing HRD score and fraction of genome altered.

9. J

   RAD51 IHC in the metastatic patient's tissue. Compare RAD51 nuclear expression in normal tissue (arrowhead points at gastric gland) but not in tumorous gastric tissue.

10. K

    Oncoprint depicting the molecular alterations in HRR genes in the metastatic sample.

Data information: n = 2–3 biological replicates in technical triplicates (A–D), n indicates biological replicates. Scale bars, 100 μm (E), 25 μm (J), 10 μm (A, B, F). Data are mean ± s.d. A one‐way ANOVA with Sidak's multiple comparison test revealed statistically significant differences in (C) (F(3, 8) = 3, P = 0.09 for USZ‐21_LG1; F(3, 8) = 12.6, P = 0.002 for UWB1.289; F(3, 8) = 52.9, P < 0.0001 for USZ‐21_MFS2; F(3, 8) = 7.4, P = 0.01 for USZ‐21_UPS1) and (D) (F(3, 8) = 10.4, P = 0.004 for USZ‐21_LG1; F(3, 8) = 2.2, P = 0.17 for UWB1.289; F(3, 8) = 1.3, P = 0.34 for USZ‐21_MFS2; F(3, 8) = 37.4, P < 0.0001 for USZ‐21_UPS1). *P < 0.05; **P < 0.01; ***P < 0.0001; ns, not significant (P > 0.05). Source data are available online for this figure.