METHOD FOR GENERATING IMMUNOREGULATORY CELLS IN A BLOOD-DERIVED SAMPLE

Abstract

The present invention relates to a method comprising the steps of provision of a sample derived from a blood sample of a subject that has received a checkpoint-inhibitor therapy and is suspected of developing or has developed symptoms of immune-related adverse events (irAE), adding a photosensitizing agent to the sample, and subjecting the sample to irradiation, which preferably generates immunoregulatory NK cells in said sample. In embodiments, the photosensitizing agent is 8-methoxypsoralen and/or the irradiation is UVA irradiation. In another aspect, the invention relates to immunoregulatory NK cells obtained from a method comprising the steps of provision of a sample derived from an isolated blood sample of a subject, adding a photosensitizing agent to the sample, and subjecting the sample to irradiation. Furthermore, the invention encompasses immunoregulatory NK cells for use in the treatment and/or prevention of irAE in a subject that has received a checkpoint-inhibitor therapy.

Claims

1.-15. (canceled)

16. A method for treating a subject in need thereof that is receiving or has received a checkpoint-inhibitor therapy and shows symptoms of or suffers from irAE comprising autoimmune hepatitis using extracorporeal photopheresis (ECP), comprising: adding a photosensitizing agent to a sample derived from a blood sample of the subject in need thereof.

17. The method according to claim 16, further comprising subjecting the sample to irradiation.

18. The method according to claim 17, wherein adding the photosensitizing agent to the sample and subjecting the sample to irradiation generates or induces the formation of immunoregulatory NK cells in the sample.

19. The method according to claim 16, wherein the photosensitizing agent is 8-methoxypsoralen.

20. The method according to claim 18, wherein the irradiation is UVA irradiation.

21. The method according to claim 16, wherein the subject in need thereof suffers from malignant melanoma or another cancer treatable by checkpoint-inhibitor therapy.

22. The method according to claim 16, wherein the checkpoint-inhibitor therapy was discontinued after symptoms of irAE occurred in the subject.

23. The method according to claim 16, wherein symptoms of irAE occurred after the checkpoint-inhibitor therapy was discontinued.

24. The method according to claim 16, wherein symptoms of irAE were maintained after the checkpoint-inhibitor therapy was discontinued.

25. The method according to claim 16, wherein the checkpoint-inhibitor therapy comprises at least one of anti-CTLA4 antibodies and anti-PD-1 antibodies.

26. The method according to claim 16, wherein the subject is concurrently receiving an immunosuppressive drug and/or is refractory to immunosuppressive drugs.

27. The method according to claim 26, wherein the immunosuppressive drug is selected from the group consisting of a steroid, a corticosteroid, a cyclosporine, an anti-TNF antibody, and combinations thereof.

Description

DESCRIPTION OF THE FIGURES

[0197] FIG. 1: Refractory autoimmune colitis responding to ECP which leads to the expansion of immunoregulatory NK cells. [0198] (a) The diarrhea severity of the patient according to common toxicity criteria (y-axis) over time (x-axis) with different treatments is shown. [0199] (b) Colonoscopy image (left panel) and H&E stained biopsy section (right panel) from the initial diagnosis of the immune checkpoint inhibitor-related colitis showing mucosal edema and ulcerations. [0200] (c) Colonoscopy image (left panel) and H&E stained biopsy section (right panel) after successful ECP treatment showing no signs of an autoimmune colitis. Colonic crypts show regular morphology without granulocytic infiltrations, apoptosis or crypt loss. [0201] (d) tSNE plot visualizing the peripheral lymphocyte compartment of the patient before and 8 weeks after start ECP. [0202] (e) Relative NK cell numbers before and after ECP treatment. [0203] (f) Expression intensity of CD16, CD56 and CD57 on NK cells (defined as single/live/CD45.sup.pos/CD14.sup.neg/CD3.sup.neg/CD19.sup.neg/CD56.sup.pos) from healthy, age-matched donors (HD, n=5) and the patient at week 53 and week 75 after ECP start visualized by UMAP. FlowSOM clustered NK cell subsets are overlaid on the two plots on the right. [0204] (g) MFI (median marker expression, value range: 0-1) of CD16 on CD56.sup.dim mature NK cells. HD (n=5), whiskers of boxplot represent the minimum and maximum values of the HD dataset. [0205] (h) Survival of mice injected with T cells (Tc) with or without NK cells (NK dose of 410.sup.4 or 410.sup.1 per mouse) in a model of anti-PD1 antibody-induced irAE as described in Suppl. Appendix. **p=0.003, ****p<0.0001.

