Fucosylation of HLA-DRB1 regulates CD4+ T cell-mediated anti-melanoma immunity and enhances immunotherapy efficacy

Immunotherapy efficacy is limited in melanoma, and combinations of immunotherapies with other modalities have yielded limited improvements but also adverse events requiring cessation of treatment. In addition to ineffective patient stratification, efficacy is impaired by paucity of intratumoral immune cells (itICs); thus, effective strategies to safely increase itICs are needed. We report that dietary administration of l-fucose induces fucosylation and cell surface enrichment of the major histocompatibility complex (MHC)-II protein HLA-DRB1 in melanoma cells, triggering CD4+ T cell-mediated increases in itICs and anti-tumor immunity, enhancing immune checkpoint blockade responses. Melanoma fucosylation and fucosylated HLA-DRB1 associate with intratumoral T cell abundance and anti-programmed cell death protein 1 (PD1) responder status in patient melanoma specimens, suggesting the potential use of melanoma fucosylation as a strategy for stratifying patients for immunotherapies. Our findings demonstrate that fucosylation is a key mediator of anti-tumor immunity and, importantly, suggest that l-fucose is a powerful agent for safely increasing itICs and immunotherapy efficacy in melanoma.

Immunotherapy efficacy is limited in melanoma, and combinations o f i mm un ot he rapies with other modalities have yielded limited improvements but also adverse events requiring cessation of treatment. In addition to ineffective patient stratification, efficacy is impaired by paucity of intratumoral immune cells (itICs); thus, effective strategies to safely increase itICs are needed. We report that dietary administration of lfucose induces fucosylation and cell surface enrichment of the major h is to co mp at ibility complex (MHC)II protein HLADRB1 in melanoma cells, triggering CD4 + T cellmediated increases in itICs and antitumor immunity, enhancing immune checkpoint blockade responses. Melanoma fucosylation and fucosylated HLADRB1 associate with intratumoral T cell abundance and antiprogrammed cell death protein 1 (PD1) responder status in patient melanoma specimens, suggesting the potential use of melanoma fucosylation as a strategy for stratifying patients for immunot herapies. Our findings demonstrate that fucosylation is a key mediator of antitumor immunity and, importantly, suggest that lfucose is a powerful agent for safely increasing itICs and immunotherapy efficacy in melanoma.
We confirmed an essential role for tumorspecific fucosylation by overexpressing murine fucokinase (mFuk) in SW1 melanoma cells to exclusively increase tumor fucosylation. mFuk expression alone suppressed tumor growth and increased total itICs comparably to oral lfuc administration alone. Again, CD4 + and CD8 + T cells were the most increased itICs (Extended Data Fig. 1l,m and Fig. 1e-h). These data indi cate that melanomaspecific fucosylation is an essential determinant of lfuctriggered itIC induction and tumor suppression, regardless of any other physiological host effects that lfuc may elicit (for example, microbiome or metabolic effects).
Correlations between tumor fucosylation and CD3 + T cells in humans were assessed by immunofluorescently analyzing a 40patient melanoma microarray. Patients with higherthanmedian tumor fuco sylation levels exhibited significantly increased intratumoral CD3 + T cell densities (Fig. 1i). Intriguingly, average melanoma fucosylation levels were lower in male patients ( Fig. 1j) but exhibited a stronger association with intratumoral CD3 + T cells (Fig. 1k).
These data indicate that melanoma fucosylation substantially shapes the itIC landscape, correlates with increased intratumoral CD3 + T cells in mice and humans and can be boosted by oral lfuc to increase itICs and suppress BRAF and NRASmutant melanomas.

l-fuc triggers CD4 ± T cell-induced itICs and alters CD4 ± T cell biology
The contribution of CD4 + and CD8 + T cells to lfuctriggered tumor suppression was assessed by immunodepletion in the SW1 model. lfuc reduced tumor growth by >50% in control and CD8 + T celldepleted mice, whereas this effect was completely abrogated by CD4 + T cell depletion ( Fig. 1l-n, immunodepletion confirmed by splenic profiling, and Extended Data Fig. 1n,o). Consistent with known roles for CD4 + T cells in recruiting and activating tumorsuppressive itICs 13 , CD4 + T cell depletion also blocked lfucinduced increases in total itICs, including intratumoral NK cells, DCs and CD8 + T cells, observed in control mice (Extended Data Fig. 1p and Fig. 1o). Similarly, in the SM1 model, CD4 + but not CD8 + T cell depletion abrogated lfuctriggered tumor suppression and increases in total itICs and itIC subpopulations (immunodepletion confirmed by splenic profiling, Extended Data Fig. 1q-w)).
Phosphoproteomic and fucosylated proteomic analyses revealed that lfuc mechanistically regulates CD4 + T cell biology by significantly altering protein kinase A (PKA) and (to a lesser extent) actin signaling, potentially via integrin B5, an upstream regulator of both of these pathways 14 that we discovered to be one of five proteins most highly bound to Aleuria aurantia lectin (AAL) (and likely fucosylated) in human peripheral blood monocyte (PBMC)derived, CD3-CD28activated CD4 + T cells, as well as Jurkat cells treated with lfuc (Extended Data Fig. 2a-f). The fact that integrin, PKA and actin signaling have been of immunotherapies. Although biomarkers of responsiveness remain under active investigation, one commonality of poor response is insuf ficient abundance and tumorsuppressive activity of itICs 4 . Therefore, elucidating itIC biology and developing safe and effective strategies to increase tumorsuppressive itICs are crucial for improving the efficacy of immunotherapies and related biomarkers.
Fucosylation, the conjugation of glycoproteins with the sugar lfucose (lfuc) at asparagine or serine-threonine residues (N or Olinked, respectively) is mediated by 13 fucosyltransferases (FUTs) and impacts protein functions that are crucial for immune and devel opmental processes 5,6 . Whereas altered fucosylation has been reported in a number of cancers, our understanding of its mechanisms and functional contributions is limited 7,8 . We previously found that global fucosylation decreases during melanoma progression, and increased tumor fucosylation levels correlate with favorable patientsurvival outcomes 9 . Furthermore, increasing melanoma fucosylation in a syngeneic mouse model reduced tumor growth and metastasis and significantly increased itICs. How fucosylation regulates antitumor immunity, however, was unknown. Here, we report that dietary lfuc can regulate the biology and interactions between CD4 + T and mela noma cells via cell surface stabilization of an MHCII protein, which robustly induces itICs and antimelanoma immunity. Tumoral MHCII protein expression, which is known to trigger CD4 + T cellmediated responses, is associated with immunemediated tumor suppression and increased responsiveness to immunotherapies 10,11 . Our findings demonstrate the ability of lfuc to improve the efficacy of immuno therapies by promoting MHCII-CD4 + T cellmediated responses and identify fucosylationbased biomarkers that may enhance patient stratification.

