Increased serum cholesterol and long-chain fatty acid levels are associated with the efficacy of nivolumab in patients with non-small cell lung cancer

Background Lipids have immunomodulatory functions and the potential to affect cancer immunity. Methods The associations of pretreatment serum cholesterol and long-chain fatty acids with the objective response rate (ORR), progression-free survival (PFS), and overall survival (OS) were evaluated in 148 patients with non-small cell lung cancer who received nivolumab. Results When each lipid was separately evaluated, increased low-density lipoprotein (LDL)-cholesterol (P < 0.001), high-density lipoprotein (HDL)-cholesterol (P = 0.014), total cholesterol (P = 0.007), lauric acid (P = 0.015), myristic acid (P = 0.022), myristoleic acid (P = 0.035), stearic acid (P = 0.028), linoleic acid (P = 0.005), arachidic acid (P = 0.027), eicosadienoic acid (P = 0.017), dihomo-γ-linolenic acid (P = 0.036), and behenic acid levels (P = 0.032) were associated with longer PFS independent of programmed death ligand 1 (PD-L1) expression. Meanwhile, increased LDL-cholesterol (P < 0.001), HDL-cholesterol (P = 0.009), total cholesterol (P = 0.036), linoleic acid (P = 0.014), and lignoceric acid levels (P = 0.028) were associated with longer OS independent of PD-L1 expression. When multiple lipids were evaluated simultaneously, LDL-cholesterol (P = 0.003), HDL-cholesterol (P = 0.036), and lauric acid (P = 0.036) were independently predictive of PFS, and LDL-cholesterol (P = 0.008) and HDL-cholesterol (P = 0.031) were predictive of OS. ORR was not associated with any serum lipid. Conclusions Based on the association of prolonged survival in patients with increased serum cholesterol and long-chain fatty acid levels, serum lipid levels may be useful for predicting the efficacy of immune checkpoint inhibitor therapy. Supplementary Information The online version contains supplementary material available at 10.1007/s00262-021-02979-4.


Introduction
Immune checkpoint inhibitors (ICIs) are increasingly used as new standard treatments for cancers, including nonsmall cell lung cancer (NSCLC) [1,2]. However, only some patients benefit from ICI therapy, and the benefits are not sustained in most responders. To address unmet needs, novel biomarkers and/or combination therapies that can enhance the efficacy of ICIs have been intensively investigated [3,4]. A better understanding of factors associated with the therapeutic benefits of ICIs could provide improved precision medicine and new insights into the mechanisms of response and resistance to ICIs.
Unlike cytotoxic chemotherapy with direct anti-tumor effects, ICIs induce anti-tumor responses via immune cells. Therefore, to understand the therapeutic effects of ICIs, assessments of host-tumor interactions are essential in addition to tumor characteristics. It has become increasingly evident that the functions of immune cells are strongly associated with metabolism. Specifically, metabolites have essential roles in controlling the function, differentiation, or longevity of immune cells, thereby affecting cancer immunity [5][6][7][8][9][10][11]. For examples, tryptophan and its metabolites are known to affect cancer immunity [12,13]. Decreased tryptophan levels induce anergy in effector T cells, and increased levels of the tryptophan metabolite kynurenine result in regulatory T cell activation and attenuate anti-tumor immunity [12]. Previously, we found that decreased levels of 3-hydroxyanthranilic acid, a downstream metabolite of the tryptophan-kynurenine pathway, are associated with longer progression-free survival (PFS) in nivolumab-treated patients with NSCLC [14]. The evaluation of immunometabolism will provide important information for ICI therapy.
Lipids, which serve as energy resources and components of the cell membrane, have immunomodulatory potential. Lipids are necessary for immune cell activation, including effector T cells, natural killer T cells, macrophages, and dendritic cells [5][6][7][8][9][10]. Decreased cholesterol levels inhibited the proliferation and activation of T cells [15]. Inversely, increased cholesterol content fostered Th1 differentiation [16]. Lauric acid, a long-chain fatty acid, induced the polarization of naïve T cells toward Th1 and Th17 cells [17,18], and it activated bone marrow-derived dendritic cells and increased T cell activation capacity [19]. Cancer cells require lipids for their proliferation and progression; therefore, these cells aggressively compete with immune cells for lipid uptake. Consequently, cancer cells deprive immune cells of these essential lipids and inhibit their activation [7,[20][21][22][23][24].
However, little is known about the roles of serum lipids in ICI efficacy. The influence of fatty acids on immunity differs by type. It is well known that ω-3 polyunsaturated fatty acids (PUFAs) promote anti-inflammatory responses, whereas ω-6 PUFAs and saturated fatty acids promote pro-inflammatory responses [19,[25][26][27]. Given the diverse immunoregulatory functions of lipids, we hypothesized that different lipids have different effects on the efficacy of ICI therapy. The current study assessed the pretreatment serum levels of multiple lipids and evaluated their associations with the efficacy of nivolumab in patients with previously treated NSCLC. In addition, we also evaluated pretreatment sera from patients with NSCLC who received cytotoxic chemotherapy.