[0206] FIG. 2: [0207] (a) tSNE plot displaying 1000 stochastically selected CD45.sup.+ lymphocytes (lymphocytes defined by FSC/SSC) of each time point with their expression of the indicated markers. [0208] (b) Heatmap showing the median marker expression (value range: 0-1) for each annotated population. [0209] (c) Absolute counts of NK cells before and after ECP treatment. [0210] (d) Relative numbers of B cells, CD4.sup.+ T cells, CD8.sup.+ T cells and CD4.sup. CD8.sup. T cells before and after ECP treatment. [0211] (e) Absolute counts of B cells, CD4.sup.+ T cells, CD8.sup.+ T cells and CD4.sup. CD8.sup. T cells before and after ECP treatment. [0212] (f) Expression intensity of indicated markers on NK cells (defined as single/live/CD45.sup.pos/CD14.sup.neg/CD3.sup.neg/CD19.sup.neg/CD56.sup.pos) from HD (n=5) and the patient at week 53 and week 75 visualized by UMAP. [0213] (g) Heatmap showing the median marker expression (value range: 0-1) for the annotated NK cell subsets defined by FlowSOM clustering. [0214] (h-j) MFI (median marker expression, value range: 0-1) of indicated markers on CD56.sup.brightCD16.sup.dim immature NK cells, CD56.sup.dim mature NK cells and CD57.sup.dim terminally differentiated NK cells of healthy, age-matched donors (HD) and the patient (week 53 and week 75 after start ECP). HD (n=5), whiskers of boxplot represent the minimum and maximum values of the HD dataset.

[0215] FIG. 3: [0216] (a)-(d) Percentage of cytokine-expressing cells within CD56.sup.bright NK cells, CD56.sup.dim NK cells, CD4.sup.+ memory T cells, CD4+nave cells and CD8+ T cells of healthy, age-matched donors (HD) and the patient (week 53 and week 75 after start ECP). HD (n=5), whiskers of boxplot represent the minimum and maximum values of the HD dataset.

[0217] FIG. 4: [0218] (a) Experimental model with adoptive transfer of patient T cells alone or along with NK cells into Rag2.sup./Il2rg.sup./ mice and induction of immune-checkpoint inhibitor-associated autoimmunity by treatment with an anti-PD-1 antibody. [0219] (b) Histopathological neutrophil and lymphocyte infiltration score of liver, skin, lung and colon isolated from Rag2.sup./Il2rg.sup./ mice on day 15 after patient T and NK cell injection as shown in (a).

[0220] FIG. 5: [0221] (a) Treatment scheme of Rag2.sup./Il2rg.sup./ mice treated with an anti-PD-1 antibody without injection of patient-derived cells (negative control). [0222] (b) Neutrophil infiltration score of liver, lung, skin and colon isolated from Rag2.sup./Il2rg.sup./ mice on day 15 after start of treatment as shown (a). [0223] (c) Survival of Rag2.sup./Il2rg.sup./ mice treated as described in (a).

[0224] FIG. 6: [0225] (a) Treatment schedule: Mice were treated with DSS (3%), anti-PD1 and ECP as indicated. [0226] (b) Weight curve of mice that were untreated, or treated with DSS (3%), anti-PD1 alone or in combination with ECP as indicated. [0227] (c) Colon length was quantified in 5 mice per group. Groups as indicated in panel B. [0228] (d) Representative colon of one mouse per group. Groups as indicated in panel B. [0229] (e) Representative HE stained section of the colon of groups as indicated in panel B. [0230] (f) Histopathology scores of the colon of groups as indicated in panel B.

[0231] FIG. 7: [0232] (a) Treatment schedule: Mice were injected iv with B16 melanoma cells and afterwards treated with anti-PD1, prednisolone or ECP as indicated. [0233] (b) Survival of mice injected iv with B16 melanoma cells and afterwards treated with anti-PD1, prednisolone or ECP.

EXAMPLES

[0234] The invention is further described by the following examples. These are not intended to limit the scope of the invention, but represent preferred embodiments of aspects of the invention provided for greater illustration of the invention described herein.