(Extended Data
Article https://doi.org/10.1038/s43018-022-00506-7 reported to mediate T cell activation, motility and immune synapse formation 15,16 suggests that lfuc promotes T cell trafficking to the tumor, a notion confirmed using an SW1 melanoma C3H mouse model treated with or without FTY720 (an inhibitor of lymph node egress). Inhibition of lymph node egress completely abrogated lfuctriggered tumor suppression (Fig. 2a,b). Strikingly, lfuctriggered tumor sup pression was associated with increases in intratumoral CD4 + T central and effector memory subpopulations that were abrogated by FTY720 (Fig. 2a, Table 1). Finally, lfuc also transiently but significantly increased cytotoxic CD4 + T cells at the midpoint (day 28) of the experi ment (Fig. 2d,e). These data confirmed that CD4 + T cells play a key role in induction of itICs and suppression of melanomas by lfuc, suggesting that lfuc triggers key changes in CD4 + T cell signaling and biology at the tumor and lymph node levels that are important for tumor suppression. Importantly, the fact that mFuk expression alone in melanoma cells resulted in smaller tumors with increased itICs (Fig. 1e-h) suggests that melanomaspecific fucosylated protein(s) can also promote antitumor immunity, although the mechanism was unclear.

Fucosylated HLA-DRB1 induces itICs and melanoma suppression
To identify melanoma proteins that contributed to fucosylation triggered, CD4 + T cellmediated melanoma suppression, we subjected fucosylated proteins from human melanoma cells to liquid chroma tography-mass spectrometric (LC-MS/MS) analysis followed by Inge nuity Pathway Analysis 21 (Extended Data Fig. 3a, left). These analyses identified 'Antigen presentation pathway' as the only immunerelated pathway, in which the MHCI and MHCII proteins HLAA and HLADRB1, respectively, were identified as the only antigenpresentation and plasma membrane proteins with T cellmodulating functions 22 (Extended Data Fig. 3a, right). We confirmed their expression in human melanocytes and melanoma cells by immunoblot (IB) analysis (Fig. 3a). Furthermore, lectin pulldown (LPD) using AAL and Ulex europaeus agglutinin I (UEA1) lectins, which bind to common core and terminal fucosylated glycans, respectively [23][24][25][26][27][28] , revealed association of both proteins with AAL (and to a lesser extent, UEA1), suggesting N′linked core glycosylation-fucosylation (Fig. 3b). Finally, immunoprecipita tion and IB analysis of V5tagged HLAA or HLADRB1 revealed direct recognition of HLADRB1 by AAL, indicating that a fraction of total HLADRB1 but not HLAA is directly fucosylated in melanoma (Fig. 3c).
Consistent with roles of HLADRB1 in CD4 + T cell activation [32][33][34] , our findings demonstrate that HLADRB1 is expressed and fucosylated in melanoma and required for lfuctriggered CD4 + T cellmediated itIC induction and melanoma suppression.

Fucosylation regulates HLA-DRB1 localization and immunological effects
We reasoned that determining how HLADRB1 is regulated by fuco sylation would provide important insight into its crucial role in lfuctriggered antitumor immunity. Using NetNGlyc version 1 and NetOGlyc version 4 (https://services.healthtech.dtu.dk) 35 , we pre dicted N and Olinked glycosylation sites at Asn48 (N48) and Thr129 (T129), respectively, which are conserved sites within constant regions of human and mouse HLADRB1 (Fig. 4a, top) 29,36 . Importantly, EB1 exhibits ~80% sequence homology with HLADRB1 and contains the conserved glycosylation-fucosylation site at N46 (ref. 29 ). Modeling of HLADRB1 interactions with prominent binding partners HLADM or CD4-TCR suggests that fucosylation of neither site affects interac tion interfaces or peptide loading or presentation (Fig. 4a, bottom).
To determine how fucosylation might regulate HLADRB1, we assessed its subcellular localization in WM793 cells that were pharmacologically modulated for fucosylation by treatment with 2Fperacetylfucose (a FUT inhibitor (FUTi) 40 ) versus vehicle (dimethyl sulfoxide, DMSO; control). Cells treated with FUTi exhibited dimmer, more central, endoplasmic reticulum (ER)colocalization of HLADRB1 than vehicletreated cells, suggesting less accumulation at the cell surface (Fig. 4d). Furthermore, flow cytometry revealed that cell sur face fucosylation and HLADRB1 both decreased or increased after FUTi or lfuc treatments, respectively, whereas mRNA and protein levels remained unchanged; thus fucosylation promotes cell surface localization of HLADRB1 ( Fig. 4e and Extended Data Fig. 4b). Finally, global proteomic profiling to identify interactors that might mediate fucosylationregulated cell surface localization of HLADRB1 revealed that N48 glycosylation-fucosylation promotes binding to calnexin, which has been reported to mediate maturation and trafficking of MHCII complexes to the surface 41 (Extended Data Fig. 5a-d).
To assess how HLADRB1 glycosylation-fucosylation con tributes to tumor suppression and itICs, we compared control or EB1knockeddown SW1 tumors reconstituted with WT or glycofuco mutant (N46G) EB1 (confirmation of knockdown reconstitution and fucosylation by IB and LPD, respectively, in Extended Data Fig. 5e).  Abrogation of lfucinduced itIC and tumor growth suppression by EB1 knockdown was rescued by reconstitution with only WT but not glycofucomutant EB1, demonstrating that glycosylation-fucosyla tion of EB1 or HLADRB1 is essential for lfuctriggered itIC induction and melanoma suppression ( Fig. 4f and Extended Data Fig. 5f,g). This is consistent with our finding that loss of glycosylation-fucosylation of HLADRB1 or EB1 abrogates its cell surface localization and impairs its ability to induce antitumor immunity. Thus, despite the other fucosylated proteins identified in melanoma cells (Extended Data Fig. 3), these data confirm that the N48 glycosylation-fucosylation of HLADRB1 is a key regulator of antimelanoma immunity and tumor suppression. Despite other potential host physiological effects of dietary lfuc (for example, microbiome, metabolome, etc.), these data confirm that lfucinduced itIC increases and melanoma suppression are critically mediated by melanomaintrinsic expression and fucosyla tion of HLADRB1, which promotes its cell surface accumulation to trigger CD4 + T cellmediated antitumor immune responses.