Study design
This was a post hoc analysis of a prospective, multicenter, observational study conducted in 14 hospitals in Japan between July 1, 2016, and December 11, 2018 [28]. Each patient provided written informed consent. Additionally, a retrospective cohort of patients who received cytotoxic chemotherapy (chemotherapy cohort) treated at Hamamatsu University Hospital between January 1, 2009, and December 31, 2019, was evaluated. The study followed the ethical standards of the Declaration of Helsinki. The study protocol was approved by the Institutional Review Board of Hamamatsu University School of Medicine (No. 19-083). The study was registered with the University Hospital Medical Information Network Clinical Trial Registry (000,039,188). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline [29].

Patients
The study inclusion criteria have been described in a previous publication [28]. Briefly, patients with an Eastern Cooperative Oncology Group performance status (ECOG-PS) of 0-2 who were scheduled to receive nivolumab monotherapy for previously treated advanced NSCLC were included. Patients lacking pretreatment serum samples were excluded. Response was assessed every four cycles by local investigators using Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1.
The chemotherapy cohort included patients with advanced NSCLC who received first-line platinum-based chemotherapies. Patients who received non-platinum therapy or had histories of previous chemotherapy or ICI therapy were excluded. The clinical data were retrospectively evaluated via medical record review.

Evaluation of clinical features and outcomes
Clinical factors, including age, sex, smoking status, body mass index (BMI), pathology, clinical stage, ECOG-PS, and line of treatment, were recorded. Tumor-programmed death ligand-1 (PD-L1) protein expression was evaluated immunohistochemically using the tumor proportion score (TPS. E1L3N antibody (Cell Signaling Technology) was used for PD-L1 immunohistochemistry until Japanese approval of the 22C3 pharmDX assay (Agilent). The objective response rate (ORR, sum of the partial and complete response rates was evaluated using RECIST version 1.1.), PFS and overall survival (OS) were evaluated from the time of treatment initiation.

Statistical analyses
Student's t test and Fisher's exact test were used for continuous and categorical variables, respectively. Pearson's correlation analysis was used to assess the correlations between lipid levels and clinical factors. Kaplan-Meier analyses were used for PFS and OS. In these analyses, lipid levels were divided into three groups by concentration as follows: low (below the interquartile range [IQR]), intermediate (within the IQR), and high (exceeding the IQR). The log-rank trend test was used to examine the associations of PFS or OS with stepwise increases in lipid levels. Logistic regression analysis was used to evaluate predictive factors for ORR, and the Cox proportional hazard analysis was used to assess predictors of PFS and OS. Variables significant at P < 0.100 in univariate analyses were employed for multivariate analyses. Additionally, sex and ECOG-PS were associated with lipid levels; therefore, they were included in the multivariate analyses. Two types of multivariate analysis were performed using different combinations of variables. First, to evaluate predictive utility of a single lipid, each individual lipid was analyzed with adjustment for sex, ECOG-PS, and variables significant at P < 0.100 in univariate analysis (single-lipid analysis). Second, to identify major contributing factors among the lipids, multiple lipids significant at P < 0.100 in univariate analysis were analyzed together with adjustment for sex, ECOG-PS, and variables significant at P < 0.100 in univariate analysis (multiple-lipid analysis). When variables had strong correlations with each other (Pearson's correlation coefficient > 0.7), only one was selected for the multiple-lipid analysis to avoid multicollinearity. Candidate combinations for multiple-lipid analysis were created by grouping lipids without strong correlations with each other. Among the candidate combinations, the best model based on the Akaike information criterion was selected as a representative combination of lipids (Supplementary Methods). P < 0.05 (two-sided) denoted significance. All values were analyzed using JMP v13.2.0 (SAS Institute Japan, Tokyo, Japan), excluding log-rank trend test data, which were analyzed using PRISM Version 7.0d (GraphPad Software, CA, USA).