[0235] We here report a patient suffering from ipilimumab/nivolumab-induced colitis, refractory to multiple immunosuppressive drugs, who achieved a complete response after extracorporeal photopheresis (ECP) coinciding with an expansion of immuno-regulatory natural killer (NK) cells.

Results of the Examples

[0236] A 29-year old male patient was treated with ipilimumab and nivolumab for metastatic melanoma. After two doses, the patient developed dermatitis, thyroiditis, hepatitis and colitis. Colitis was diagnosed based on macroscopic mucosal ulcerations and intraepithelial apoptosis and crypt loss identified in the biopsy (FIG. 1a,b). While dermatitis, thyroiditis and hepatitis resolved after stopping ipilimumab/nivolumab and corticosteroid treatment, the patient experienced three colitis episodes within the next 20 weeks (CTC-AE II-III). These were treated with corticosteroids (in total 23 weeks prior to ECP), infliximab (2 single doses 18 and 15 weeks prior to ECP) and cyclosporine (14 weeks prior to ECP, FIG. 1a and Table 1).

[0237] Since there was no durable response, the patient received ECP. During the next 8 months, he underwent 2 cycles of ECP on consecutive days every 2-4 weeks. ECP was well tolerated and led to a complete response (FIG. 1a). Immunosuppression was tapered without symptom rebound. Continuous remission of colitis was confirmed by colonoscopy (FIG. 1c). Immune checkpoint-inhibitor treatment was discontinued upon first manifestation of irAE and never resumed.

[0238] We analyzed the peripheral blood leukocyte compartment at multiple time points before and during ECP treatment. We observed a 4-fold increase in NK cells (FIG. 1d-e, FIG. 2a-j), with an immuno-regulatory phenotype (FIG. 1f,g). Also multiple pro-inflammatory cytokines were lower in the patient compared to age-matched healthy donors (FIG. 3a-e). Supporting the notion that NK cells regulated autoimmunity, adoptive transfer of the patient's NK cells prevented irAE in a murine irAE model triggered by human T cells and anti-PD-1 antibody-treatment in a dose-dependent manner (FIG. 1h and FIG. 4a,b). Control experiments confirmed that morbidity was mediated by the patient-derived T cells (FIG. 5).

[0239] ECP is an established therapy for the treatment of graft-versus-host disease (GVHD).sup.4 and leads to an increase of NK cells in GVHD patients.sup.5. Data regarding safety and efficacy of ECP for irAE treatment have been lacking so far. This case report implicates that ECP can be an efficient therapy for refractory checkpoint inhibitor-associated colitis through the expansion of a protective NK cell population.

ECP Reduces Immune Mediated Adverse Events without Blocking the Anti-Melanoma Effect

[0240] Based on the results presented above (FIGS. 1-5) showing that extracorporal photopheresis (ECP) reduced immune related colitis in a patient that had been treated with combined immunotherapy (nivolumab, ipilimumab) for metastatic melanoma, we next aimed to test this in an in vivo model for irAEs. To induce colitis, mice were treated with 3% DSS and anti-PD1 (FIG. 6A) according to previous reports (19). The treatment reduced the body weight of the mice consistent with the development of colitis and weight loss was reduced by ECP treatment (FIG. 6B). ECP treatment also increased colon length compared to the group treated with anti-PD1 only (FIG. 6C, D). Colon length was reported to be a surrogate parameter for the severity of the immunotherapy-induced colitis (19). In agreement with reduced colitis severity, we observed reduced infiltration of neutrophils in the colon wall of mice treated with ECP compared to the group treated with anti-PD1 only (FIG. 6E, F). These findings indicate that ECP reduces anti-PD1 induced colitis in mice.

[0241] To understand if the immunomodulatory effect of ECP was connected to a loss of anti-tumor activity we next treated melanoma bearing mice with anti-PD1 alone or in combination with the glucocorticoid prednisolone or ECP (FIG. 7A). We observed that prednisolone reduced the survival of melanoma-bearing mice compared to the group treated with anti-PD1 only (FIG. 7B). In contrast, the group treated with anti-PD and ECP had a comparable outcome as the group treated with anti-PD1 only (FIG. 7B). These findings indicate that ECP does not interfere with the anti-melanoma response induced by anti-PD1 treatment.