Oral l-fuc augments anti-PD1-mediated melanoma suppression
Expression of MHCII reportedly correlates with increased antiPD1 efficacy 42,43 . Indeed, patients who failed antiPD1 therapy exhibited relative >45% reduced cell surface MHCII but not MHCI (Extended Data Fig. 5h). As antiPD1 efficacy can be limited by itIC abundance 44 , par ticularly of CD4 + T and memory CD4 + T cells 10,45-49 , we tested whether the ability to increase CD4 + T cellmediated itIC induction and tumor suppression using oral lfuc could be leveraged to augment antiPD1 efficacy. In the SW1 model, oral lfuc suppressed tumors as much as antiPD1 but did not enhance efficacy of antiPD1 therapy (~50-60%; Fig. 5a, left). By contrast, in the SM1 model, lfuc was less tumor suppres sive than antiPD1 therapy alone but rather augmented durable suppres sion in combination with antiPD1 therapy (Fig. 5a, right). Importantly, we also found that lfuc does not alter cell surface levels of programmed cell death ligand 1 (PDL1) in mouse or human melanoma cells (Extended Data Fig. 6a), suggesting that the lfucassociated tumor suppression in these models is attributed to determinants beyond the PD1-PDL1 axis.
To clarify how the combination of lfuc and antiPD1 therapy enhanced suppression, we characterized immune cell profiles in the tumors and lymph nodes of SM1 tumorbearing mice over a time course of treatment with lfuc, with or without antiPD1 therapy. Adminis tration of lfuc (1) alone increased intratumoral CD4 + T central and effector memory cells, an effect that was increased when combined with antiPD1 therapy (Fig. 5b (blue dashed boxes) and Supplemen tary Table 2) and (2) initially expanded intratumoral cDC2 cells, fol lowed by later expansion of cDC2 cells and moDCs in the lymph nodes when combined with antiPD1 therapy (Fig. 5b (orange dashed boxes) and Supplementary Table 2). In addition to expanding the absolute numbers of intratumoral CD4 + and CD8 + T cells at the endpoint (day 63), the combination of lfuc and antiPD1 therapy increased the rela tive percentage of intratumoral CD8 + T central memory cells (Fig. 5b (green dashed box) and Supplementary Table 2). Thus, lfuc can sup press some melanomas as effectively as antiPD1 therapy, whereas, in others, it can enhance efficacy, which is associated with increased intratumoral CD4 + T central and effector memory subpopulations and lymph node cDC2 and moDC populations, consistent with the effects of lfuc observed in Fig. 2.

Clinical implications of tumor fucosylation and fucosylated HLA-DRB1
Given the potent enhancement of antiPD1 efficacy by oral lfuc admin istration in mice, we investigated whether tumor fucosylation or total/ fucosylated HLADRB1 might correlate at all with responsiveness to antiPD1 therapy in human patient biopsies, as the identification of preliminary correlations might support their subsequent development into predictive biomarkers for antiPD1 responsiveness. To this end, we devised a new technique: we modified the conventional proximity ligation assay (PLA) 50 to facilitate immunofluorescent visualization of fucosylated HLADRB1 by applying antiHLADRB1 antibody together with biotinylated AAL, which has previously been successfully used to stain tissues specifically for corefucosylated glycans 51 (Fig. 6a). This technique, lectinmediated PLA (LPLA), revealed cytoplasmic and/or membranous localization of endogenous fucosylated HLADRB1 in mel anoma cells (Fig. 6b) that is lost upon FUTi treatment (Fig. 6c), confirm ing lfucstimulated cell surface localization of HLADRB1 (Fig. 4d,e and Extended Data Fig. 4b). The cytoplasmic and/or 'vesicularappearing' staining is consistent with HLADRB1 that was fucosylated in the endo plasmic reticulum (ER)-Golgi and is en route to the surface via the secretory pathway. In applying this technique further to formalinfixed, paraffinembedded (FFPE) melanoma tissue specimens, we observed similar staining patterns for fucosylated HLADRB1 (Fig. 6d,e), which were completely abolished by washing the tissue with lfuc, confirming specificity for fucosylated HLADRB1 (Fig. 6f).
To assess correlations of (1) tumorspecific fucosylation and total/ fucosylated HLADRB1 of individual tumor cells and (2) intratumoral numbers CD4 + T cells with responder status to singleagent antiPD1 therapy, we implemented LPLA on primary melanoma biopsies from two distinct responder and two nonresponder patients followed by singlecell segmented signal quantitation (Fig. 7a,b). Tumors of responders clearly contained tumor cell populations with high levels of fucosylation and total HLADRB1 as compared to nonresponders ( Fig. 7b(i,ii)). Although the tumor of only one of two responders con tained melanoma cells with increased levels of fucosylated HLADRB1 compared with those of the nonresponders (Fig. 7b(iii)), this trend mirrored that of intratumoral CD4 + T cell counts ( Fig. 7b(iv)), consist ent with the role for fucosylated HLADRB1 in CD4 + T cellmediated tumor suppression.
We assessed potential associations between tumor fucosylation, total/fucosylated HLADRB1, CD4 + T cells and responder status in expanded cohorts of patients with melanoma treated with antiPD1 therapy. Levels of tumor fucosylation and total and fucosylated

| Administration of combination l-fuc and anti-PD1 therapy suppresses tumors and increases intratumoral CD4 + T central and effector memory cells. a,
Volumetric growth curves for SW1 tumors in C3H/HeN mice (left) and SM1 tumors in C57BL/6 mice (right) fed with or without lfuc and treated with PBS (control) or antiPD1 therapy (concurrent initiation of lfuc with or without antiPD1 therapy (red triangle)). The tumor growth curves show mean ± s.e.m. per mouse group. For each group, n = 7 mice except PBS with lfuc and antiPD1 therapy with lfuc groups, which each have eight mice. b, Volumetric growth curves for SM1 tumors in C57BL/6 mice fed with or without lfuc and treated with PBS (control) or antiPD1 therapy (PD1) (concurrent initiation of lfuc with or without antiPD1 therapy (red triangle)). The tumor growth curves show mean ± s.e.m. from ≥7 mice per group. At day 7 (before administration of lfuc or antiPD1 therapy, n = 3 mice), day 21 (endpoint for tumors of controltreated mice, n = 5 mice per group), day 31 (endpoint for tumors of lfuctreated mice, n = 5 mice per group) and day 63 (endpoint for tumors of antiPD1treated mice, n = 5 mice per group), the primary tumors (tumor) and draining lymph nodes of 4-5 mice per treatment group were analyzed by flow cytometry for intratumor levels of CD4 + and CD8 + T subpopulations (naive or terminal; stem central, central or effector memory) and DC subpopulations (cDC1, cDC2 and moDC) as in Fig. 2. Proportions of CD4 + , CD8 + and DC subpopulations in each organ at each time point are represented by the colorcoded pie charts (each pie chart represents 4-5 mice). Absolute numbers of the subpopulations per 10 6 cells of tumor or tissue homogenate at each time point are represented in the color coded column charts. Corresponding raw flow cytometric data for these charts are shown in Supplementary      HLADRB1 in tumor cells were generally higher in antiPD1 responders than in nonresponders from Massachusetts General Hospital (MGH; n = 31; Fig. 7c, top) and the MD Anderson Cancer Center (MDACC; n = 11; Fig. 7c, bottom). Total tumor fucosylated HLADRB1 exhibited weak or no association with tumoral CD4 + T cells (Fig. 7d, top and bottom), although the association was modestly increased when restricted to CD4 + T cells localized at the periphery of the tumors (Extended Data  Fig. 6b,c; absolute CD4 + T numbers in Supplementary Table 3). Importantly, we found that some specimens containing divergent tumor-stroma content did exhibit divergent correlation strengths (for example, core biopsies containing only tumor versus noncore biop sies containing substantial stroma). For example, a 'highly correlated' antiPD1 responder (noncore biopsy) containing substantial tumorstromal interface exhibited correlated high levels of fucosylated HLADRB1 and CD4 + T cells, whereas a 'noncorrelated' responder (a core biopsy) did not (Extended Data Fig. 6d), suggesting that variable stromal content within biopsies may have at least partially undermined the strength of the correlations that we assessed. The lack of significant correlation may also be attributed to the dynamic relationship between fucosylated HLADRB1 and CD4 + T cell infiltration that is further weakened by suboptimal inclusion criteria and/or patient stratification. Comparison of these markers in five patientmatched tumors before and after antiPD1 therapy revealed no significant correlation in total HLADRB1 levels. However, before treat ment, tumor cell fucosylation was significantly higher in the complete responder versus partial responders and nonresponders; this dropped to the equivalently lower levels of the other patients after treatment.
With the exception of one nonresponder, the complete responder also exhibited significantly increased fucosylated HLADRB1 in tumor cells before treatment (Extended Data Fig. 6e). The consistent trends in tumor fucosylation and fucosylated tumor HLADRB1 observed across the three independent cancer center cohorts appear to support potential utility but importantly point to the need for further study in expanded patient pretreatment biopsy cohorts that are controlled for a number of specific clinical variables, which will be discussed below.

