Nutrient availability within the tumor microenvironment (TME) can influence cancer progression, therapeutic response, and metastasis (Muir and Vander Heiden, 2018; Muir et al., 2017; Abbott et al., 2023; Cantor et al., 2017; Vande Voorde et al., 2019; Davidson et al., 2016; Gui et al., 2016; Ferraro et al., 2021; Jin et al., 2020; Ngo et al., 2020; Faubert et al., 2020; Rossiter et al., 2021). How metabolite availability is regulated in the TME has been a topic of extensive research (Apiz Saab and Muir, 2023; Sullivan et al., 2019; Lyssiotis and Kimmelman, 2017; Sullivan and Vander Heiden, 2019; Reinfeld et al., 2021); however, a comprehensive analysis of what nutrients are found in cancer tissue interstitial fluid, and how this deviates from nutrients available in the corresponding normal tissue, has not been conducted. Thus, whether global tumor nutrient availability is actively modified by tumor cells, the accompanying immune, endothelial, and stromal cells, or predominantly influenced by the conditions of the originating resident tissue, is not known. Better understanding the determinants of nutrient availability in the TME will inform efforts to understand preferences for specific cancers to develop in specific tissues (Faubert et al., 2020; Bergers and Fendt, 2021; Mosier et al., 2021), and could enable better matching of patients with cancer therapies whose efficacy is influenced by what nutrients are present in tumor tissue (Muir and Vander Heiden, 2018; Muir et al., 2017; Abbott et al., 2023; Cantor et al., 2017; Vande Voorde et al., 2019; Rossiter et al., 2021).

Surgical management of human renal cell carcinoma (RCC) allows for the simultaneous sampling of tumor interstitial fluid (TIF), adjacent healthy kidney tissue interstitial fluid, and plasma collected from patients undergoing nephrectomy. This afforded the opportunity to assess metabolites across these different samples and assess how nutrients found in tumor tissue relate to those found in corresponding normal tissue and blood. Notably, a number of distinct types of primary renal cancers, driven by a diverse set of genetic programs, arise from the kidney (Linehan et al., 2010; Linehan, 2012). This study was unselected with regard to tumor histology, although the diversity of tumors offered a further opportunity to examine the range of nutrient variation across kidney tumors versus normal tissue, as well as among tumors with a variety of histologies. Supported by mass spectrometry-based quantification of polar metabolites and lipids in normal and cancerous tissue interstitial fluid and plasma, we find that nutrient availability in TIF closely mirrors that of interstitial fluid in adjacent normal kidney tissue (kidney interstitial fluid, KIF), but that nutrients found in both interstitial fluid compartments differ from those found in plasma. This analysis suggests that the nutrients in kidney tissue differs from those found in blood, and that nutrients found in kidney tumors are largely dictated by factors shared with normal kidney tissue. We additionally present metabolite measurements in these conditions as a resource to support further study and modeling of the local environment of RCC and normal kidney physiology.

We isolated interstitial fluid from RCC tumors (TIF) and from adjacent non-tumor bearing kidney tissue (KIF), then performed quantitative polar metabolomics (n = 104 metabolites) and lipidomics (n = 210 metabolites) of these fluids alongside plasma collected from patients with kidney cancer at the time of nephrectomy (Figure 1A; Supplementary files 1–3). Principle component analysis (PCA) of polar and lipid metabolite levels measured in these samples revealed some differences between TIF and KIF, but these tissue analytes principally clustered distinctly from plasma (Figure 1B,C), highlighting that the nutrients found in circulation differ extensively from those found in tissue interstitial fluid. In particular, lipid features were tightly concordant between KIF and TIF, in contrast to plasma where lipid components have widespread variation.