Discussion of the Examples

[0242] Discussion of the mechanism: NK cells exert a variety of heterogenic immunological functions including anti-inflammatory activity. In a murine model of GVHD, transfer of NK cells improved survival, which was dependent on intact TGF- signaling.sup.6. In another preclinical GVHD study, NK cells induced perforin and Fas ligand-mediated reduction of alloreactive T cell proliferation and increased T cell apoptosis.sup.7. A c-Kit.sup. CD27.sup. CD11b.sup.+ population was identified as a specific effector NK cell subset that was capable to control GVHD without interfering with the graft-versus-leukemia (GVL) effect.sup.8. In humans, killer cell immunoglobulin-like receptor (KIR) ligand mismatch in haploidentical allo-HCT in GVH direction reduced the risk for GVHD which was mediated by NK cells.sup.9.

[0243] ECP is an efficient treatment for GVHD. An increase in NK cells during ECP for extensive chronic GVHD has been previously observed.sup.10. Patients with acute GVHD have a higher frequency of CD56.sup.bri NK subsets with stronger NKG2D and CD62L expression.sup.11. In the same study, CD56-CD16.sup.+ NK cells with higher expression of CD57 and CD11b were increased in patients with chronic cGVHD. ECP shifted the NK cell populations towards a more immuno-regulatory phenotype with protection of a specialized anti-viral and anti-leukemic CD57.sup.+NKG2C.sup.+CD56.sup.dim subset.sup.11. We hypothesize that similar mechanisms might be responsible for the protective effects of NK cells against irAE. When comparing the NK cell compartment in the patient with that of age- and sex-matched healthy controls, we observed a downregulation of CD16 expression, in particular on CD56.sup.dim NK cells. CD16 is the FcRIII, an activating NK cell receptor that can induce strong cytokine production. Previous studies show that shedding of CD16 might be an immuno-regulatory mechanism to prevent autoimmunity.sup.12. CD16 downregulation modulates NK cell responses and contributes to maintenance of the immune homeostasis of both antibody and T cell-dependent pathways.sup.13. Supporting this hypothesis, expression of GM-CSF, IFN-, TNF and IL-2 was lower in the NK cells of the patient when compared to the control group.

Methods Employed in the Examples

ECP Procedure:

[0244] Extracorporeal photopheresis was performed on a Therakos CellEx photopheresis system with administration of Methoxsalen (Uvadex). Two procedures were performed on consecutive days with 1500 ml blood being processed during each procedure. We collected all human samples after approval by the ethics committee of the Albert Ludwigs University, Freiburg, Germany (protocol number 300/16) and after written informed consent in accordance with the Declaration of Helsinki.

Murine Model of irAE

[0245] T cells were isolated from the patient's peripheral blood using negative selection with the Pan T cell isolation kit (Miltenyi Biotec) according to the manufacturer's instructions. NK cells were isolated from the patient's peripheral blood using the NK cell isolation kit, human (Miltenyi Biotec) according to the manufacturer's instructions. Rag2.sup./Il2rg.sup./ mice were injected intravenously with 310.sup.5 T cells with or without 410.sup.4 or 410.sup.5 NK cells. From day 1 to day 22 after injection, mice were treated twice weekly with 8 mg/kg body weight anti-PD-1 antibody (clone J43) and once weekly with 1 mg/kg body weight LPS, both applied by an intraperitoneal injection (FIG. 4a). Sections of skin, liver, colon and lung collected on day 15 after human T cell injection were stained with hematoxylin-eosin and scored on the basis of histopathological characterization of human irAEs, including lymphocyte and neutrophil infiltration, crypt abscesses and apoptotic cells.sup.14,15 by an experienced pathologist. All animal studies had been approved by the University institutional review board on the Use and Care of Laboratory Animals at the Albert-Ludwigs University Freiburg, Germany (Protocol approval numbers: G17-049, X13-07J, X15-10A).

Flow Cytometry

[0246] For monitoring of lymphocyte lineage populations during and after ECP therapy, peripheral blood lymphocytes of the patient were isolated and stained with a standardized panel of antibodies against CD45, CD19, CD3, CD4, CD8, CD16, CD56 and HLA-DR as a part of the routine diagnostics. Data was compensated in FlowJo (V10), lymphocytes were exported and using the R environment.sup.17. tSNE and FlowSOM clustering were performed as previously described.sup.16.