Discussion
Here, we report the administration of a dietary sugar as a way to increase itICs and enhance efficacy of the immune checkpoint blockade antiPD1 agent. These studies reveal insights into the posttranslational regu lation and immunological roles of melanoma cellexpressed MHCII proteins, further highlighting their relationship with itICs 42,43,[45][46][47][48][49] . Specifically, fucosylation regulates the cell surface abundance of HLADRB1, which triggers robust CD4 + T cellmediated itIC induc tion and melanoma suppression. It is important to acknowledge that our reliance on AAL lectin predominantly focuses our study on α1,6fucosylated proteins. Although this does not diminish the crucial role that α1,6fucosylated HLADRB1, which was identified as fucosylated via lectinagnostic click chemistry mass spectrometric screening, plays in lfuctriggered antitumor immune responses, it is possible that proteins with other fucosylation linkages might contrib ute to aspects of antitumor immunity. It is also likely that the statistical strength of our analyses of tumor fucosylation with patient outcomes (Fig. 7) was limited by use of only AAL lectin, which precludes the detec tion of other structural forms of fucosylation. Nonetheless, the abil ity to leverage this mechanism using oral lfuc administration may help to enhance other immunotherapeutic modalities (that is, other checkpoint inhibitors or adoptive celltransfer therapies). Notably, as a nontoxic dietary sugar with a past safety precedent as an experi mental therapy for children with leukocyte adhesion deficiency II 52,53 , lfuc appears to be a potentially safe and tolerable therapeutic agent.
The consistent trends that we observed in higher tumor fucosyla tion and fucosylated HLADRB1 across antiPD1 responders versus nonresponders between the three independent cancer center cohorts support their potential utility as biomarkers of antiPD1 responsive ness. However, further analyses in expanded patient biopsy cohorts are clearly needed. Considering the variable tumorsuppressive effects of lfuc observed in our antiPD1treated SM1 and SM1 mouse models, there are likely similar biological and clinical variables in patients that must be further explored and that may have precluded statistical sig nificance in our small analyses.
In terms of biological variables, how T cell biology is regulated by fucosylation, for example, has heretofore been unclear. Reported divergent effects of fucosylation on T cell activation versus exhaus tion (that is, via regulation of PDL1 expression) point to FUTspecific expression and roles that remain to be elucidated [54][55][56] . The fact that lfuc does not alter the cell surface levels of PDL1 in human or mouse melanoma cells (Extended Data Fig. 6a), suggesting that the discrep ant tumor suppression by singleagent versus combination lfuc and antiPD1 therapy in our SW1 and SM1 mouse models (Fig. 5a), is attrib uted to determinants beyond the PD1-PDL1 axis. Indeed, our global fucosylated and phosphoproteomic analyses suggest that fucosylation in CD4 + T cells impacts integrin β5, PKA and actin signaling (Extended Data Fig. 2) and that this is associated with increased intratumoral T cell presence and memory phenotypes in our models (Figs. 2 and 5), consistent with previous reports that those functions are regulated by those pathways in T cell biology [15][16][17]57 . The fact that lfuc can increase CD4 + T central memory cells also partially explains how it can augment antiPD1 efficacy, which is associated with the presence of these cells 10 . How lfuc may regulate these signaling pathways and enrich for CD4 + T memory subsets within the tumor microenvironment and, further more, how lfuc alters DC biology and induces their intratumoral accumulation (Figs. 2 and 5) may contribute to antitumor immune responses and tumor suppression in this context are unclear and war rant further lines of study. In addition, sex might be a determinant, as melanoma fucosylation levels are lower but correlate more strongly with intratumoral CD3 + T cells in male versus female patients (Fig. 1j,k). Reduced melanoma fucosylation, which is expected to lower itICs, might explain increased lethality in male patients (American Cancer Society Facts and Figures, 2022).
The availability of pretreatment antiPD1 tumor tissue specimens for this study was extremely limited. Thus, the specimens that we acquired were subject to clinical variability that may have undermined statistical robustness in our analyses. Subsequent studies investigating tumor fucosylation, total/fucosylated HLADRB1 and CD4 + T cells as biomarkers will need to factor for clinical variables including therapies received before antiPD1 therapy and preexisting medical conditions as well as time from biopsy to antiPD1 treatment. Because we were unable to control for these confounders in our specimens, it is unclear how they may have impacted tumor and HLADRB1 fucosylation and CD4 + T cell biology and thus the strength of correlations between these markers and responsiveness to treatment. Likewise, the importance and contribution of the immune environment of the peritumoral stroma in this setting remains to be elucidated, as some of our biopsies contained stroma, whereas other tumor core biopsies did not. Prior studies focusing on tumor-immune interactions and immunothera pies (including antiPD1 therapy) have highlighted the importance of analyzing biomarker staining patterns at tumor-immune and tumorstromal interfaces contained within biopsies, as these are areas of enriched immunological activity and signaling [58][59][60] . Indeed, our obser vation of 'highly correlated' and 'noncorrelated' antiPD1 responder biopsies containing disparate amounts of tumor stroma highlight how lack of sufficient stroma in tumor biopsies likely undermined the statistical robustness of the correlations between fucosylated HLADRB1 and CD4 + T cells (Extended Data Fig. 6d). The acquisition of such biopsy specimens that are controlled for the variables detailed above is an important consideration for subsequent studies. Our find ings highlight the need for a prospective clinical trial with defined protocols for collection of monotherapy antiPD1 pretreatment biopsies at defined time points proximal to therapy and clear biopsy protocols to yield tumor specimens that contain substantial intact stromal interface.
In conclusion, fucosylation of HLADRB1 is a key regulator of itIC abundance in melanomas, and this mechanism, together with fucosylationregulated CD4 + T cell biology, can be therapeutically exploited using oral lfuc administration. Elucidation of the mecha nistic determinants is expected to advance our understanding of the immunobiology of melanoma and other cancers and to inform efforts in implementing fucosylation and/or fucosylated HLADRB1 as bio markers and of lfuc as a therapeutic agent. Article https://doi.org/10.1038/s43018-022-00506-7

Methods
Our research complies with all relevant ethical regulations: all ani mal experiments were approved by the Moffitt IACUC committee. All patient specimenstaining analyses are considered as IRB exempt: investigators were blinded from all patient health information, and the specimens were previously collected under IRBapproved protocols per respective institutional IRB committees. All cell line and antibody information is provided in the Nature Portfolio Reporting Summary.

