Article https://doi.org/10.1038/s41467-022-33975-6 A phase II trial of weekly nab-paclitaxel for progressive and symptomatic desmoid tumors Received: 31 March 2022 A list of authors and their affiliations appears at the end of the paper Accepted: 10 October 2022 Desmoid fibromatosis (DF) are mesenchymal neoplasms, with potential aggressive course and relevant clinical impact. New systemic therapy mod- Check for updates alities are needed in this symptomatic/progressive population. In this multi- 1234567890():,; 1234567890():,; center, phase II trial (NCT03275818), patients with symptomatic/progressing DF received three cycles of weekly nab-paclitaxel. Brief pain inventory short form (BPI-SF) was collected at baseline and in every visit. MRI was performed every 3 months. Primary composite endpoint was RECIST 1.1 overall response rate (ORR) and/or clinical response (improvement ≥ 2 points in BPI-SF). If 40% of patients achieved clinical/radiological response, further investigation would be warranted. Toxicity, progression-free survival (PFS), pattern of response and its correlation with clinical best response and BPI, variation of physical function, and analgesic consumption were secondary endpoints. The transla- tional research reported was not a pre-specified secondary outcome. Forty eligible patients started therapy, being 35 radiologically and clinically evalu- able. The study achieved its primary endpoint, as 7(20%) patients obtained RECIST partial response, whereas 31(89%) experienced pain reduction of ≥2 points in BPI-SF worst pain. Therapy was well tolerated. With a median follow- up of 30(14–44) months, median 12 and 24-months PFS rates were 91%(CI 95%, 82–100) and 84%(CI 95%, 71–97). For clinical progression, 12 and 24-months PFS rates were 85% (CI 95%, 73–97) and 74% (CI 95%, 58–90) respectively. Short course of nab-paclitaxel is active, safe and achieves quick and durable responses in progressing/symptomatic DF patients. Desmoid fibromatosis (DF) constitutes a ubiquitous mesenchymal On the one hand, large surgical series from sarcoma expert centers clonal neoplasia demonstrating a fibroblastic/myofibroblastic pro- reported worrisome figures, with 5-year relapse-free survival ranging liferation in long fascicles, sometimes with notorious collagen fibers, from 52.8%3 to 69%4, or an ~90% incidence of relapse if surgery was with infiltrating local growth but a null metastatic potential1. This offered in recurrent DF5. Furthermore, recurrent DF cases usually intriguing tumor may exhibit different behaviors, from indolent subtle display a more clinically apparent disease. On the other hand, obser- growth to fast aggressive infiltration or even to spontaneous regres- vational studies reported a similar 3-year event-free survival between sion. In general, 50% of DF tumors show a pauci-symptomatic com- conservative vs up-front surgical approaches, 68 vs 65%, respectively6. portment, and this has been more evident since the wait-and-see Tumor location has prognostic influence, with the DF of abdominal policies have become widespread2. wall being the most frequently related to spontaneous regression The paradigm switch from surgery as cornerstone to a non- (28.4% in a large series), and the best tolerated for wait-and-see interventional preferred approach in DF is related to several factors. policies7,8. At the other extreme, extra-abdominal DF tumors have the e-mail: jmartin@atbsarc.org Nature Communications | (2022)13:6278 1 Article https://doi.org/10.1038/s41467-022-33975-6 highest recurrence rate3, especially DF tumors located in girdles or nab-paclitaxel every 3 weeks, was superior to paclitaxel in terms of extremities9. Furthermore, DF can present as a particularly painful response and time to progression. mass or can provoke a functional impairment of an extremity that Our hypothesis is that a low dose and a short course of nab- forces the use of therapy. Likewise, wait-and-see policy is also not paclitaxel would be advantageous over other prolonged chemother- recommended if the DF tumor growth is threatening vital structures apy administrations and might be enough to induce a quick sympto- (such as major vessels, nerves, visceral organs etc.), or in the context of matic response and/or radiological response in measurable and disturbing symptoms. symptomatic DF patients. Of note, DF may frequently induce new symptoms or clinical Here we show the results of a phase II trial exploring 3-month nab- progression in the absence of RECIST progressive disease (PD)10. This paclitaxel administration, in this specific subset of symptomatic DF singular fact of DF has barely been addressed in studies11,12 despite its patients. paramount relevance, since persistent symptoms can determine a poorly perceived quality of life in DF patients. Particularly, pain is the Results most debilitating symptom reported by DF patients13. Substantial Patients and treatment symptomatic relief with different treatments is frequently seen in DF From May 2017 to September 2019, 43 patients were assessed for patients in the absence of RECIST response, somewhat already pro- eligibility in 8 hospitals belonging to GEIS network. Finally, 40 spectively described two decades ago14. patients were enrolled in the trial (first and last patient enrolled on The median time to clinical improvement or to response is lacking 9 May 2017 and 27 September 2019 respectively) as 2 patients did in most studies; however, this factor is critical to appreciate the added not meet the inclusion criteria and 1 patient refused to be enrolled. value of a treatment against DF. In prospective studies with systemic The safety population was constituted by these 40 patients since all treatment, the median time to radiological response has been received at least one nab-paclitaxel administration. There were 39 observed as around 10 months15–17. In contrast, the median time to patients evaluable for RECIST response, one patient only received radiological stabilization with a wait-and-see strategy takes longer than three nab-paclitaxel doses and then decided to discontinue due to a 1 year17,18, or even more than 2 years19, according to different prog- grade 2 allergic reaction. Four out of the 40 patients did not reach nostic factors in the series. 