[0247] For multiparametric NK cell phenotypisation and cytokine analysis peripheral blood lymphocytes were isolated using density gradient medium (Lymphoprep, STEMCELL Technologies) according to the manufacturer's instructions. Thawed peripheral blood lymphocytes were stained with antibodies listed in Table S2. Zombie Aqua Fixable Viability kit (Biolegend) was used for live/dead discrimination. For production of cytokines, cells were stimulated with 50 ng/ml PMA (Axon Lab) and 500 mg/ml lonomycin (Sigma) in the presence of GolgiPlug (BD Biosciences) for 4 h. Intracellular staining was performed using the BD Cytofix/Cytoperm kit (BD Biosciences) according to the manufacturer's protocol. Data was acquired on an Aurora flow cytometer (Cytek) and compensated using FlowJo (Flowjo V10.6.1, LLC) software. Cell populations specified in the figures were exported and analyzed using the R environment.sup.17. Data was processed for FlowSOM clustering as described.sup.16. For dimensionality reduction the UMAP package was used.sup.18.

Statistics

[0248] Statistical analysis was performed using the GraphPad Prism Lab Software V7.0. Comparisons of two groups were performed by two-tailed unpaired Student's t tests. Differences in survival (Kaplan-Meier survival curves) were evaluated using the Mantel Cox (log-rank) test. Data are presented as meanSEM if not otherwise indicated. A p-value <0.05 was considered to be significant.

Tables of the Examples

TABLE-US-00001 TABLE 1 Disease and treatment course. Time Maximal point diarrhea prior to severity Treatment at the New treatment ECP (CTC- beginning of added during Episode (weeks) AE) the episode the episode 1 20 3 Methylprednisolon Methylprednisolon 0.3 mg/kg BW 2 mg/kg BW (previously given for Infliximab autoimmune hepatitis) 5 mg/kg BW 2 16 3 Methylprednisolon Prednison 0.1 mg/kg BW 1 mg/kg BW Infliximab 5 mg/kg BW Cyclosporine A 3 mg/kg BW 3 2 3 Cyclosporine A ECP 1.25 mg/kg BW Prednison 7.5 mg abs

[0249] The patient had his first episode of colitis 20 weeks prior to the start of ECP. At that time point he had been treated for 3 weeks with steroids for his previous immune checkpoint inhibitor-related hepatitis, thyroiditis and dermatitis. The colitis occurred during tapering of the steroids. Therefore, the steroid dose was increased and due to an insufficient response, infliximab 5 mg/kg BW was administered once. The symptoms resolved. Upon steroid tapering, the patient experienced his second episode of colitis, 16 weeks prior to begin of ECP. He was treated with an increased dose of methylprednisolone and a second dose of infliximab. The diarrhea was refractory to this therapy and consequently cyclosporine A was added. The symptoms resolved again. As the cyclosporine A dose was reduced, the patient had a third episode of colitis. Here, ECP treatment was initiated. Two weeks after ECP start, the patient had normal bowel movement frequency. Cyclosporine A treatment was discontinued 8 weeks after ECP start, corticosteroid treatment was discontinued 12 weeks after ECP start without any symptom rebound. ECP was performed for a total of 32 weeks. With a follow-up of 11 months after the last ECP, the patient remained in complete remission with respect to both, the irAE and the melanoma manifestation.

TABLE-US-00002 TABLE 2 Antibodies used for flow cytometry with human cells. Antigen Flurochrome Clone Manufacturer CD16 BUV495 3G BD CD14 BUV563 M5E2 BD CD45RO BUV615 UCHL1 BD CD3 BUV661 UCHT1 BD CD45 BUV805 HI-30 BD NKp46 BV421 9E2 Biolegend CD56 BV480 NCAM16.2 BD CD8 BV570 RPA-T8 Biolegend CD4 BV711 OKT4 Biolegend CD4 APC-Cy7 RPA-T4 BD CD94 BV786 HP-3D9 BD CD57 PerCP-Cy5.5 HNK-1 Biolegend TIGIT PE MBSA43 eBioscience CD62L PE-Cy5 DREG-56 BD KLRG1 PE-Cy7 13F12F2 ThermoScientific NKG2D APC 1D11 Biolegend NKG2C AF488 134591 R&D CD19 BUV737 SJ25C1 BD CD19 APC-Vio770 REA675 Miltenyi IL-2 BV711 MQ1-17H12 Biolegend TNF BV785 Mab11 Biolegend IFN gamma PE-Cy7 4S.B3 eBioscience GM-CSF PE BVD2-21C11 BD