General cell culture
The following cell lines were from the American Tissue Type Collection: A375, HEMN (normal adult epidermal melanocytes) and Jurkat. The following cell lines were from Rockland Immunochemicals: WM983A, WM983B, WM1366, WM115, WM2664, WM164, WM793 and Lu1205. SW1 melanoma cells (gift from the Ronai laboratory at the Sanford Burnham Prebys Medical Discovery Institute) and SM1 melanoma cells (gift from the Smalley laboratory at the Moffitt Cancer Center) were cultured in DMEM containing 10% FBS, 1 g ml −1 glucose and 4 mM lglutamine at 37 °C with 5% CO 2 . HEMN cells were grown in Lonza MGM4 growth medium; before collection for IB analysis, the cells were switched to the same medium as the other cells overnight. Cell lines were transfected using Lipofectamine 2000 (Invitrogen). Primary CD4 + T cells were collected using the EasySep Human CD4 + negative selection isolation kit (Stemcell Technologies) according to the manufacturer's protocols. Upon arrival at the laboratory, all cell lines are quarantined until they have passed footprint identification and mycoplasma testing (as mycoplasma negative). The identities of all cell lines (human and mouse) in the Lau laboratory are verified annually by short tandem repeatbased authentication 'CellCheck' services provided through IDEXX BioResearch.

Cloning and mutagenesis
The gene encoding mFuk was cloned using cDNA from SW1 cells into the pLentiCMycDDKIRESPuro expression vector (Ori Gene Technologies) using BamHI and NheI restriction sites. Mouse EB1 constructs were cloned using cDNA from SW1 cells into the pLentiCMycDDKIRESPuro expression vector (OriGene Technolo gies) using AscI and XhoI restriction sites. pLKO shNT, pLKO shK11, pLKO shK12, pLKO shEB11 and pLKO shEB12 were obtained from MilliporeSigma. pLX304::EV was obtained from OriGene Technologies. pLX304::HLAA and pLX304::HLADRB1 constructs were obtained from DNAasu 61 . HLADRB1 N48G and HLADRB1 T129A as well as EB1 N46G mutants were generated using the QuikChange II XL sitedirected mutagenesis kit according to the manufacturer's protocol (Agilent Technologies).

Lectin pulldown
Control beads and AAL or UEA1 lectinconjugated agarose beads were preblocked for 2 h in blocking buffer (2% IgGfree BSA ( Jackson Immu noResearch Laboratories)) on a rotator at 4 °C. Cells were lysed on ice in 1% Triton X100 lysis buffer (1% Triton X100, 20 mM TrisHCl, pH 7.4, 150 mM NaCl in ddH 2 O with protease and phosphatase inhibitors (Thermo Fisher Scientific)), briefly sonicated and pelleted. The result ing lysates were normalized in protein concentration to the sample with the lowest concentration, diluted to a final concentration of 0.25% Triton X100 with dilution buffer (0% Triton X100, 20 mM TrisHCl, pH 7.4, 150 mM NaCl in ddH 2 O with protease and phosphatase inhibitors), incubated with 15 µl preblocked beads (beads were centrifuged out of a block and resuspended in dilution buffer) and rotated overnight at 4 °C. Next, the beads were washed twice with dilution buffer and subjected to (12%) SDS-PAGE and IB analysis using the indicated antibodies.

Mass spectrometric analyses
Profiling fucosylated proteins. EV, pLentiFUKGFP or shFUKexpressing WM793 cells were cultured in biological triplicate in the presence of 50 µM lfucalkyne for ~72 h to ~80% confluence. Cells were lysed in 1.5% Ndodecylβdmaltoside, 20 mM HEPES, pH 7.4, and protease and phosphatase inhibitors. Lysates were precipi tated with acetone overnight, and pelleted proteins were resuspended and subjected to click chemistry labeling with biotin-azide using the ClickiT kit according to the manufacturer's protocol (Invitrogen). The negative control included EV cells not labeled with lfucalkyne. All biotinylated-fucosylated samples were pulled down using neu travidin beads (preblocked with 2% IgGfree BSA). Samples were sub mitted to the Sanford Burnham Prebys proteomics core facility for onbead digestion and LC-MS/MS analysis. Hits that were increased by >1.5 fold in pLentiFUKGFPexpressing cells and unchanged or decreased in pLentiEVGFPexpressing cells or decreased in pLentishFUKexpressing cells were subjected to Ingenuity Pathway Analysis (Qiagen).
Profiling HLA-DRB1 glycosylation-fucosylation. Stained bands (~1 µg) of exogenously expressed V5-HLADRB1 purified from WM793 cells were minced, reduced and alkylated using 20 mM TCEP (Tris(2carboxyethyl)phosphine) and iodoacetamide in 50 mM TrisHCl. Gel pieces were washed with 20 mM ammonium phosphate in 50% methanol overnight at 4 °C, followed by dehydrating for 30 min with 100% acetonitrile. Samples were next digested with trypsin for 4 h at 37 °C and eluted through C 18 ZipTips with 50% methanol and 0.1% formic acid (FA). Five microliters of the elution were diluted in 0.1% FA and injected into a Q Exactive Orbitrap mass spectrometer equipped with an Easy NanoLC HPLC system and a reversephase column (Thermo Fisher Scientific). A binary gradient solvent system consisting of 0.1% FA in water (solvent A) and 90% acetonitrile with 0.1% FA in water (solvent B) was used to separate peptides. Raw data were analyzed using both Proteome Discoverer version 2.1 (Thermo Fisher Scientific) with the Byonic (Protein Metrics) module and Byonic stan dalone version 2.10.5. All extracted ion chromatograms were generated using Xcalibur Qual Browser version 4.0 (Thermo Fisher Scientific). The UniProt sequence Q5Y7D1_Human was used as the reference sequence for peptide analysis.
Phosphoproteomic profiling of CD4 ± T cells. The indicated CD4 + T cells were lysed in RIPA buffer with protease and phosphatase inhibi tors. Briefly, ~1 mg of each lysate was reduced with 4.5 mM dithiothreitol for 30 min at 60 °C, alkylated with 10 mM iodoacetamide at room temperature in the dark for 20 min and digested overnight at 37 °C with an enzymetoprotein ratio of trypsin of 1:20 (Worthington). Resulting peptides were desalted using a reversedphase SepPak C 18 cartridge (Waters) and lyophilized for 48 h. Lyophilized peptides were enriched for global phosphopeptides (pSTY) using IMAC FeNTA magnetic beads (Cell Signaling Technology) on a KingFisher Flex Purification System (Thermo Fisher Scientific), followed by SpeedVac concentra tion and resuspension in loading buffer (5% ACN and 0.1% TFA) before autosampling and LC-MS/MS as described below.
Fucoproteomic profiling of CD4 ± T cells. The indicated CD4 + T cells were lysed in RIPA buffer with protease and phosphatase inhibitors and subjected to LPD using control or AAL beads. The beads were washed with PBS and subjected to onbead trypsin digestion. Resulting pep tides were further denatured with 30 mM ammonium bicarbonate at 95 °C for 5 min and then acidified with TFA at a final concentration of 1%. ZipTippurified, eluted peptides were concentrated with a Speed Vac and resuspended in loading buffer (5% ACN and 0.1% TFA) before autosampling and LC-MS/MS as described below.
Profiling HLA-DRB1 interactors. V5tagged WT or N48G glycofucomu tant HLADRB1expressing WM793 cells were lysed and subjected to V5 bead pulldown. Five percent of pulled down protein was immunoblot ted to ensure equal sample submission for processing as described above and LC-MS/MS as described below.