2-points in the worst pain form of BPI-SF and, as a result, 36 con- The gain-of-function mutation of the CTNNB1 gene in sporadic DF stituted the evaluable population for clinical response. Finally, 35 entails, among other effects, an increase of HIF-1α transcription20. patients were fully evaluable for both primary endpoints (Fig. 1). Additionally, Wnt pathway activation is related to metabolic remo- The clinical cut-off for the final analyses of data was April 30, 2021. deling and accentuates the Warburg effect that, ultimately, stimulates Median dose intensity for nab-paclitaxel in the 35 evaluable neoangiogenesis21. Microtubule-targeting drugs, such as taxanes, patients for the two main endpoints was 100% (33–100%). In this inhibit angiogenesis through HIF-1α protein translation22. These find- population, the number of patients with dose reductions was 4 ings provide rationale for the use of taxanes in DF patients. Besides, (11%), while those with any dose delay or discontinuation numbered taxanes can also inhibit metalloproteinase-7 (MMP-7) and vascular 10 (29%). endothelial growth factor (VEGF), both targets of beta-catenin and Demographics and clinical characteristics are listed in Table 1. At overexpressed in DF. entry, there were 13 out of 40 patients (32.5%) with exclusive radi- Among taxanes, nanoparticle albumin-bound paclitaxel (nab- ological RECIST progression, 12 out of 40 patients (30%) with exclusive paclitaxel) is a novel formulation of paclitaxel that does not require symptomatic progression and 15 out of 40 patients (37.5%) with both solvents for its formulation. In breast cancer there is considerable radiological and symptomatic progression. Previous surgical resection accrued experience with this compound. In the pivotal phase III trial, had been performed in 21 patients (52%), while 10 (25%) had previously Assessed for enrolment (n=43) Excluded (n=3) 2 patient did not meet inclusion criteria 1 patient decision Patients enrolled (n=40) Safety population (n=40) Excluded (n=4) Excluded (n=1) 4 patients with Baseline Brief Pain Inventory 1 patient did not complete cycle 1 of therapy (BPI) < 2 in ‘’Worst pain’’ item Evaluable for RECIST response Evaluable for clinical response (n=39) (n=36) Evaluable for RECIST and clinical response (n=35) Fig. 1 | Consort diagram. This figure summarizes the flux of patients in the study. Nature Communications | (2022)13:6278 2 Article https://doi.org/10.1038/s41467-022-33975-6 Table 1 | Patient demographics patients (89%). Additionally, ≥4-point and ≥6-point decreases from baseline were seen in 20 (57%) and 11 (31%) patients (Supple- N = 40 mentary Table 1). Median age, years (range) 38 (18–76) Regarding the 39 RECIST-evaluable patients, there were 10 (26%) Gender (M/F) 14 (35%)/26 (65%) PR and 29 (74%) SD as a result of local assessment, while there were 9 Reason for inclusion: (23%) PR, 29 (74%) SD, and 1 (3%) PD after central radiological assess- Progression 28 (70%) ment (Fig. 2). Remarkably, the median time to obtain radiological PR Pain 26 (65%) was 3 (0.5–12.2) and 6.3 (2.3–18.6) months for local and central Functional impairment 20 (50%) assessments, respectively. And the median time to obtain a clinical response was 0.9 months (0.2–2.9). Median time diagnosis to enrollment, months (range) 17 (1–187) Secondary endpoints Median previous relapses 1 (0–10) As listed in Table 2, the proportion of patients that experienced at least Previous therapy: 23 (57%) one episode of grade 3 or 4 toxicities was exceptionally low. In detail, Previous surgery: 21 (52%) grade 3 neutropenia was seen in 7.5% of patients, grade 3 mucositis in More than 1 7 (17%) 2.5% and grade 3 peripheral neuropathy in 2.5%. No grade 4 or 5 toxi- Previous radiotherapy: 5 (12%) cities were observed in the trial. The most frequent hematological Previous chemotherapy: 6 (15%) toxicities were neutropenia (57.5%) and anemia (52.5%). Among non- hematological toxicities, fatigue (55%), alopecia (42.5%), increased Location: alanine aminotransferase (ALT) (37.5%), and diarrhea (35%) were the Thoracic wall 10 (25%) most frequent treatment emergent adverse effects. Intraabdominal 7 (17%) With a median follow-up of 30 (14–44) months, 5 patients (14%) Upper extremity 7 (17%) experienced a radiological progression in accordance with central Lower extremity 7 (17%) assessment while 8 patients (23%) experienced a clinical progression Head and neck 6 (15%) within the efficacy population (n = 35). Neck and proximal upper Abdominal wall 3 (7%) extremities had significantly worse 30-month PFS rate, 20% compared Location 2: with 80% in other tumor locations (Supplementary Table 2). The median PFS at 12 and 24 months for RECIST and central review were High risk (Head and neck 91% (CI 95%, 82–100) and 84% (CI 95%, 71–97), respectively in the Or proximal Upper extremity) 11 (27%) efficacy population. For clinical progression or increase of at least Other 29 (72%) 2-points in the worst pain questionnaire of BPI-SF, the proportion of ECOG BASELINE: patients with PFS rate at 12 and 24 months were 85% (CI 95%, 73–97) 0 15 (38%) and 74% (CI 95%, 58–90), respectively. 1 24 (61%) Considering analgesic drug consumption, 11 patients (31%) Extension at baseline: reduced the score by at least one in the AQA, 21 patients (60%) did not Localized 18 (45%) change the score and 2 (5%) patients increased the score by one (Supplementary Table 1). Locally advanced 22 (55%) In the univariate analysis, the only two factors with significant Resectable baseline: favorable prognosis in 30-month PFS rate for RECIST were younger age Resectable 3 (7%) (≤50 years old) and the achievement of a partial radiological response Unresectable 37 (92%) (Table 3). Univariate analysis for primary endpoint is shown in Sup- Nuclear β-catenin expression plementary Table 2. The variation of physical function was a pre- Positive 39 (98%) specified outcome. However, further secondary endpoint analysis, Negative 0 (0%) including this latter, will be presented in a separate article, more Not assessablea 1 (2%) focused on quality of life parameters. No other relevant deviations from the planned analysis have been identified. CTNNB1 mutation: T41A 22 (55%) Translational research S45F 3 (8%) Differential gene expression analysis was carried out by grouping NA 15 (37%) samples according to RECIST response [responders (n = 5) vs. non- M male, F female, NA not available. responders (n = 11)]. A total of 865 genes were significantly (p < 0.05) a The diagnosis of this case was confirmed by sequencing CTNNB1 gene. and differently expressed according to RECIST response; 383 and 482 were overexpressed or underexpressed, respectively, in the samples of patients that had a PR to nab-paclitaxel, compared to non-responders. Among these genes, 43 genes were significantly and differently received at least one class of