Patient characteristics
Among 200 patients who were enrolled in the original prospective study, 52 patients who did not have sufficient serum samples for lipid measurements were excluded. Therefore, 148 patients with assessable pretreatment serum samples were included in this post hoc analysis. Patient characteristics are presented in Table 1. Most patients were men (82.4%), and most had a history of smoking (88.5%) and ECOG-PS of 0-1 (94.6%). Ninety-three patients (62.8%) had non-squamous cell carcinoma. Tumor PD-L1 expression was assessed in 144 patients (97.3%), and the TPS was 1%-49% in 49 patients (33.1%) and ≥ 50% in 21 patients (14.2%). One hundred and thirty-eight patients (93.2%)
There was no significant association between lipid levels and ORR (Supplementary Table 8).

Associations between lipid levels and the efficacy of cytotoxic chemotherapy
One hundred and thirteen patients in the chemotherapy cohort had comparable demographics as nivolumab-treated patients (Supplementary Table 9).
In univariate Cox proportional hazard analysis, total cholesterol (P = 0.044), linoleic acid (P = 0.018), eicosadienoic  Table 10). After adjustment for age, sex, ECOG-PS, and the receipt of cisplatin-based therapy, no lipids were significantly predictive of PFS (Supplementary Table 10). Univariate Cox proportional hazard analysis identified no predictive factors for OS, and univariate logistic regression analysis indicated that no factors were predictive of ORR (Supplementary Tables 11-12).

