SERS analysis of cancer cell-secreted purines reveals a unique paracrine crosstalk in MTAP-deficient tumors

Significance

This research unveils a unique paracrine crosstalk in the tumor microenvironment (TME) involving cancer cells with methylthioadenosine phosphorylase (MTAP) deletions, a common trait linked to poor prognosis. Using surface-enhanced Raman Scattering (SERS), we discovered that these cells secrete 5′-methylthioadenosine (MTA), which can be further metabolized by fibroblasts, leading to molecular changes consistent with cancer aggressiveness and affecting macrophage polarization. Our study not only highlights the potential of SERS for exploring secreted signatures in cancer research but also offers valuable insight to guide more targeted experiments toward deciphering the role of such secreted metabolites in the TME, which could inform future therapeutic strategies.

Abstract

The tumor microenvironment (TME) is a dynamic pseudoorgan that shapes the development and progression of cancers. It is a complex ecosystem shaped by interactions between tumor and stromal cells. Although the traditional focus has been on the paracrine communication mediated by protein messengers, recent attention has turned to the metabolic secretome in tumors. Metabolic enzymes, together with exchanged substrates and products, have emerged as potential biomarkers and therapeutic targets. However, traditional techniques for profiling secreted metabolites in complex cellular contexts are limited. Surface-enhanced Raman scattering (SERS) has emerged as a promising alternative due to its nontargeted nature and simplicity of operation. Although SERS has demonstrated its potential for detecting metabolites in biological settings, its application in deciphering metabolic interactions within multicellular systems like the TME remains underexplored. In this study, we introduce a SERS-based strategy to investigate the secreted purine metabolites of tumor cells lacking methylthioadenosine phosphorylase (MTAP), a common genetic event associated with poor prognosis in various cancers. Our SERS analysis reveals that MTAP-deficient cancer cells selectively produce methylthioadenosine (MTA), which is taken up and metabolized by fibroblasts. Fibroblasts exposed to MTA exhibit: i) molecular reprogramming compatible with cancer aggressiveness, ii) a significant production of purine derivatives that could be readily recycled by cancer cells, and iii) the capacity to secrete purine derivatives that induce macrophage polarization. Our study supports the potential of SERS for cancer metabolism research and reveals an unprecedented paracrine crosstalk that explains TME reprogramming in MTAP-deleted cancers.

Contrary to the long-held belief that cancer was simply a mass of transformed cells, the current view defines solid tumors as dynamic pseudo-organs governed by intricate relationships between their components (1). These interactions between the tumor and its ecosystem (a heterogeneous mixture of infiltrating and resident host cells, secreted factors, and extracellular matrix) create a unique niche, collectively known as the tumor microenvironment (TME) (2). As a result, the TME is now recognized as a crucial determinant of the disease outcome, with some stromal compartments promoting malignant cell progression and others impeding it (3, 4). Notwithstanding, the TME can also exhibit a certain degree of plasticity, transitioning from tumor-suppressive to tumor-promoting environments by triggering various signaling cues. Cancer cells themselves are active participants in TME remodeling, through the secretion of different cytokines and metabolites (5, 6). Consequently, targeting such paracrine signals has emerged as a promising approach toward more sophisticated treatments, including innovative cell-based therapies (7, 8).

Small metabolites have been identified as relevant paracrine signals, capable of rewiring the tumor local milieu. Through such metabolic interactions involving the exchange of nutrients and intermediate metabolites between cells, the components of the TME can act cooperatively to support tumor growth and impair immunity. For example, Kynurenine production in IDO-1 overexpressing tumors, or lactate accumulation, have been demonstrated to induce immunosuppression, highlighting their potential as biomarkers of poor prognosis (9–11). Metabolomics approaches (i.e., large-scale studies of intracellular or extracellular small molecules) are attracting attention as a source of information relative to the composition of the extracellular milieu. However, the application of these technologies faces some specific challenges: 1) the sample destructive nature of these approaches, 2) the need for invasive interventions, and 3) the elaborate sample preparation procedures that slow down the process of data collection and hinder routine implementation (12, 13). Consequently, the development of novel methods to rapidly detect multiple metabolites is still a pending challenge.