Liquid chromatography-mass spectrometry
Tryptic peptides were analyzed using a nanoflow ultrahigh performance liquid chromatograph and an electrospray Orbitrap mass spectrometer (RSLCnano and Q Exactive Plus, Thermo) for tan dem MS peptide sequencing. Peptide mixtures were loaded onto a precolumn (100µm ID × 2cm column packed with C 18 reversedphase resin; particle size, 5 µm; pore size, 100 Å) and washed for 5 min with aqueous 2% acetonitrile and 0.1% FA. Solvent A comprised 98% ddH 2 O, 2% acetonitrile and 0.1% FA, and solvent B comprised 90% acetonitrile, 10% ddH 2 O and 0.1% FA. Trapped peptides were eluted or separated on a C 18 analytical column (75µm ID × 50 cm; particle size, 2 µm; pore size, 100 Å; Thermo Fisher Scientific) using a 90min gradient at a flow rate of 300 nl min −1 of 2% to 3% solvent B over 5 min, 3% to 30% solvent B over 27 min, 30% to 38.5% solvent B over 5 min, 38.5% to 90% solvent B over 3 min and then held at 90% for 3 min, followed by 90% to 2% solvent B in 1 min and reequilibrated for 18 min. MS resolution was set at 70,000, and MS/MS resolution was set at 17,500 with a maximum IT of 50 ms. The top 16 tandem mass spectra were collected using datadependent acquisition following each survey scan and 60s exclu sion for previously sampled peptide peaks. For phosphoproteomic, fucoproteomic and HLADRB1 WT and glycofucomutant interactor profiling, MaxQuant 62 software (version 1.6.2.10) was used to identify and/or quantify phosphopeptides and proteins for the datadependent acquisition runs. The false discovery rate was set to 1%.

Human donor peripheral CD4 ± T cell-isolation protocol
Human CD4 + T cells were (1) isolated from fresh PBMCs using a CD4 + T cell negative selection isolation kit (Stemcell Technologies) accord ing to manufacturer's protocols, (2) cultured in the presence of vehicle or 250 µM lfuc and (3) activated using antiCD3/CD28 Dynabeads (Thermo Fisher Scientific) at a bead:CD4 + T cell ratio of 1:1. After 48 h, cell pellets were collected and lysed for fucoproteomic or phospho proteomic profiling.

Flow cytometry
Gating schemes are provided in the Supplementary Information. Unless otherwise indicated, cytometry was performed using an LSR Flow Cytometer (BD Biosciences), and analysis was performed using FACSDiva version 9, CellQuest version 6 and FlowJo version 9 software (BD Biosciences).

Assessment of cell surface fucosylation, HLA-DRB1 and PD-L1.
The indicated cells were treated for 72 h with DMSO, 250 µM FUTi (MilliporeSigma) or 250 µM lfuc (Biosynth), followed by staining with 0.1 µM PKH26 (MilliporeSigma) before fixation in 4% formal dehyde solution. Cells were stained with antiHLADRB1 antibody and fluorescein AAL or antihuman or antimouse PDL1 antibodies overnight, followed by three washes and staining with Alexa Fluor 594 donkey antirabbit antibody. Cells were washed and subjected to flow cytometric analyses using a FACSCalibur (BD Biosciences) as in Supplementary Fig. 4.

Assessment of cell surface pan-MHC-I and pan-MHC-II.
Surgi cally resected patient tumors were minced to fragments less than 1 mm and digested in 1× tumor digest buffer. Singlecell suspensions were strained through 40µm nylon mesh and counted for viability by trypan blue exclusion. Strained tumor homogenates were stained using Live/Dead Zombie NIR, PE antipanMHCI (HLAA-HLAC), FITC antipanMHCII, PerPCy5.5 antiCD45, APC antiCD90 and BV421 antiEpCAM antibodies and subjected to flow cytometric analysis as in Supplementary Fig. 5.

Immunoblot analyses
Cells were lysed on ice in standard RIPA buffer (25 mM TrisHCl, pH 7.6, 150 mM NaCl, 5 mM EDTA, 1% NP40 or 1% Triton X100, 1% sodium deoxycholate, 0.1% SDS in diH 2 O with protease and phosphatase inhibi tors), sonicated and pelleted, and the resulting lysates were normalized by protein concentration using the DC assay (BioRad Laboratories). Lysates were subjected to (12%) SDS-PAGE and IB using the indicated antibodies. IB imaging and analysis was performed using either an Odyssey FC scanner and Image Studio (LICOR Biosciences) or film.

Quantitative PCR with reverse transcription
RNA from the indicated cells was extracted using the GenElute Mam malian Total RNA extraction system (MilliporeSigma) and reversed transcribed using the HighCapacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). RT-qPCR was performed using FastStart Universal SYBR Green Master Mix (Rox) (Roche Diagnostics) using CFX Manager version 3.1 on a BioRad CFX96 RealTime system (BioRad Laboratories). RT-qPCR cycles were as follows: 95 °C for 10 min, 35 cycles of 95 °C for 15 s, 55 °C for 60 s and 72 °C for 30 s. Gene expres sion was normalized to histone H3A expression. Primers for RT-qPCR were generated using NCBI Primer BLAST software (National Center for Biotechnology Information). Oligonucleotide sequences are provided in Supplementary Table 4.