Discussion
This was the first study to evaluate the association between multiple lipids and the efficacy of nivolumab in patients with previously treated NSCLC. Interestingly, we found that pretreatment serum cholesterol and long-chain fatty acid levels were significantly associated with PFS and OS, but not ORR, in patients treated with nivolumab, independent of ECOG-PS and PD-L1 expression. Conversely, no lipids were associated with the efficacy of cytotoxic chemotherapy. Serum lipid levels can be easily and noninvasively measured to assess patient status. In addition, lipids are natural products that can be administered easily and safely, and therefore, they have potential as therapeutic agents in combination with ICI therapy. Our data indicated the potential utility of lipids for predicting the efficacy of ICI therapy.
Increased serum lipid levels may compensate for lipid deprivation in immune cells in the tumor microenvironment. Cancer cells accumulate lipids in competition with immune cells, leading to in lipid deficiency in immune cells [7,[20][21][22]. It has been reported that increased exogenous lipids can potentially restore the activity of lipiddeficient immune cells in the tumor microenvironment. Mice with diet-induced hypercholesterolemia exhibited increased numbers of CD8+ and natural killer (NK) cells in hepatocellular carcinoma and adjacent tissues, leading to enhanced cytotoxic activity against cancer cells [30]. In patients with hepatocellular carcinoma, serum total cholesterol and LDL-cholesterol levels were positively correlated with the counts and cytotoxicity of intratumoral NK cells [30].
Another possible mechanism of the correlation between lipid levels and the efficacy of nivolumab was that cholesterol-induced T cell exhaustion can paradoxically increase the efficacy of anti-PD-1 therapy. It has been reported that increased extracellular cholesterol content induces immune checkpoint molecules on T cells including PD-1, leading to T cell exhaustion [31]. It is speculated that patients with cancer and elevated cholesterol levels exhibit PD-1-induced immune escape, increasing their susceptibility to anti-PD-1 therapy, whereas those without high cholesterol levels may experience immune escape caused by non-PD-1-induced pathways, leading to insusceptibility. Contrarily, cytotoxic chemotherapy does not confer any therapeutic advantage for PD-1-induced T cell exhaustion, and thus, this treatment did not influence the association between serum lipids and treatment efficacy.
Similar paradoxical effects were also reported in obesity, an established risk factor for several cancers. In 331 patients who received immunotherapy for melanoma, obese patients (BMI ≥ 30 kg/m 2 ) exhibited significantly longer PFS and OS than those with normal BMI (18.5-24.9 kg/m 2 ), but this was not replicated in patients who received chemotherapy [32]. Similar results were reported in 250 patients with a variety of cancers who received immunotherapies [33]. Mice with diet-induced obesity displayed increased PD-1-positive T cell exhaustion in the tumor microenvironment and greater tumor growth control diet-fed mice. Inversely, the efficacy of anti-PD-1 treatment was better in obese mice than in controls [33]. The study focused on increased leptin levels attributable to obesity as the cause of T cell exhaustion; however, the precise mechanisms underlying the improved efficacy of anti-PD-1 treatment in obesity were not clarified. The current study did not include any patients with BMI ≥ 30 kg/m 2 , and BMI was not associated with the levels of any lipids or the efficacy of nivolumab. However, obesity is often associated with hyperlipidemia, which can affect obesity-associated immune alterations.
Interestingly, ω-6 PUFAs (linoleic acid, arachidic acid, eicosadienoic acid, and dihomo-γ-linolenic acid) and saturated fatty acids (lauric acid, myristic acid, stearic acid, behenic acid, and lignoceric acid), which are known to promote pro-inflammatory response, displayed positive associations with treatment efficacy. On the contrary, ω-3 PUFAs (linolenic acid, eicosatetraenoic acid, eicosapentaenoic acid, docosapentaenoic acid, or docosahexaenoic acid), which are known to promote anti-inflammatory responses, exhibited no association with efficacy. Because of their counter-regulatory roles in immunity, ω-3 PUFAs are considered beneficial for inflammatory diseases, cardiovascular diseases, and cancers, whereas ω-6 PUFAs and saturated fatty acids are considered risk factors [17,18,34]. However, in the context of immune therapy, pro-inflammatory responses have potential benefits.
However, controversies exist regarding the roles of cholesterol and fatty acids in cancer immunity. Lipid cannot always provide favorable conditions for cancer immunity. It is reported that LDL-cholesterol inhibited T cell activation, resulting in impaired antitumor function [35]. Increased cholesterol and/or fatty acid levels have the potential to attenuate anticancer immunity via activation suppressor immune cells, such as regulatory T cells (Tregs), myeloid-derived The hazard ratio was adjusted for sex, body mass index, Eastern Cooperative Oncology Group performance status, pathology, and programmed death ligand-1 expression. Total cholesterol, myristic acid, linoleic acid, and lignoceric acid were excluded from the multiple-lipid analysis because of their strong correlations with other lipids, and the results of the combination of the excluded lipids are presented in Supplementary  [20,22,24]. The difference in cancer immunity between LDL-and HDL-cholesterol is also unclear. Generally, LDL and HDL exert counterregulatory effects on cholesterol transport: LDL promotes inflammation via cholesterol influx into immune cells, whereas HDL reduces inflammation via cholesterol efflux [8]. If cholesterol-induced pro-inflammatory responses are beneficial for anticancer immunity, then increased levels of HDL-cholesterol would be unfavorable. However, in the current study, increased HDL-cholesterol levels were associated with longer PFS and OS independent of LDLcholesterol levels. The tumor microenvironment has complicated cross-interactions among tumor cells, effector cells, and regulatory immune cells, and metabolism is uniquely regulated in each cell type. Further studies are warranted to understand the complex mechanism of lipid metabolism in cancer immunity. The current study had three main limitations. First, it is unknown which lipids have actual and/or the most relevant immunomodulatory activities that affect the efficacy of nivolumab. The multiple-lipid analysis using the selected combinations identified certain lipids as major contributing factors. However, selection based on correlation coefficients could not completely eliminate potential confounding. This is because lipids have a complicated metabolic cascade in which they are directly and/or indirectly associated with each other, and lipid composition varies among foods (major source of lipids for humans). In addition, it is also possible that selection based on correlations ignored the potential importance of the excluded lipids. Second, the differences in lipid composition and/or concentration between peripheral blood and the tumor microenvironment are unknown. In fact, cancer tissue is reported to have twofold higher free cholesterol and 3.5-fold higher esterified cholesterol levels, probably because of increased intake by tumor cells [36]. Third, serum lipids levels were provisionally determined according to the statistical distribution of the limited number of study patients. It is unknown whether higher serum levels will always lead to better clinical outcomes. In addition, it is also unknown whether increased lipid levels following supplemental treatment have beneficial effects. The regulation of lipid metabolism can potentially enhance the efficacy of immunotherapy, and this strategy has attracted attention as a novel therapeutic option [37][38][39]. Further studies are needed to evaluate the optimal levels and therapeutic benefits of lipids in ICI therapy.
In conclusion, increased serum cholesterol and long-chain fatty acid levels were associated with better PFS and OS in patients with previously treated NSCLC who received nivolumab. Serum lipids may be useful for predicting the efficacy of ICIs.