We also performed PCA of polar and lipid metabolite levels measured in material derived from the subset of patients with clear cell RCC (ccRCC), as this is the most common subtype of RCC (Linehan, 2012; Muglia and Prando, 2015) and represented the largest fraction of the patient samples available for analysis. We found similar clustering patterns for ccRCC metabolite levels (Figure 1—figure supplement 1A,B) as those found when we analyzed the data collected from all patient-derived material. Further analysis by t- and chi-squared tests revealed that levels of most polar metabolites do not significantly differ between TIF and KIF (Figure 1D, Figure 1—figure supplement 1C). These same analyses revealed a difference in the number of statistically different metabolites when comparing TIF and KIF lipid composition, but these differences were fewer than when comparing the lipid composition of either TIF or KIF compartments to plasma (Figure 1E, Figure 1—figure supplement 1D). Taken together, these data suggest that metabolites found in TIF and KIF are largely similar based on these measurable analytes, although there are individual polar and lipid metabolites that differ between TIF and KIF (Figure 2A,B, Figure 2—figure supplement 1A,B). These data support the notion that the human RCC tumor nutrient environment is primarily driven by factors shared with kidney tissue, rather than being extensively modified by cancer cells or other factors unique to the tumor.

Focusing on differences between TIF and KIF using data derived from the entire patient dataset, we found several metabolites to be elevated in TIF compared to KIF that agree with known metabolic characteristics of RCC. These include increased levels of 2-hydroxyglutarate (Shim et al., 2014; Wang et al., 2021), kynurenine (Wettersten et al., 2015; Ganti et al., 2012; Lucarelli et al., 2017), and glutathione (Wettersten et al., 2015; Hakimi et al., 2016; Figure 2C), all consistent with prior reports and known biology of RCC. Of note, we observed depletion of glucose and elevation of lactate in TIF compared to plasma, again consistent with known increases in glucose fermentation to lactate in this tumor type (Kaushik et al., 2022; Masson and Ratcliffe, 2014; Kaelin, 2008; Courtney et al., 2018; Sanchez and Simon, 2018; Figure 2D). However, the levels of glucose and lactate measured in TIF were similar to those measured in KIF, arguing that glucose was not further depleted in TIF compared to KIF (Figure 2D). These data challenge the notion that cells in the TME are necessarily glucose starved (Burgess and Sylven, 1962; Gullino et al., 1964; Hirayama et al., 2009; Urasaki et al., 2012; Ho et al., 2015), as the average glucose concentration is near or above the threshold identified as limiting for proliferation in a range of cancer cell lines (Birsoy et al., 2014). It is important to note that the levels of glucose measured in bulk TIF may not reflect local depletion in subregions of the tumor. Regardless, these findings fit with other data from RCC, as well as from other tumor types, demonstrating that while glucose availability is reduced in the TME relative to blood, it is not completely depleted (Sullivan et al., 2019; Reinfeld et al., 2021; Siska et al., 2017). This finding is somewhat unexpected, given that VHL loss in ccRCC leads to increased HIF1α-driven expression of glucose transporters and glycolytic enzymes (Linehan et al., 2010; Wettersten et al., 2015; Hakimi et al., 2016; Mandriota et al., 2002; Semenza, 2009; Rathmell et al., 2018), but exchange of glucose between blood and interstitial fluid may be sufficient to prevent glucose depletion to levels that are less than those found in normal kidney tissue. Of note, these data do not necessarily argue glucose metabolism in tumors, or in malignant cancer cells within the tumor, is similar to that of normal tissue, only that availability of glucose across the regions of tissue sampled is similar. Indeed, evidence for increased glucose metabolism in some forms of human RCC is evident from [¹⁸F]Fluorodeoxyglucose (FDG)-positron emission tomography scans (Reinfeld et al., 2021; Hou et al., 2021; Lee et al., 2017; Wang et al., 2012); however, FDG-glucose uptake is variable in the more common ccRCC subtye and can display regional variation (Brooks et al., 2016). In vivo tracing studies of isotope-labeled glucose have demonstrated distinct metabolic fluxes in ccRCC tumors compared to matched adjacent kidney tissue at the time of surgery (Courtney et al., 2018). Lastly, these data are notable with respect to the finding that ccRCC silences the key gluconeogenic enzyme fructose-1,6-bisphosphatase 1 (Li et al., 2014), as it suggests that loss of gluconeogenesis in cancer cells has minimal effect on local glucose availability.