Surface-enhanced Raman scattering (SERS) spectroscopy is an optical ultrasensitive analytical method that can be implemented to monitor metabolites within cell environments, with a low-impact intervention. Its strength lies in the ability to exponentially amplify the specific inelastic scattering—or Raman fingerprint—of molecules when they are adsorbed onto corrugated metal surfaces, such as those built upon metal nanoparticles. Since its discovery almost 50 y ago, it has gained increasing attention as a powerful tool to enhance the sensitivity of Raman spectroscopy in general, and toward biomedical applications in particular (14). Furthermore, SERS can work as a label-free approach that, with the adequate plasmonic transducer, offers a cost-effective method capable of detecting the presence of several biomolecules in a single measurement, upon laser irradiation.

The state of the art in SERS bioapplications validates this technology as an alternative strategy for imaging metabolic profiles, with the distinct advantage that it can provide a more holistic view of biochemical changes that could be otherwise ignored in more targeted techniques such as LC–MS (15). However, most SERS studies have focused on cellular processes already described by other routine methods, i.e., they did not exploit SERS to reveal additional insights about metabolic interactions (16, 17). To fully unleash this potential, we propose using SERS as a front-line tool to readily interrogate the extracellular environment under multiple cell conditions, which also include coculture with representative cells from the surrounding tumor stroma—e.g., fibroblasts and macrophages.

In this proof-of-concept approach, we have narrowed down our focus to tumor cell lines with an altered methionine salvage pathway. Specifically, we studied cells with a deletion in the rate-limiting methylthioadenosine phosphorylase (MTAP) enzyme, which is commonly deleted in many cancers (18–20). MTAP mutation elicits the accumulation of its main substrate, 5′-methylthioadenosine (MTA), which under normal physiological conditions serves as the convergence point for the methionine cycle and polyamine production (21, 22). Intriguingly, the role of MTAP in cancer and the selective advantage underlying its deletion is not yet fully understood, which made it compelling for our proof-of-concept SERS metabolic screening (23, 24). Recent studies have highlighted that the subsequent excess of MTA could be secreted to the TME, where it is metabolized by other MTAP-positive cells (25, 26). Nevertheless, the consequences of such an MTA elevation, either intracellularly or in the extracellular space, are still a subject of intense debate in the literature, as well as the kinetic regulation of its influx and efflux (27).

We sought to assess whether SERS could be employed to explore MTA dynamics in complex cell environments with no need for specific targeting, and the collected information be utilized to guide the inspection of potential MTA-mediated paracrine communication, supporting further examination by complementary, targeted analytical approaches. By doing so, we attempted not only to validate the applicability of SERS for the multiplex detection of MTA and other related metabolites but also to shed light on metabolic networks of MTAP-deficient environments.

Results and Discussion

The Metabolic Consequences of MTAP Loss Can be Monitored by SERS.

The high frequency of MTAP deletion was long considered a secondary or “passenger” event in tumors with deletion in the CDKN2A locus, a more widespread event with sound evidence to promote cancer. Both genes are closely located on chromosome 9p21, which motivated the original idea that loss of MTAP occurred only coincidentally during CDKN2A deletions (28, 29). However, over the years, an increasing number of reports have revealed a direct association between homozygous deletion of MTAP and poor prognosis, uncovering strong evidence that links MTAP loss to tumor growth (26). Thus, in an effort to understand this unexpected relationship, numerous studies have investigated the metabolic alterations that provide MTAP-negative (MTAP−/−) cells with a competitive advantage, mostly focusing on intracellular mechanisms (27). Recently, interest has expanded to the extracellular environment and the secretomes of MTAP−/− cells. Building on our previously reported achievements in identifying secreted metabolite signatures using SERS, we propose harnessing such established protocols and other conventional methods to monitor cell lines with varying degrees of MTAP expression, and their impact on neighboring cells (16, 30). In this implementation, the plasmonic substrates used to enhance the Raman signal of secreted metabolites comprised 30 nm AuNPs, drop cast on glass slides. This straightforward method ensures continuous accessibility to the substrates while avoiding time-consuming fabrication protocols, thus facilitating SERS measurements under multiple conditions (SI Appendix, Fig. S1).