Fluorescent immunocytochemical and immunohistological staining and analysis
General immunofluorescent cell staining. Cells were grown on German glass coverslips (Electron Microscopy Services) and fixed in fixation buffer (4% formaldehyde, 2% sucrose in PBS) for 20 min at room temperature. Cells were washed with PBS, permeabilized in permeabilization buffer (0.4% Triton X100 and 0.4% IgGfree BSA Article https://doi.org/10.1038/s43018-022-00506-7 ( Jackson ImmunoResearch Laboratories) in PBS for 20 min at room temperature, washed with PBS again and incubated with the indicated antibodies. Unless otherwise indicated, images were acquired using a Keyence BZX710 microscope and processed and analyzed using ImageJ version 1.53a (NIH).
General immunofluorescent tissue staining. FFPE tumor sec tions (or tumor microarray (TMA) slides) were melted at 70 °C for 30 min, deparaffinized using xylene and rehydrated in serial alcohol washes. The slides were pressure cooked at 15 psi for 15 min in 1× DAKO antigenretrieval buffer (Agilent Technologies). Slides were subjected to two 5min standing washes in PBS before blocking in 1× CarboFree Blocking Solution (Vector Laboratories) for 2-3 h, followed by two more washes and staining with the indicated lectin and/or antibodies before washing and mounting with VECTASHIELD and DAPI (Vector Laboratories). Fig. 1a,d,k, tumors were immunostained with FITCconjugated AAL lectin (0.4 µg ml −1 ) and rabbit antiMART1 and rabbit antiS100 antibod ies. Images were acquired using a Keyence BZX710 microscope, and images were processed and analyzed using ImageJ version 1.53a (NIH) as follows: melanoma markerpositive regions were assigned as regions of interest in which we measured the integrated density of the AAL signal. Integrated densities of control tumors were assigned as 1, and integrated AAL density values of experimental tumors were divided by control values to produce relative fold changes and plotted as column charts. Fig. 1i-k, melanoma TMA (ME1002b; US Biomax; 18 female and 22 male patients, ages 25-88 years, stages 1A-4, cutaneous and mucosal tumors) was immunostained with FITCconjugated AAL lectin (0.4 µg ml −1 ), rabbit antiMART1, rabbit antiS100 and antiCD3 antibodies, followed by Alexa Fluor 568 (Cy3) donkey antirabbit and Alexa Fluor 647 (Cy5) donkey antimouse sec ondary antibodies. Slides were mounted with VECTASHIELD and DAPI. An Aperio ScanScope FL scanner (Leica Biosystems) was used to scan the TMA at 20×. Definiens Tissue Studio version 4.7 (Definiens) was used to identify individual cores followed by singlecell segmentation. Mean fluorescence intensity (MFI) values for fucosylation, melanoma markers and CD3 channels were extracted from each segmented cell; minimum MFI thresholds were set to enumerate melanoma and CD3 + T cells. Average MFI values for each core were reported for fucosylation and melanoma marker channels.

Melanoma TMA analysis. For
We used nonparametric Wilcoxon ranksum test to compare melanomaspecific fucosylation levels between CD3 + T cell high ver sus low groups. Density values of CD3 + T cells were all log 2 transformed in the statistical analysis. Multivariable linear regression was used to assess association between fucosylation and T cells while adjusting for confounding factors including sex, age and stage. The Spearman correlation coefficient was used to measure the correlation between melanomaspecific fucosylation and T cells in different sex groups.
Lectin-mediated proximity ligation assay. Coverslipgrown cells or FFPE tumor sections subjected to LPLA were processed upfront as described above. Both approaches used mouse antiHLADRB1 anti body (0.2 µg ml −1 ) and biotinylated AAL lectin (0.2 µg ml −1 ), staining overnight at 4 °C. Coverslips and/or slides were washed twice with PBS and incubated with phalloidin Alexa Fluor 488 with goat antibiotin antibody for 2 h at 4 °C. Subsequent steps of the protocol were adapted from the Duolink In Situ Green PLA kit's manufacturer's protocol (Mil liporeSigma). PLA antigoat MINUS and PLA antimouse PLUS probes were applied at 1:5 for 1 h at 37 °C. The coverslips and/or slides were washed twice with wash buffer A before ligation with 1:5 ligation buffer and 1:40 ligase in ddH 2 O for 30 min at 37 °C. Coverslips and/or slides were washed twice with wash buffer A before incubation in amplifica tion mix (1:5 amplification buffer and 1:80 polymerase in ddH 2 O for 1.5 h at 37 °C). Coverslips and/or slides were washed twice with wash buffer B before mounting with VECTASHIELD and DAPI. For Fig. 5c, the indicated FFPE sections were immunostained with antiHLADRB1 antibody or by LPLA as detailed above with the addition of antiCD4 antibody. WTS imaging was per formed using the Vectra 3 Automated Quantitative Pathology Imaging System (PerkinElmer). Tiles (20× region of interest) were sequentially scanned across the slide and spectrally unmixed using inForm (Perki nElmer). HALO (Indica Labs) was used to fuse tile images together. For each wholetumor image, (1) every individual melanoma marker (MART1 and S100)positive cell was segmented and quantitatively measured for total fucosylation and total and fucosylated HLADRB1, and (2) every CD4 + T cell within the melanoma markerpositive tissue region and melanoma markernegative periphery was counted. For each patient, marker values were displayed in box plots to visualize staining distribution of individual tumor cells. The total numbers of melanoma cells per patient section measured and analyzed were as follows: patient 1, 557,146 cells; patient 2, 743,172 cells; patient 3, 95,628 cells; and patient 4, 13,423 cells.

Anti-PD1-treated patient specimens
Moffitt Cancer Center patient specimens. For Fig. 5d,e and Extended Data Fig. 4d,e, deidentified specimens from MCC patients with advanced stage melanoma were collected and analyzed following patient consent under MCC Institutional Review Board (IRB)approved protocols: For Fig. 7b, 'responder' patients exhibited >20 months of progression free survival, whereas 'nonresponder' patients progressed after <6 months after receiving antiPD1 therapy.
For Extended Data Fig. 5h, nonresponse status to PD1 checkpoint blockade (nivolumab or pembrolizumab) was defined by RECIST 1.1 as disease progression on therapy or within 3 months of the last dose.

University of Texas MD Anderson Cancer Center patient specimens.
Biospecimens were retrieved, collected and analyzed after patient consent under University of Texas MDACC IRBapproved protocols. University of Texas MDACC patients with advanced (stages III-IV) melanoma between 1 July 2015 and 1 May 2020 who received >1 dose of PD1 checkpoint blockade agent (nivolumab or pembrolizumab) were identified from detailed review of clinic records. Responders or nonresponders were defined as those with a complete or partial response or stable or progressive disease, respectively, by RECIST 1.1. Pathologic response was defined by the presence or absence of viable tumors on pathologic review when available.
Massachusetts General Hospital patient specimens. Patients ini tiating antiPD1 therapy (pembrolizumab) as frontline treatment for metastatic melanoma at MGH provided written informed consent for the collection of tissue and blood samples for research (DF/HCC IRBapproved protocol 11181). Responders showed radiographic decrease in disease at initial staging for ≥12 weeks. Nonresponders did not exhibit radiographic response and/or had rapid progression. Progressionfree survival was defined as days from treatment start to radiographic scan when progression was first noted (uncensored) or last progressionfree scan (censored). Overall survival was defined as days from treatment start to date of death (uncensored) or last followup (censored).