REFERENCES

[0250] 1. Larkin J, Chiarion-Sileni V, Gonzalez R, et al. Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma. N Engl J Med 2015; 373:23-34. [0251] 2. Wolchok J D, Chiarion-Sileni V, Gonzalez R, et al. Overall Survival with Combined Nivolumab and Ipilimumab in Advanced Melanoma. N Engl J Med 2017; 377:1345-56. [0252] 3. Postow M A, Chesney J, Pavlick A C, et al. Nivolumab and ipilimumab versus ipilimumab in untreated melanoma. N Engl J Med 2015; 372:2006-17. [0253] 4. Zeiser R, Blazar BR. Acute Graft-versus-Host DiseaseBiologic Process, Prevention, and Therapy. N Engl J Med; 2017; 377:2167-79. [0254] 5. Ni M, Wang L, Yang M, et al. Shaping of CD56(bri) Natural Killer Cells in Patients With Steroid-Refractory/Resistant Acute Graft-vs.-Host Disease via Extracorporeal Photopheresis. Front Immunol 2019; 10:547. [0255] 6. Asai O, Longo D L, Tian Z G, et al. Suppression of graft-versus-host disease and amplification of graft-versus-tumor effects by activated natural killer cells after allogeneic bone marrow transplantation. J Clin Invest 1998; 101:1835-42. [0256] 7. Olson J A, Leveson-Gower D B, Gill S, Baker J, Beilhack A, Negrin R S. NK cells mediate reduction of GVHD by inhibiting activated, alloreactive T cells while retaining GVT effects. Blood 2010; 115:4293-301. [0257] 8. Meinhardt K, Kroeger I, Bauer R, et al. Identification and characterization of the specific murine NK cell subset supporting graft-versus-leukemia- and reducing graft-versus-host-effects. Oncoimmunology 2015; 4:e981483. [0258] 9. Ruggeri L, Capanni M, Urbani E, et al. Effectiveness of donor natural killer cell alloreactivity in mismatched hematopoietic transplants. Science 2002; 295:2097-100. [0259] 10. Alcindor T, Gorgun G, Miller K B, et al. Immunomodulatory effects of extracorporeal photochemotherapy in patients with extensive chronic graft-versus-host disease. Blood 2001; 98:1622-5. [0260] 11. Ni M, Wang L, Yang M, et al. Shaping of CD56(bri) Natural Killer Cells in Patients With Steroid-Refractory/Resistant Acute Graft-vs.-Host Disease via Extracorporeal Photopheresis. Front Immunol 2019; 10:547. [0261] 12. Romee R, Foley B, Lenvik T, et al. NK cell CD16 surface expression and function is regulated by a disintegrin and metalloprotease-17 (ADAM17). Blood 2013; 121:3599-608. [0262] 13. Goodier M R, Lusa C, Sherratt S, Rodriguez-Galan A, Behrens R, Riley EM. Sustained Immune Complex-Mediated Reduction in CD16 Expression after Vaccination Regulates NK Cell Function. Front Immunol 2016; 7:384. [0263] 14. Beck K E, Blansfield J A, Tran K Q, et al. Enterocolitis in patients with cancer after antibody blockade of cytotoxic T-lymphocyte-associated antigen 4. J Clin Oncol 2006; 24:2283-9. [0264] 15. Johncilla M, Misdraji J, Pratt D S, et al. Ipilimumab-associated Hepatitis: Clinicopathologic Characterization in a Series of 11 Cases. Am J Surg Pathol 2015; 39:1075-84. [0265] 16. Brummelman J, Haftmann C, Nunez N G, et al. Development, application and computational analysis of high-dimensional fluorescent antibody panels for single-cell flow cytometry. Nat Protoc 2019; 14:1946-69. [0266] 17. Team R D C. A language and Environment for Statistical Computing. R Foundation for Statistical Computing 2010. [0267] 18. Mcinnes L, Healy J, Melville J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. arXiv 6, 03426v03422 (2018). [0268] 19. Perez-Ruiz E, Minute L, Otano I, Alvarez M, et al. Prophylactic TNF blockade uncouples efficacy and toxicity in dual CTLA-4 and PD-1 immunotherapy. Nature. 2019 May 569(7756):428-432.