systemic treatment including overexpressed in the responders group, when considering the genes chemotherapy. with a logFC >1 (Supplementary Table 3). No gene was significantly and differently underexpressed in the responders group when a logFC < −1 Primary endpoints cut-off was applied. Radiological response observed in the subset of 35 patients, eva- EnrichR found two MSigDB Hallmarks that significantly over- luable for both main endpoints, was distributed as partial response lapped with genes differently expressed between patients that (PR) in 9 (26%) and stable disease (SD) in 26 (74%) according to local responded to the treatment or experienced disease progression. evaluation, whereas following central review, there were 7 PR (20%), Those two pathways are PI3K/AKT/mTOR signaling and Angiogenesis 27 SD (77%) and 1 PD (3%). The clinical response in the same popu- (Supplementary Fig. 1). Angiogenesis enrichment was supported by lation, ≥2-point decrease in worst pain from baseline, was seen in 31 THBD, VTN, SPP1, LPL and FGFR1. PI3K/AKT/mTOR signaling was Nature Communications | (2022)13:6278 3 Article https://doi.org/10.1038/s41467-022-33975-6 Maximum diameter variation 40% Change from baseline in longest diameter of 20% target lession dimensión (%) 0% -20% -40% Progression -60% Stable disease Partial Response -80% -100% Patients Fig. 2 | Water fall plot. Among the 39 evaluable patients with evaluable data, the of diameters of target lesions) and for partial response (at least 30% decrease), Overall Response Rate (ORR) by RECIST 1.1 was 23%. Lower and upper dashed lines respectively. Partial response (blue); Stable disease (yellow) and Progression represent the cut-off for progressive disease (an increase of at least 20% in the sum (purple). Source data are provided as a Source Data file. supported by TIAM1, PIKFYVE, GRK2, PRKCB, CAMK4, MKNK2, ITPR2, studies address this question. One prospective study acknowledged MAPK1, TRIB3, and FGF22. that time to clinical response was a matter of a few weeks after sys- The unsupervised clustering carried-out using the 43 gene-set temic treatment with gamma-secretase inhibitor, but it was not grouped patients into 3 different subsets (Fig. 3). Interestingly, a sig- quantified16. In a retrospective study, the median time to clinical nificant correlation was found favoring group 3 with respect to RECIST response with toremifene, defined as any pain improvement, was response achievement, 56% vs 0% (p = 0.034) for joined groups 1 and 2. 4 months and the authors communicated a total of 75% of pain improvement in 16 symptomatic patients11. Other retrospective series Discussion treated with hormone therapy, in this case tamoxifen, reported 31% of In this phase II trial, we found an early clinical and radiological, cen- symptomatic relief from a series of 35 DF patients27. Additionally, the trally reviewed response, of 89 and 20%, respectively. This outcome time to radiological response reported in a series with systemic was achieved in a series of 35 progressing patients by radiological treatments was around 10 months15–17. Our prospective trial compares RECIST 1.1, clinical criteria or both, and evaluable for the two main favorably with these previous studies, showing a median time to clin- endpoints. Hence, this short-term regimen with nab-paclitaxel was ical and radiological responses as short as 0.9 and 6.3 months, active and safe in DF, with the accomplishment of the alternative respectively. Notably, the time to spontaneous regression or stabili- hypothesis of obtaining at least 40% response in either of the two main zation is barely mentioned in published series, but it takes more than endpoints. The 2-point decrease in the BPI-SF score of worst pain was one year17–19. conservatively selected following previous recommendations23,24 and Other outcomes such as PFS and duration of response are also was probed to be clinically meaningful in a previous phase II trial crucial in DF, indicating how long certain therapeutic strategies are exploring denosumab in aggressive giant cell tumor of bone patients, a able to concede disease control for both symptomatic and radi- similar context of a locally aggressive mesenchymal tumor with null or ologically aspects. Reports of PFS in some larger series of DF with very low metastatic potential25. active surveillance were 2-year PFS of 71%28, 3-year PFS of 38%29, or Composite main endpoints are rarely seen in clinical trials, but in 5-year PFS rate of 49.9%7. Regarding PFS reports in series treated with recent times are increasingly being considered26. In the case of DF, this systemic therapies, the 2-year PFS rates usually exceed 80%11,17, and is of paramount importance since a single primary endpoint does not 5-year PFS rates ranged from 55 to 81%15,30. These previous reports have provide a comprehensive clear image of what active treatments are focused on PFS following RECIST; in this respect our study reports a potentially attaining. In many circumstances, clinical impairment in the competitive 2-year PFS rate of 86% by central radiology review. absence of radiological RECIST progression of DF represents a worthy Importantly, a recent phase 3 trial in DF communicated 2-year PFS argument for starting therapy11. By contrast, patients may dramatically rates of 81 and 36% for sorafenib and placebo arms, respectively. This improve their symptoms even when no RECIST response was side of the coin should also be taken into account in some contexts obtained16. This has also been the case of our patients where the pro- where active surveillance may not be the best choice. In line with this, portion of subjects with clinical response was 89%, a fourfold increase our series might represent an extreme in the spectrum of biological in comparison with radiological responses. Additionally, the time to behavior of DF: all patients were symptomatic, and with radiological or reach a response may also importantly be perceived by patients. Few clinical progression, or both. Moreover, the proportion of abdominal Nature Communications | (2022)13:6278 4 Article https://doi.org/10.1038/s41467-022-33975-6 Table 2 | Toxicity profile (n = 40) Toxicity Any grade (%) Grade 1 (%) Grade 2 (%) Grade 3 (%) Grade 4 (%) Hematological – n (%) Neutropenia 23 (57.5) 10 (25.0) 10 (25.0) 3 (7.5) 0 Anemia 21 (52.5) 16 (40.0) 5 (12.5) 0 0 Leukopenia 16 (40.0) 13 (32.5) 1 (2.5) 2 (5.0) 0 Thrombocytopenia 2 (5.0) 2 (5.0) 0 0 0 Lymphocytopenia 1 (2.5) 0 1 (2.5) 0 0 Non-hematological – n (%) Fatigue 22 (55.0) 13 (32.5) 9 (22.5) 0 0 Alopecia 17 (42.5) 3 (7.5) 14 (35.0) 0 0 ALT increased 16 (40.0) 12 (30.0) 4 (10.0) 0 0 Diarrhea 14 (35.0) 11 (27.5) 3 (7.5) 0 0 Neuropathy 14 (35.0) 11 (27.5) 2 (5.0) 1 (2.5) 0 Nausea 13 (32.5) 12 (30.0) 1 (2.5) 0 0 AST increased 10 (25.0) 10 (25.0) 0 0 0 Myalgia/arthralgia 10 (25.0) 7 (17.5) 3 (7.5) 0 0 Hypernatremia 9 (22.5) 9 (22.5) 0 0 0 Mucositis oral 9 (22.5) 8 (20.0) 0 1 (2.5) 0 Hypoglycemia 7 (17.5) 7 (17.5) 0 0 0 Rash and other skin disorders 6 (15.0) 6 (15.0) 0 0 0 Fever 5 (12.5) 5 (12.5) 0 0 0 Headache 5 (12.5) 4 (10.0) 1 (2.5) 0 0 Hyponatremia 5 (12.5) 3 (7.5) 2 (5.0) 0 0 Pruritus 5 (12.5) 4 (10.0) 1 (2.5) 0 0 Constipation 4 (10.0) 4 (10.0) 0 0 0 Dysgeusia 4 (10.0) 3 (7.5) 1 (2.5) 0 0 Nail changes 4 (10.0) 3 (7.5) 1 (2.5) 0 0 Vomiting 4 (10.0) 4 (10.0) 0 0 0 Anorexia 3 (7.5) 2 (5.0) 1 (2.5) 0 0 Dizziness 3 (7.5) 3 (7.5) 0 0 0 Insomnia 3 (7.5) 3 (7.5) 0 0 0 Other gastrointestinal disorders 3 (7.5) 2 (5.0) 1 (2.5) 0 0 Hyperkalemia 2 (5.0) 1 (2.5) 1 (2.5) 0 0 Hypocalcemia 2 (5.0) 1 (2.5) 1 (2.5) 0 0 Hypokalemia 2 (5.0) 2 (5.0) 0 0 0 Localized edema 2 (5.0) 2 (5.0) 0 0 0 Other respiratory disorders 2 (5.0) 2 (5.0) 0 0 0 ALP increased 1 (2.5) 1 (2.5) 0 0 0 Nab-paclitaxel allergic reaction 1 (2.5) 0 1 (2.5) 0 0 Anxiety 1 (2.5) 1 (2.5) 0 0 0 Blood bilirubin increased 1 (2.5) 0 1 (2.5) 0 0 CPK increased 1 (2.5) 1 (2.5) 0 0 0 Epistaxis 1 (2.5) 1 (2.5) 0 0 0 Hypercalcemia 1 (2.5) 1 (2.5) 0 0 0 Hypersomnia 1 (2.5) 1 (2.5) 0 0 0 Hypotension 1 (2.5) 1 (2.5) 0 0 0 Odynophagia 1 (2.5) 1 (2.5) 0 0 0 Pneumonitis 1 (2.5) 0 1 (2.5) 0 0 Vision loss 1 (2.5) 1 (2.5) 0 0 0 ALT alanine aminotransferase, AST aspartate aminotransferase, ALP alkaline phosphatase, CPK creatine phosphokinase. wall DF, the most favorable site for regression or PFS8, constituted only substantial proportion of abdominal cases 28 and 43% for regorafenib a 7% in our study. In the case of regorafenib trial, the patients were and placebo arms, respectively17. On the other hand, the pazopanib enrolled in progression, defined as size increase of ≥10%, or sympto- trial enrolled progressive, but not all symptomatic, patients31. In matic or with unresectable tumors. Further, the trial enrolled a addition, neck DT accounted for 2% compared to 15% in our study. Nature Communications | (2022)13:6278 5 Article https://doi.org/10.1038/s41467-022-33975-6 Table 3 | Univariate progression-free survival analysis Intention-to-treat population (n = 40) Intention-to-treat sub-population (n = 35) 30 months- P value 30 months- P value 30 months- P value 30 months- P value central PFS CR PFS* central PFS CR PFS* Median age (years): 0.11 0.057 0.15 0.051 18–38 94% (83–100) 77% (58–97) 93% (79–100) 80% (59–100) >38 77% (56–97) 46% (20–73) 76% (55–97) 45% (19–71) Age (years): <0.001 0.25 <0.001 0.33 18–50 93% (83–100) 66% (48–83) 92% (81–100) 65% (46–84) >50 57% (20–94) 57% (20–94) 57% (20–94) 57% (20–94) Sex (M/F): 0.23 0.093 0.2 0.15 Male 77% (53–100) 48% (21–75) 73% (47–99) 69% (47–91) Female 91% (78–100) 72% (52–92) 90% (76–100) 50% (22–78) Median time diagnosis to enroll- 0.85 0.93 0.62 0.95 ment (range): 0–17 84% (67–100) 62% (39–84) 81% (61–100) 61% (36–86) >17 87% (71–100) 64% (40–88) 87% (70.100) 63% (39–87) Previous relapses: 0.37 0.81 0.41 0.9 No 79% (58–100) 65% (40–91) 78% (56–100) 62% (35–89) Yes 90% (77–100) 61% (39–83) 89% (74–100) 62% (39–85) Previous surgery: 0.64 0.38 0.62 0.58 No 83% (65–100) 73% (53–93) 81% (61–100) 70% (48–92) Yes 88% (75–100) 52% (27–77) 87% (70–100) 55% (29–81) Previous chemotherapy: 0.37 0.6 0.37 0.44 No 83% (70–97) 63% (45–81) 81% (67–93) 63% (44–82) Yes 100% (NA) 60% (17–100) 100% (NA) 50% (1–99) Location: 0.15 <0.001 0.16 <0.001 Head and neck & Proximal upper 71% (43–100) 20% (0–51) 71% (43–100) 20% (0–51) extremity Other 90% (77–100) 81% (63–98) 90% (77–100) 80% (63–98) ECOG baseline: 0.32 0.18 0.22 0.43 0 79% (59–100) 80% (60–100) 74% (49–99) 75% (50–99) 1 90% (77–100) 50% (27–73) 89% (76–100) 54% (30–77) Local response: 0.72 0.28 0.72 0.35 PR 90% (71–100) 80% (55–100) 89% (68–100) 78% (51–100) SD 84% (69–99) 56% (36–76) 82% (66–98) 56% (35–78 f) Central response: <0.001 <0.001 <0.001 <0.001 PR 100% (NA) 100% (NA) 100% (NA) 100% (NA) SD 84% (69–98) 53% (33–73) 83% (67–98) 54% (33–75) PD 0% (NA) 0% (NA) 0% (NA) 0% (NA) Change in T2 enhance sequence 0.23 0.13 0.17 0.056 (MRI)a: 0–7.5 94% (82–100) 79% (57–100) 94% (82–100) 79% (57–100) >7.5 77% (57–97) 54% (30–78) 73% (51–96) 46% (20–71) New score: 0.63 0.42 0.53 0.25 No change 100% (NA) 67% (13–100) 100% (NA) 67% (13–100) 0–4 90% (71–100) 80% (55–100) 90% (71–100) 80% (55–100) 5–9 100% (NA) 100% (NA) 100% (NA) 100% (NA) >10 77% (57–97) 54% (30–78) 73% (51–96) 46% (20–71) Reduction in pain (BPI worst pain 0.98 0.73 0.45 0.37 reduction) At least −2 87% (65–100) 56% (17–95) 85% (72–99) 50% (1–99) <−2 85% (72–99) 65% (47–83) 75% (32–100) 65% (47–83) a Change in T2: defined as the percentage in changes of the fibrous component of the tumor in T2-enhanced sequence during therapy. Two-sided Fisher’s exact test. Multiple comparisons were not used for this analysis. PFS progression-free survival, CR PFS clinical/radiological PFS, M male, F female, PR partial response, SD stable disease, PD progressive disease, BPI Brief Pain Inventory. Nature Communications | (2022)13:6278 6 Article https://doi.org/10.1038/s41467-022-33975-6 RECIST_response 10 RECIST_response Gender PARTIAL RESPONSE Location PROGRESSIVE DISEASE PPM1A 8 CASK STABLE DISEASE GTF3A MTRNR2L9 Gender LRRC17 6 SERPINF1 Female RALGAPA1 Male AKAP12 SPARCL1 4 RBM6 Location DPT SMARCE1 gastric ACTA2 2 HNRNPA1 thoracic/extrem/head&neck CNN3 RBM39 STEAP2 FRG1 CHD1 DDX3Y RYK MX1 ADGRL4 SLC12A2 ACSL5 LRRC23 MAP3K4 HNRNPM PDXDC2P NKTR NPIPA5 NPIPB15 NPIPB9 VCL HNRNPU FILIP1L ATP6V1E1 ISG20L2 SPATS2L CBWD1 MYLK UHRF2 X107.CEL X1004.CEL X1005.CEL X1101.CEL X504.CEL X126.CEL X506.CEL X1006.CEL X101.CEL X131.CEL X117.CEL X119.CEL X110.CEL X902.CEL X1012.CEL X607.CEL Group 1 Group 2 Group 3 Fig. 3 | Heatmap clustering of patients based on gene expression profiling. cluster annotation, samples are clustered, with k-mean clustering method, Patients (columns; n = 16) were clustered in three groups based on a combination of according to three features: gender (m/f), location of the tumor and type of mRNA differentially expression data (rows), modeled according to patient response [RECIST classification: Partial Response (PR), Stable Disease (SD) or Pro- response (fitting a linear model through limma package). Zooming to column-wise gression Disease (PD)]. Source data are provided as a Source Data file. In contrast to the