We registered the supernatants of four distinct cancer cell lines that exhibit significant variations in MTAP levels (confirmed by qPCR, see SI Appendix, Fig. S2). As noted in Fig. 1A, we captured the signal from two cell lines with homozygous loss of MTAP (U87, MDA-MB-231) and two other lines with diploid MTAP (HeLa, PC3) (31). In accordance with previous reports (18, 27), we observed vibrational differences among all the interrogated cell lines, reflecting their unique subset of secreted metabolites. Still, unlike in such previous studies, we also intended to determine the common spectral features of MTAP−/− cells. In particular, applying partial least squares discriminant analysis (PLS-DA) to spectral data, we could identify the key vibrations, characteristic of each condition, namely MTAP−/− vs. MTAP+ cells. Through the PLS-DA approach, the most important wavenumbers (loadings) discriminating cells with MTAP loss were plotted in Fig. 1B (Upper panel). As a result, we could confirm that the SERS spectra of U87 and MDA-MB-231 presented distinct vibrations at 735 cm⁻¹ and 1,320 cm⁻¹, typically associated with out-of-plane bending vibrations of aromatic rings and CH₃ bending vibrations, respectively, as well as other bands at 1,475 cm⁻¹ (amine bending vibrations in primary or secondary amines) and at 640 cm⁻¹ (sulfur-containing groups or out of plane bending-vibrations) (32). Notably, these wavenumbers were compatible with the SERS vibrations of an MTA standard in PBS (100 µM), as displayed in the lower panel in Fig. 1B. In SI Appendix, Fig. S3, we also present the Raman spectra of MTA along with the density functional theory (DFT) simulations of its main vibrations at various conformations.

Fig. 1.

Dynamics of extracellular MTA can be monitored by SERS. (A) SERS spectra acquired from MDA-MB-231, PC3, U87, and HeLa cell supernatants on CTAC-AuNP substrates; colors refer to MTAP expression levels: red for MTAP-negative (MTAP−/−), orange for MATP-positive cells (MTAP+). (B) The upper panel showcases wavenumbers with the highest scores derived from the PLS-DA model, highlighting discriminative features between MTAP−/− and MTAP+ cells. In contrast, the lower panel presents the SERS spectrum of MTA at a concentration of 100 µM in water. The highlighted wavenumbers exhibit close alignment with MTA's characteristic vibrations. An Inset provides the molecular structure of MTA for reference. (C) Scheme depicting the strategy followed to monitor MTAP loss in cancer cells by SERS. (D) Calibration curve of MTA at varying physiological concentrations (1–20 µM); asterisks mark the signal used for internal calibration, resulting from CTAC vibrations within the plasmonic substrate. (E) SERS spectra of selected cell lines after 24 h of supplementation with MTA (10 µM). (F) Upon dimensional reduction by t-SNE, spectra from cells supplemented with MTA form two well-separated clusters, according to the levels of MTAP expression.

These initial insights led us to conclude that extracellular MTA, selectively released by MTAP-deficient cells, could be effectively tracked by SERS. Hence, we could follow the strategy depicted in Fig. 1C to monitor MTAP loss using the extracellular media as a proxy. Moreover, among the different AuNP substrates, gold nanoparticles capped with cetyltrimethylammonium chloride (CTAC-AuNPs, see experimental section for synthesis protocol and characterization) allowed us to approximate the levels of extracellular MTA at a physiologically relevant range (1–20 µM, see Fig. 1D and SI Appendix, Fig. S4). MTA spectra recorded on different AuNP-based plasmonic substrates [drying drop and superlattices (33)] are also provided in SI Appendix, Fig. S5. Using CTAC-AuNPs, the peaks from CTAC adsorbed on the NPs surface (labeled with an asterisk, *) served as a suitable internal control to approximate the concentration of MTA in the samples. SERS results were confirmed using quantification of extracellular metabolites by conventional methods (LC–MS, SI Appendix, Fig. S6).