Animal models
All animals were housed at the Vincent A. Stabile Research building animal facility at the MCC, which is fully accredited by the Association  63 were from the Lau laboratory breeding colony. All mice were housed in rooms on a stand ard 12h-12h light cycle, with a temperature range of 68-72 °F and humidity range of 30-70%.
Power calculations determined cohort sizes to detect significant changes in tumor sizes. Generally, by using ten mice per group, we estimate that we will be able to detect a 10% difference in tumor devel opment between any two conditions with a P value of 0.05 and a power value of 0.80 and a 20% change with a P value of 0.05 and a power of 0.95. This calculation has been used previously to designate groups of ten mice 9 , which accommodates for incidental loss of mice due to issues beyond our control (for example, tumor ulceration that requires exclusion from the study). Mouse tumor volumes were measured using length, width and depth divided by 2. At each experimental endpoint or if mice showed signs of metastatic disease, mice were humanely euthanized using CO 2 inhalation in accordance with American Veteri nary Medical Association guidelines.
For all mouse models, 1 × 10 6 melanoma cells were injected subcu taneously in the right hind flanks of each mouse. Between 7 and 14 d, when tumor volumes reached ~150 mm 3 , mice were supplemented with or without 100 mM lfuc (Biosynth) via drinking water, which was provided ad libitum 9 . This dosage is within previously reported ranges for dietary supplementation with lfuc and other similar dietary sugars (for example, dmannose) in rodent studies [64][65][66][67][68] . When tumors reached the endpoint volume (~2 cm 3 ), animals were killed, and tumors and organs were processed for flow cytometric profiling or histological analysis as indicated. The maximum permitted tumor volume was not exceeded. Fig. 1 and Extended Data Fig. 1, SW1 or SM1 mouse melanoma cells were injected into syngeneic C3H/HeN (or NSG) female or C57BL/6 male mice, respectively, as follows: parental SW1 cells for Fig. 1a, parental SM1 cells for Extended Data Fig. 1e, SW1 cells stably expressing either EV or mFuk for Fig. 1e and parental SW1 cells for Extended Data Fig. 1k.

Models of control versus mFuk with or without l-fucose. For
Models of control versus l-fucose with or without FTY720. For Fig. 2b, SW1 mouse melanoma cells were injected into syngeneic C3H/HeN female mice. FTY720 (Cayman Chemicals) was administered at 20 µg every 2 d to inhibit lymph node egress 69 starting on day 12, just before the initiation of lfuc administration, until endpoint. Fig. 1l-o and Extended Data Fig. 1s-w, 3 d before tumor engraftment, PBS (control) or ~300 µg antiCD4 (20 mg per kg) or antiCD8 (20 mg per kg) antibodies were administered by intraperitoneal injection into the indicated cohorts of mice. Immunodepletion antibody or PBS injections were continued every 3-4 d until endpoint. Syngeneic recipient C3H/HeN female or C57BL/6 male mice were injected with SW1 or SM1 cells, respectively.
Anti-PD1 mouse model. For Fig. 5, SW1 or SM1 mouse melanoma cells were injected into syngeneic C3H/HeN female or C57BL/6 male mice, respectively. After approximately 7 d, when the mouse tumors reached ~150 mm 3 , the mice were supplemented either with or with out 100 mM lfuc via drinking water, which was provided ad libitum. Simultaneously, PBS (control) or antiPD1 antibody (20 mg per kg) was administered via intraperitoneal injection every 3-4 d until endpoint.

Statistics and reproducibility
GraphPad Prism version 8 was used for statistical calculations unless otherwise indicated. For all comparisons between two independ ent conditions, ttests were performed to obtain P values and s.e.m. For comparisons between ≥2 groups, oneway or twoway ANOVAs were performed, and P values and s.e.m. were obtained. For the TMA data, the Wilcoxon signedrank test was used to determine sig nificance. Data distribution was assumed to be normal but this was not formally tested; we have included individual data points in all relevant plots.
For molecular and cellbased assays, experiments were performed in three biologically independent replicates with similar results and outcomes. No data were excluded from analyses. Only when cell lines were either contaminated or lost knockdown or exogenous expres sion were the cell lines regenerated and the experiments performed again (that is, more than three times total). For cost feasibility, three biologically independent specimens were pooled for mass spectro metric profiling experiments, which were performed once. However, postprofiling validation experiments were performed in three biologi cally independent experiments with similar results.
Each of the 12 unique mouse models was performed once. How ever, with the exception of the NSG mouse model (Extended Data  Fig. 1k), each mouse model contained builtin repeat control groups (for example, control and lfucfedonly groups) that were repeated in at least one of the other models. Furthermore, SW1 mouse mod els were replicated in SM1 mouse models, which show conservation of our results across different melanoma mutational background and strains and sex of mice. No mice were excluded from the analy ses unless tumors ulcerated or did not graft successfully before treatment.
In general, for molecular, cellbased and mousebased experi ments, the investigators were not blinded to allocation during experi ments and outcome assessment; randomization was not used in these cases. However, the investigators were blinded to allocation during TMA and patient specimen immunostaining, signal measurement and initial analysis.

Reporting summary
Further information on research design is available in the Nature Port folio Reporting Summary linked to this article.

Data availability
Mass spectrometry data have been deposited in ProteomeXchange with the primary dataset identifiers as follows: Extended Data Fig. 2a Fig. 5a, PXD038068. All other data sup porting the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper.

Code availability
No custom code was used in this study. Correspondence and requests for materials should be addressed to Eric K. Lau.
Peer review information Nature Cancer thanks Charles Dimitroff for their contribution to the peer review of this work.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.

Nature Cancer
Article https://doi.org/10.1038/s43018-022-00506-7 Extended Data Fig. 3 | Fucosylated mass spectrometric analysis and knockdown efficiency of H2K1 and H2EB1. (a) (left) Schematic for proteomic analysis of fucosylated proteins in human WM793 melanoma cells using pLenti GFP empty vector (EV), pLentiFUKGFP (o/e), or shFUKexpressing WM793 cells from 8 . Click chemistry biotinylatedfucosylated proteins that were pulled down using Neutravidin beads from the 6alkynylLfucoselabeled cells were subjected to LCMS/MS, and hits were subjected to the indicated filtering scheme followed by Ingenuity Pathway Analysis (Qiagen). (right) Top 20 pathways, plasma membrane and immunerelated proteins identified by Ingenuity Pathway Analysis (Qiagen) to be significantly altered by fucosylation. (b) qRTPCR analysis confirming knockdown of H2K1 (shH2K1; left) or H2EB1 (shEB1; right) using 2 shRNAs per target compared to control nontargeting (shNT) shRNA. n = 3 biologically independent experiments. Red arrows indicate the specific shRNA clones used in functional experiments in the remainder of the study. All error bars represent standard error of the mean.   antiPD1 responder biopsy with high fucoHLADRB1 and CD4 + T cell (upper) vs. a 'noncorrelated' responder biopsy with low fucoHLADRB1 and CD4 + T cells were stained for indicated markers. Yellow dashed lines represent the tumor:stromal interface surrounding melanoma markernegative stroma in the highly correlated responder that is absent in the noncorrelated responder. Yellow asterisks indicate nonnucleated nonspecific staining on the noncorrelated responder slide. (e) Mean tumor cellular (MTC) total fucosylation (upper), total HLADRB1 (middle), or fucosylated HLADRB1 (lower) levels in MDACC patient matched pre/postantiPD1 tumor specimens. C/P/N = Complete/partial/non responder, respectively. Pvalues shown are twosided Pvalues derived from the Spearman correlation test.