long-lasting schemes of systemic treatment that our short scheme was not enough to maintain clinical and/or radi- typically have been used in DF, our proposal intended to offer a short- ological benefits in DF of the neck or proximal upper extremities. In term regimen, whilst the other benefits, such as activity and safety, are line with this, DF of the neck have been related to a higher recurrence maintained. The classic weekly low-dose of methotrexate plus vinca rate in previous publications35,36, and DF of the limbs resulted in a more alkaloids duration is around 1 year14,32, and it is not free of concerning unfavorable location than the trunk in some reports8. toxicity as neutropenia grade 3 or peripheral neuropathy31 that A limitation of this trial is that, even when all cases were centrally impelled investigators to make modifications such as a biweekly reviewed before accrual and exhibited a characteristic immunostain- administration, but still with a median duration that exceeded one ing profile, the mutational status was available for only 55% of cases. year15. The duration of targeted therapies in DF trials has been similar The difficulties in selecting an adequate questionnaire for measuring or even longer. Thus, imatinib duration in DF ranged between 1 and 2 functionality in all the DF locations hindered the systematic collection years12,33, sorafenib was administered without specific duration and of some dramatic improvements in this area. widely exceeded one year17, pazopanib was prescribed for one year31, The finding of enrichment for angiogenesis and mTOR-related and the median duration with gamma secretase inhibitors was more genes among those differently expressed in RECIST responders or than two years34. Despite these studies communicating manageable non-responders, is in line with our expectations. Nab-paclitaxel was treatment-emergent adverse effects, they also reported grade 3/4 supposed to act as an antiangiogenic through the inhibition of some toxicities in the range of 11 to 45% for Imatinib, 56% for pazopanib, 31% cytokines such as bFGF and VEGF, and by other interfering mechan- for sorafenib, and 47% for gamma secretase inhibitors. Accordingly, isms. Interestingly, the inhibition mTOR pathway can reduce the pro- shorter schemes, such as our tested regimen, represent a rational cess of angiogenesis through the inhibition of HIF-1α37. The statistically alternative attaining earlier and durable responses while avoiding significant correlation favoring group 3 with higher probability to sustained toxicity related with more durable treatments. Nevertheless, obtain a RECIST reponse, gives us new insights on potential new Nature Communications | (2022)13:6278 7 Article https://doi.org/10.1038/s41467-022-33975-6 targets and predictive gene- signature that should be analyze on the inhibitors or hormone therapy were allowed. Availability of a paraffin whole series and then externally validated with other series using embedded tumor block for central pathology confirmation and antiangiogenic in DF. translational research was mandatory. An effective method of con- Since DF behavior is heterogeneous, wait-and-see options should traception had to be used in men or women with childbearing not be universally recommended. Developing new strategies, such as potential, before entry into the study, throughout the treatment and shorter regimens, can be beneficial for patients if toxicity can be for 6 months after ending the study treatment. Most relevant exclusion minimized whilst maintaining adequate activity. criteria were prior taxanes therapy, more than one previous che- Three-month nab-paclitaxel administration showed to be active, motherapy line, and unavailability to undergo MRI, DF tumors with ill- quick to achieve responses, durable for maintaining responses (except defined margins, previous irradiated target lesion or women who are for neck and proximal upper extremities) and safe. This regimen pregnant or breast-feeding. Secondary side effects were collected warrants further investigation in symptomatic and progressing DF. according to CTCAE 4.0. The first patient was enrolled in 9 May 2017 and the last patient was enrolled in 27 September 2019. Methods Study design Statistics and reproducibility ABRADES (Abraxane in desmoid and desmoplastic tumors) study, a The primary endpoint of the study was double: overall response rate phase II, non-randomized, open-label, multicenter clinical trial with (ORR) according to RECIST 1.1 or clinical response with an improve- two different cohorts, was sponsored by the Spanish Group for ment of at least 2 points in the BPI-SF worst pain score, in the absence Research on Sarcoma (GEIS) and conducted in 14 Spanish sarcoma of PD. Secondary objectives were progression-free survival (PFS) expert centers, in pediatric and adult populations (aged ≥6 months). measured as a median, variation of symptoms other than pain and also (ClinicalTrials.gov identifier NCT03275818). Data from the DF cohort is scored in the BPI-SF and analgesic consumption during the first year, presented here, from which 8 centers were actively accruing patients. analysis of the safety profile of nab-paclitaxel and central assessment The study protocol was approved by a central ethics committee of the pattern of responses in different sequences of MRI. For sample (“Provincial de Sevilla” Research Ethics Committee (Seville, Spain)), size estimation, a feasible Simon two-stage design was used, selecting covering all participating centers at national level in Spain. Procedures error rates alpha equal to 5% and beta equal to 20%. The option related were performed in accordance with good clinical practice (GCP) to efficacy was 40% of either ORR and/or clinical response, whatever of guidelines and in compliance with the Declaration of Helsinki. All both. Success, defined as ORR and/or clinical response in 20% of the patients signed a written informed consent form to participate in the cases or less will be considered as unacceptable and would not warrant study, including consent for use of tumoral samples. Patients aged further investigation (null hypothesis, H0 20%). While success defined ≥6 months and ≤20 years received nab-paclitaxel 240 mg/m2, while as ORR and/or clinical response in 40% of the cases or more will be patients aged ≥21 years received nab-paclitaxel 125 mg/m2, on days 1, 8, considered as an acceptable result warranting further investigation of and 15 in cycles of 28 days. Patients in the DF cohort received a max- the drug in DF (alternative