Recent studies have found that arginine is depleted to very low levels in murine pancreatic cancer TIF (Sullivan et al., 2019; Apiz Saab et al., 2023; Lee et al., 2023), a phenomenon facilitated by myeloid-derived arginase activity (Apiz Saab et al., 2023). Of note, arginine is not significantly depleted in KIF or TIF from RCC patients compared to plasma, although levels of related urea cycle metabolites are reduced in TIF and KIF compared to plasma (Figure 2E). These findings suggest that myeloid cell infiltrates in RCC are not sufficient to lower arginine levels, and this observation may relate to the responsiveness of RCC to T-cell immune checkpoint blockade therapy (Santoni et al., 2018). These data are also notable with respect to the observation that primary RCC tumors commonly downregulate ASS1 expression and urea cycle activity (Yoon et al., 2007; Perroud et al., 2009; Ochocki et al., 2018; Khare et al., 2021), implying that the amount of arginine in KIF or TIF is sufficient to support cancer cell proliferation without the requirement for de novo arginine synthesis involving ASS1 upregulation. To this point, a recent study found that metastatic kidney cells in the lung upregulated ASS1 expression due to the lower availability of arginine in the lung compared to the kidney (Sciacovelli et al., 2022).

Given that distinctive lipid metabolism alterations are known to be associated with RCC (Heravi et al., 2022; Gebhard et al., 1987; Saito et al., 2016; Riscal et al., 2021; Valera and Merino, 2011), we probed our lipidomics dataset to investigate changes in lipid composition within RCC TIF compared to KIF and plasma. We found cholesterol levels in both TIF and KIF were notably lower when compared to plasma (Figure 2F), aligning with prior studies observing a roughly fivefold depletion of VLDL, LDL, and HDL cholesterol in tissue interstitial fluid versus plasma (Vessby et al., 1987; Dabbagh and Frei, 1995; Parini et al., 2006). Although cholesterol levels were dramatically lower compared to plasma, we did detect a roughly threefold elevation in cholesterol and cholesteryl esters in TIF compared to KIF (Figure 2G,H), which is intriguing given work showing ccRCC suppresses de novo cholesterol biosynthesis leading to a dependence on exogenous cholesterol import to meet metabolic demands (Riscal et al., 2021). These findings may argue that alterations in local lipid metabolism may be a feature of RCC to supply cancer cells with elevated cholesterol in the TME and raises questions about the dynamic regulation of cholesterol metabolism involving non-cancer cells in these tumors.

Lastly, we examined the composition of plasma from the RCC patients undergoing nephrectomy, and compared the polar metabolite levels to polar metabolite levels in plasma collected from healthy adult donors, and to polar metabolites levels in plasma from non-small cell lung cancer (NSCLC) patients (Figure 2—figure supplement 2A–D). Among the metabolites analyzed, intermediates in nucleotide metabolism stood out as being different between samples. Cystine concentration was also noted to be approximately twofold higher in plasma from patients with RCC compared to levels in plasma from healthy individuals based on the measurements here and available data from the Human Metabolome Database (HMDB; Wishart et al., 2022; Figure 2—figure supplement 3A). This finding demonstrates that there are influences from the tumor on the composition of nutrients in the circulating plasma. Although the RCC patients were thought to have localized disease, whether this influence is due to metabolic demands or contributions within the TME from circulating tumor cells, or as a consequence of undiagnosed metastasis, cannot be determined. The cystine finding was specifically notable because cystine levels have been demonstrated to influence sensitivity to glutaminase inhibitors (Muir et al., 2017), and RCC was a tumor type that showed early responses in glutaminase inhibitor clinical trials (Lee et al., 2022). The concentration of cystine in RCC plasma was comparable to that of NSCLC plasma, which suggests this may be a more common feature of tumor physiology than previously appreciated.

Since the plasma samples collected from both the RCC and NSCLC patients were collected following an overnight fast prior to surgery, whereas the healthy donor samples were collected from non-fasted donors, we considered the possibility that the fasting status of patients during the collection process may impact the concentration of circulating cystine. To explore this further, we collected fed and fasted plasma from a healthy donor and found a 1.6-fold increase in circulating cystine in the fasted state compared to the fed state (Figure 2—figure supplement 3B). This observation offers a plausible explanation for at least a component of the higher cystine concentration in RCC or NSCLC plasma, but a larger trial, including plasma from non-fasted cancer patients, would be needed to verify whether cystine levels are affected by fasting status. Moreover, the observed rise in cystine concentration due to fasting raises an intriguing possibility that therapies influenced by nutrient levels, such as glutaminase inhibitors (Muir et al., 2017), may also be influenced in part by a patient’s diet or fed/fasted state (Wilinski et al., 2019; LaBarre et al., 2021), which could have effects on therapeutic efficacy. Nevertheless, while considering the influence of fasting status on metabolite availability is important, other differences in plasma metabolite levels were unique to either RCC or NSCLC that cannot be explained by blood being collected in a fed or fasted state, such as the buildup of glutathione or uric acid in RCC compared to NSCLC plasma. Whether these other metabolite changes in blood have any biological or clinical significance will require further study.