hypothesis, H1 40%). A total of 21 eligible imum of three cycles. Nab-paclitaxel was administered intravenously and treated patients was estimated to be included in the first stage. If over 30 min, without corticosteroid or antihistamine premedication. ≤4 patients had radiological and/or clinical response, the trial would be Central pathology review was a compulsory requirement for trial stopped. Otherwise, with >4 patients with radiological and/or clinical entry. Radiological assessment was only allowed by MRI, with perfu- response, the trial will continue to accrue up to 35 eligible and evalu- sion and diffusion sequences being mandatory, specific thorough able patients to enter the study. If 12 or more successes were observed determinations are reflected in the protocol. Studies with MRI were in those 35 subjects, it will be concluded that the results of the trial performed at baseline and every 3 months for the first year, and then warrant further investigation38. Comparisons between qualitative every 6 months for two additional years. Central radiological anon- variables were done using the Fisher Exact Test or Chi-square. Time to ymous assessment was compulsory for all imaging studies. Brief Pain event variables (OS and PFS) were measured from the date of therapy Inventory - Short Form (BPI-SF) was collected at baseline, days 1, 8, and onset and were estimated according to the Kaplan-Meier method. 15, at the end of treatment and every 3 months for the first 3 years. Comparisons between the variables of interest were performed by the Additionally, analgesic quantification algorithm (AQA) was collected at log-rank test. All p-values reported were 2-sided, and statistical sig- the same intervals as the BPI-SF. More details can be found within the nificance was defined at p < 0.05. Statistical analysis was performed protocol through this link: https://grupogeis.org/GEIS-39_Abrades_ with SSPS version 28. Protocol_V2_220218-clean.pdf. A copy of this protocol is also available as a Supplementary Note in the Supplementary Information File. Translational research Gene expression profiling was performed using Clariom™ S Pico Patients human assay (Applied Biosystems™; ThermoFisher Scientific, Inc.; Eligible patients had centrally confirmed histology of DF; measurable Foster City, CA, USA), which accurately measured the expression levels disease according to RECIST 1.1 criteria; and which score was at least 2 of more than 20,000 well-annotated genes, in order to obtain a points in the worst pain section of the BPI-SF. It was determined that transcriptome-wide gene-level expression profile. Sixteen samples this decrease in the score was the minimally important difference were used in this exploratory analysis due to budget restrictions. (MID) that patients perceived as relevant and that would have led the Briefly, RNA was amplified and labeled using the GeneChip® WT PLUS clinician to plan a change in the patient’s management23. The MID for Reagent Kit (Thermo Fisher Scientific, Inc.; Waltham, MA, USA). BPI-SF was conservatively defined as 2 points in the score on the basis Amplification was performed with 100 ng of total RNA input, following of anchor-based methods24 and previous similar experience25. Addi- the procedures described in the WT PLUS Reagent Kit user manual. tional criteria were Eastern Cooperative Oncologic Group (ECOG) The amplified cDNA was quantified, fragmented, and labeled to performance status ≤1, deep DF from extremities, trunk or head and hybridize to GeneChip® Clariom S Human Array (Thermo Fisher Sci- neck regions, intra-abdominal DF were enrolled if harboring CTNNB1 entific, Inc.), using 5.5 μg of single-stranded cDNA product and fol- mutation. Patients had to have clinical or radiological progression in lowing manufacturers’ protocols. Washing, staining (GeneChip® the previous 6 months, and adequate bone marrow, renal and liver Fluidics Station 450, Thermo Fisher Scientific, Inc.), and scanning function. Normal cardiac function with left ventricular ejection frac- (GeneChip® Scanner 3000, Thermo Fisher Scientific, Inc.) were per- tion of at least 50% had to be demonstrated by echocardiogram or by formed following the protocols outlined in the user manual for car- multigated acquisition scan. Previous chemotherapy, prostaglandin tridge arrays. Nature Communications | (2022)13:6278 8 Article https://doi.org/10.1038/s41467-022-33975-6 Bioinformatics analysis 6. Bonvalot, S. et al. Extra-abdominal primary fibromatosis: aggressive Arrays were processed (background corrected, quantile normalized management could be avoided in a subgroup of patients. Eur. J. and summarized) via the RMA (Robust Multi-array Average) method Surg. Oncol. 34, 462–468 (2008). from the oligo package (v1.54.1). Summarization was achieved with 7. Fiore, M. et al. Desmoid-type fibromatosis: a front-line conservative the corresponding custom CDF (pd.clariomshuman.hs.entrezg approach to select patients for surgical treatment. Ann. Surg. v25.0.0) from Brainarray (http://brainarray.mbni.med.umich.edu/ Oncol. 16, 2587–2593 (2009). Brainarray). 8. Salas, S. et al. Prognostic factors influencing progression-free sur- Differential expression analyses were performed using the vival determined from a series of sporadic desmoid tumors: a wait- Ebayes function in the limma package (v3.46.2). Analyses was exe- and-see policy according to tumor presentation. J. Clin. Oncol. 29, cuted taking into account RECIST 1.1 response [responders (n = 5) 3553–3558 (2011). vs. non-responders (SD and PD; n = 11)], using non-responders as the 9. Gronchi, A. et al. Quality of surgery and outcome in extra- reference for data analyses. Pathway enrichment was accomplished abdominal aggressive fibromatosis: a series of patients surgi- with Enrichr39 on the Hallmark gene sets (MSigDB Hallmark 2020)40, cally treated at a single institution. J. Clin. Oncol. 21, restricting the input to those genes showing differential expression 1390–1397 (2003). (p-value ≤ 0.05). Benjamini and Hochberg correction for multiple 10. Desmoid Tumor Working Group. The management of desmoid comparisons was applied and a p-value threshold of 0.05 was tumours: A joint global consensus-based guideline approach for defined41. A clustering was performed in the final filtered gene-set adult and paediatric patients. Eur J Cancer 127, 96–107 (2020). using FactoMineR and Factoextra R packages in an unsupervised 11. Fiore, M. et al. Hormonal manipulation with toremifene in sporadic manner relying on both Gap Statistic and Average Silhouette desmoid-type fibromatosis. Eur. J. Cancer 51, 2800–2807 (2015). methods to determine the optimal number of groups for the 12. Penel, N. et al. Imatinib for progressive and recurrent aggressive K-means algorithm. Costumed Heat map was produced using fibromatosis (desmoid tumors): an FNCLCC/French Sarcoma pheatmap R package. Group phase II trial with a long-term follow-up. Ann. Oncol. 22, 452–457 (2011). Reporting summary 13. Husson, O. et al. Desmoid fibromatosis through the patients’ eyes: Further information on research design is available in the Nature time to change the focus and organisation of care? Supportive Care Research Reporting Summary linked to this article. Cancer 27, 965–980 (2019). 14. Azzarelli, A. et al. Low-dose chemotherapy with methotrexate and Data availability vinblastine for patients with advanced aggressive fibromatosis. The Microarray dataset is available in the ArrayExpress database under Cancer 92, 1259–1264 (2001). accession code E-MTAB-12163. The clinical variables used for the 15. Nishida, Y. et al. Desmoid with biweekly methotrexate and vin- translational research and raw gene expression data are available in the blastine shows similar effects to weekly administration: a phase II ArrayExpress database under the access [E-MTAB-12163]. Complete clinical trial. Cancer Sci. 111, 4187–4194 (2020). clinical data is provided in Supplementary Data 1 and 2. Data from 16. Villalobos, V. M. et al. Long-term follow-up of desmoid fibromatosis patients was anonymized. All the identifiers, with the exception of treated with PF-03084014, an oral gamma secretase inhibitor. Ann. gender, have been omitted. Shared public results implied statistical Surg. Oncol. 25, 768–775 (2018). results (significant tests), displayed in form of tables (i.e. data that 17. Gounder, M. M. et al. Sorafenib for advanced and refractory des- passes a filter) and statistical plots (e.g. differentially expressed genes moid tumors. N. Engl. J. Med. 379, 2417–2428 (2018). heatmaps). Shared tables contemplate normalized genes counts, fil- 18. Barbier, O. et al. Primary or recurring extra-abdominal desmoid tering for genes significance and genes differential expression. Plots fibromatosis: assessment of treatment by observation only. Orthop. include expression heatmaps. The pipeline of the study is shared in Traumatol. Surg. Res. 96, 884–889 (2010). detail in the methods section of the paper. The study protocol is 19. Kim, Y., Rosario, M. S., Cho, H. S. & Han, I. Factors associated with available in the Supplementary Information file. The remaining data disease stabilization of desmoid-type fibromatosis. Clin. Orthop. are available within the Article, Supplementary Information, Supple- Surg. 12, 113–119 (2020). mentary Data or Source Data file. 20. Kaidi, A., Williams, A. C. & Paraskeva, C. Interaction between beta- catenin and HIF-1 promotes cellular adaptation to hypoxia. Nat. Cell Code availability Biol. 9, 210–217 (2007). The custom code is permanently available in the ArrayExpress data- 21. Pate, K. T. et al. Wnt signaling directs a metabolic program of gly- base under accession code E-MTAB-12163. colysis and angiogenesis in colon cancer. EMBO J. 33, 1454–1473 (2014). References 22. Escuin, D., Kline, E. R. & Giannakakou, P. Both microtubule- 1. Fletcher CD, B. J. et al. WHO Classification of Tumours: Soft Tissue stabilizing and microtubule-destabilizing drugs inhibit hypoxia- and Bone Tumours. 5th Edition, Vol. 3 (IARC, Lyon, 2020). ISBN inducible factor-1alpha accumulation and activity by disrupting 9789283245025. microtubule function. Cancer Res. 65, 9021–9028 (2005). 2. Bonvalot, S. et al. The treatment of desmoid tumors: a stepwise 23. Guyatt, G. H., Osoba, D., Wu, A. W., Wyrwich, K. W. & Norman, G. R. clinical approach. Ann. Oncol. 23, x158–x166 (2012). Methods to explain the clinical significance of health status mea- 3. Peng, P. D. et al. Management and recurrence patterns of desmoids sures. Mayo Clin. Proc. 77, 371–383 (2002). tumors: a multi-institutional analysis of 211 patients. Ann. Surg. 24. Mathias, S. D. et al. Estimating minimally important differences for Oncol. 19, 4036–4042 (2012). the worst pain rating of the Brief Pain Inventory-Short Form. J. 4. Crago, A. M. et al. A prognostic nomogram for prediction of Support Oncol. 9, 72–78 (2011). recurrence in desmoid fibromatosis. Ann. Surg. 258, 25. Martin-Broto, J. et al. Effects of denosumab on pain and analgesic 347–353 (2013). use in giant cell tumor of bone: interim results from a phase II study. 5. Tsagozis, P., Stevenson, J. D., Grimer, R. & Carter, S. Outcome of Acta Oncol. 53, 1173–1179 (2014). surgery for primary and recurrent desmoid-type fibromatosis. A 26. Hamasaki, T., Evans, S. R. & Asakura, K. Design, data monitoring, and retrospective case series of 174 patients. Ann. Med. Surg. 17, analysis of clinical trials with co-primary endpoints: a review. J. 14–19 (2017). Biopharm. Stat. 28, 28–51 (2018). Nature Communications | (2022)13:6278 9 Article https://doi.org/10.1038/s41467-022-33975-6 27. Libertini, M. et al. Aggressive fibromatosis response to tamoxifen: Author contributions lack of correlation between MRI and symptomatic response. Clin. Concept and design (J.M.B.); assembly of data (all the authors); data Sarcoma Res. 8, 13 (2018). analysis and interpretation of the results (J.M.B., D.S.M., J.M.M., A.G., J.Z., 28. de Bruyns, A. et al. Evolving practice patterns over two decades and N.H.); manuscript draft (J.M.B., D.S.M., J.Z., and N.H.); final approval (1993-2013) in the management of desmoid-type fibromatosis in of manuscript (all the authors). British Columbia. Clin. Oncol. (R. Coll. Radiol.) 32, e102–e110 (2020). 29. Turner, B. et al. Surgical excision versus observation as initial Competing interests management of desmoid tumors: a population based study. Eur. J. J.M.-B. reports research grants from PharmaMar, Eisai, Immix BioPharma, Surgical Oncol. 45, 699–703 (2019). and Novartis outside the submitted work; honoraria for advisory board 30. Palassini, E. et al. Long-term efficacy of methotrexate plus vin- participation and expert testimony from PharmaMar, Eli Lilly and Com- blastine/vinorelbine in a large series of patients affected by pany, Bayer, GSK, Novartis, Amgen, Roche, Tecnofarma and Asofarma; desmoid-type fibromatosis. Cancer J. 23, 86–91 (2017). and research funding for clinical studies (institutional) from PharmaMar, 31. Toulmonde, M. et al. Pazopanib or methotrexate-vinblastine com- Eli Lilly and Company, BMS, Pfizer, AROG, Bayer, Eisai, Lixte, Kar- bination chemotherapy in adult patients with progressive desmoid yopharm, Deciphera, GSK, Celgene, Novartis, Blueprint, Nektar, Forma, tumours (DESMOPAZ): a non-comparative, randomised, open- Amgen, and Daichii-Sankyo. D.S.M. reports institutional research grants label, multicentre, phase 2 study. Lancet Oncol. 