Assessment of nutrient availability in other tumor types and tissues is needed to determine whether the similarity between nutrient availability in RCC and normal kidney tissue is a general feature of cancers, and examination of metastatic tumors compared to local tissue is also needed. Nevertheless, it is provocative to consider that general tissue homeostatic nutrient maintenance mechanisms are preserved in tumors, and that cancer cells may retain some dependence on nutrients defined by the adjacent origin tissue. To this end, it is notable that the metabolites found in RCC TIF differ from those reported in mouse pancreatic ductal adenocarcinoma (PDAC) TIF (Sullivan et al., 2019; Apiz Saab et al., 2023), and from those found in interstitial fluid derived from mouse brain and mammary fat pad tissue (Ferraro et al., 2021). Findings from analysis of PDAC TIF from tumors implanted in different tissue sites also noted differences in metabolite levels (Sullivan et al., 2019), supporting the notion that cancer must adapt to a specific tissue nutrient environment rather than determine the majority of nutrients present. These findings may suggest that each tissue represents a unique nutrient environment driven my interactions between resident non-cancer cell types, with potential implications for cancer metastasis. That is, this model raises the possibility that accessing a tissue nutrient environment that shares features with the primary cancer site is needed to support proliferation of metastatic cancer cells, and may contribute to the stereotyped pattern of metastases associated with cancers arising in specific tissues (Riihimäki et al., 2018). These findings may also present a challenge for therapies targeting nutrient dependencies in patients with metastatic tumors given that available metabolites could differ in each organ of metastasis.

We acknowledge some caveats with how to interpret these findings. Quantification of steady-state nutrient levels does not directly reflect nutrient use, as nutrient consumption and release rates cannot be derived from these steady-state measurements. An increase in levels of a metabolite in tissue relative to blood must represent some local production in the tissue, while a decrease in the levels of a metabolite relative to blood must represent some local consumption, however a lack of change does not mean a metabolite is not involved in metabolism in that tissue. Exchange of metabolites between blood and tissues may be faster than the rate of consumption or production, and we also did not consider differences between arterial and venous blood metabolites. Moreover, that KIF was collected from non-tumor renal tissue in specimens that were resected from patients with RCC raises the possibility that the levels of at least some of the metabolites measured in normal kidney were affected by the nearby tumor, or the fact that the patient had a cancer. Ideally, metabolites measured in KIF from healthy individuals would be compared with those measured in interstitial fluid from kidney tumors and adjacent normal tissues, although obtaining KIF from healthy patients was not possible for this study. Additionally, we also note that analysis of interstitial fluid will not take into account local variation in nutrient levels across different tissue regions (Pan et al., 2016; Okegawa et al., 2017; Wang et al., 2022a; Miller et al., 2023; Wang et al., 2022b). Microenvironmental differences will be averaged in the pooled collection method employed in this study. Known tumor heterogeneity, including the presence of both cancer and non-cancer cell types, are likely to drive gradients in nutrient availability within tumors (Morandi et al., 2016; Lyssiotis and Kimmelman, 2017). Lastly, the processing methodology employed to collect interstitial fluid may itself lead to some cell lysis and release of intracellular metabolites, which would affect the measured metabolite concentrations. Despite the limitations, this dataset provides a resource for future studies of RCC and normal kidney metabolism. The findings also emphasize that metabolites levels measured in one cancer type cannot necessarily be extrapolated to reflect which nutrients are found in the TME of other tumors. Ongoing efforts to understand how nutrient availability varies across different cancers may inform therapeutic strategies that take into account both tumor genetics and unique nutrient environments in tissues (Abbott et al., 2023).