20, from PharmaMar, Eisai, Immix BioPharma, and Novartis outside the 1263–1272 (2019). submitted work; travel support from PharmaMar, Eisai, Celgene, Bayer 32. Schneider, N., Hallin, M. & Thway, K. STAT6 loss in dedifferentiated and Pfizer, and personal fees from Tecnopharma, outside the submitted solitary fibrous tumor. Int. J. Surg. Pathol. 25, 58–60 (2017). work. Nadia Hindi reports grants, personal fees and non-financial sup- 33. Kasper, B. et al. Imatinib induces sustained progression arrest in port from PharmaMar, personal fees from Lilly and Tecnopharma, grants RECIST progressive desmoid tumours: Final results of a phase II from Eisai, and Novartis, outside the submitted work and research study of the German Interdisciplinary Sarcoma Group (GISG). Eur. J. funding for clinical studies (institutional) from PharmaMar, Eli Lilly and Cancer 76, 60–67 (2017). Company, AROG, Bayer, Eisai, Lixte, Karyopharm, Deciphera, GSK, 34. Kummar, S. et al. Clinical activity of the gamma-secretase inhibitor Novartis, Blueprint, Nektar, Forma, Amgen, and Daichii-Sankyo. The PF-03084014 in adults with desmoid tumors (aggressive fibroma- remaining authors declare no competing interests. tosis). J. Clin. Oncol. 35, 1561–1569 (2017). 35. de Bree, E. et al. Desmoid tumors of the head and neck: a ther- Additional information apeutic challenge. Head Neck 36, 1517–1526 (2014). Supplementary information The online version contains 36. Risoud, M., Mortuaire, G., Leroy, X., Leblond, P. & Fayoux, P. Des- supplementary material available at moid tumours of the head and neck in children: review of man- https://doi.org/10.1038/s41467-022-33975-6. agement. Eur. Ann. Otorhinolaryngol. Head Neck Dis. 134, 155–160 (2017). Correspondence and requests for materials should be addressed to 37. Toschi, A., Lee, E., Gadir, N., Ohh, M. & Foster, D. A. Differential Javier Martin-Broto. dependence of hypoxia-inducible factors 1 alpha and 2 alpha on mTORC1 and mTORC2. J. Biol. Chem. 283, 34495–34499 (2008). Peer review information Nature Communications thanks Frederic Chi- 38. Jung, S. H., Lee, T., Kim, K. & George, S. L. Admissible two-stage bon, Alexander Lazar and Varadan Sevilimedu for their contribution to designs for phase II cancer clinical trials. Stat. Med. 23, the peer review of this work. 561–569 (2004). 39. Xie, Z. et al. Gene set knowledge discovery with enrichr. Curr. Reprints and permission information is available at Protoc. 1, e90 (2021). http://www.nature.com/reprints 40. Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 1, 417–425 (2015). Publisher’s note Springer Nature remains neutral with regard to jur- 41. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a isdictional claims in published maps and institutional affiliations. practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B (Methodol.) 57, 289–300 (1995). Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, Acknowledgements adaptation, distribution and reproduction in any medium or format, as The authors would like to thank the Spanish group for Research on Sar- long as you give appropriate credit to the original author(s) and the comas (GEIS) for supporting the study and BMS/ Celgene (study ID AX- source, provide a link to the Creative Commons license, and indicate if CL-GEIS-007345) for providing drug supply. The sponsor (GEIS) has been changes were made. The images or other third party material in this in charge of study design, data collection, analysis and manuscript writ- article are included in the article’s Creative Commons license, unless ing. The authors also thank the donors and the University Hospital Virgen indicated otherwise in a credit line to the material. If material is not del Rocío for participating in patient recruitment. The authors would like included in the article’s Creative Commons license and your intended to thank Araceli Rodriguez Morales and Gabriela Golab for Data Man- use is not permitted by statutory regulation or exceeds the permitted agement. David S. Moura is recipient of a Sara Borrell postdoctoral fel- use, you will need to obtain permission directly from the copyright lowship funded by the National Institute of Health Carlos III (ISCIII) (CD20/ holder. To view a copy of this license, visit http://creativecommons.org/ 00155). The authors would also like to thank SELNET project. SELNET has licenses/by/4.0/. received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement No. 825806. © The Author(s) 2022 Nature Communications | (2022)13:6278 10 Article https://doi.org/10.1038/s41467-022-33975-6 Javier Martin-Broto 1,2,3,4 , Andres Redondo5, David S. Moura1, Claudia Valverde6, Jose Manuel Morales7, Antonio Lopez-Pousa8, Javier Martinez-Trufero 9, Antonio Gutierrez10, Roberto Díaz-Beveridge11, Pablo Luna12, Virginia Martinez-Marin5, David Marcilla13, Ivan Arribas 14, Patricio Ledesma15, Jose Antonio Lopez-Martin 16, Davide Di Lernia1, Jorge Zamora1 & Nadia Hindi 1,2,3,4 1 Health Research Institute-Fundación Jiménez Díaz University Hospital, Universidad Autónoma de Madrid (IIS-FJD, UAM), 28040 Madrid, Spain. 2Medical Oncology Department, Fundación Jimenez Diaz University Hospital, 28040 Madrid, Spain. 3General de Villalba University Hospital, 28400 Madrid, Spain. 4 Autonomous University of Madrid, 28049 Madrid, Spain. 5Department of Medical Oncology, Hospital Universitario La Paz, 28046 Madrid, Spain. 6Department of Medical Oncology, Vall d’Hebron University Hospital, 08035 Barcelona, Spain. 7Radiology Department, Virgen del Rocio University Hospital, 41013 Sevilla, Spain. 8Medical Oncology Department, Sant Pau Hospital, 08025 Barcelona, Spain. 9Medical Oncology Department, Miguel Servet University Hospital, 50009 Zaragoza, Spain. 10Hematology Department, University Hospital Son Espases, 07120 Mallorca, Spain. 11Medical Oncology Department, Hospital Universitari i Politècnic La Fe, 46026 Valencia, Spain. 12Medical Oncology Department, Son Espases University Hospital, 07120 Mallorca, Spain. 13 Pathology Department, Virgen del Rocio University Hospital, 41013 Sevilla, Spain. 14Universitat de València, 46010 Valencia, Spain. 15Sofpromed Investi- gacion Clinica SLU, 07009 Mallorca, Spain. 16Medical Oncology Department, 12 de Octubre University Hospital, 28041 Madrid, Spain. e-mail: jmartin@atbsarc.org Nature Communications | (2022)13:6278 11