In a subsequent experiment, we challenged the same four cell lines with varying concentrations of exogenous MTA. Our objective was to continue assessing the behavior of these cells in response to MTA supplementation. As expected, MTAP-negative cells (MDA-MB-231 and U87) exhibited defects in MTA uptake and consumption, resulting in pronounced MTA-related peaks that could be readily measured from supernatants with 10 µM supplementation (see Fig. 1E, also proven intra/extracellularly by LC–MS measurements shown in SI Appendix, Fig. S7). In contrast, HeLa and PC3 cells, i.e., MTAP competent lines, consumed MTA within the first 24 h. Additional experimental evidence suggested that the kinetics of MTA consumption were dependent on cell density and metabolite uptake (i.e., MTA is not degraded by secreted enzymes, as explained in SI Appendix, Fig. S8). In this regard, the observed fluctuations at the main MTA vibrations accounted for a slower uptake into HeLa cells. Of note, 5 µM MTA was sufficient to strongly impact the recorded SERS signals in the extracellular media (see Fig. 1F). To visualize those changes in a bidimensional plot, we carried out a multivariate spectra analysis by applying an unsupervised clustering method: t-SNE (34), which was able to classify cell lines according to MTAP status.

MTA Is Recycled Through Tumor–fibroblast Crosstalk.

Informed by our SERS approach, we further explored the impact of extracellular MTA on other stromal and immune cells within the TME. We chose to investigate the impact of extracellular MTA accumulation on fibroblasts and macrophages, two stromal components which are profoundly affected by cancer cell paracrine signals (35, 36).

First, we tested the extracellular media spectra of fibroblasts when exposed to supplemented MTA (5 and 10 µM), comparing the supernatant collected after 24 h with those under control conditions. Fibroblasts were challenged with MTA and the corresponding SERS spectra were collected after the 24-h period, which revealed strong signals at 735 cm⁻¹ (Fig. 2A). We reasoned that this signal might represent a defect in MTA uptake and metabolism in the fibroblasts, despite their detectable expression levels of MTAP (SI Appendix, Fig. S9). However, other MTA-associated bands, specifically those at 1,320 or 1,475 cm⁻¹, were absent in the spectra. These data suggest that the 735 cm⁻¹ band likely originates from a closely related molecule, but not from MTA. To better separate potentially overlapping peaks at such regions of interest, we additionally performed a deconvolution analysis, which allowed us to discern the contribution of different metabolites to the final spectra (e.g., it resolves two individual components at 725 cm⁻¹ and 735 cm⁻¹). This analysis also yielded consistent differences between the spectra of MTA-supplemented fibroblasts and MDA-MB-231 cells over the 1,475 cm⁻¹ region (Fig. 2A). To filter the potential candidates to which these spectral signatures may correspond, we analyzed the intracellular metabolic pathway of MTA (Fig. 2B). Commercial solutions of MTA and its derivatives, such as adenine (Ade), ATP, or hypoxanthine (HX), were therefore measured by SERS, and then compared with those acquired from fibroblast cultures (SI Appendix, Fig. S10). By performing a fitting analysis based on pure solutions and cell-derived spectra, we found a better spectral correspondence between fibroblast supernatants and commercial solutions of Ade and HX—both with fitting scores above the MTA fitting. Importantly, these results were supported by LC–MS measurements, which corroborated the consumption of supplemented MTA in fibroblast media and the production of HX. Our approach illustrates a unique opportunity for discovery using SERS. Whereas other targeted metabolic analyses such as LC–MS would provide data about predefined molecules, SERS allows for the evaluation of metabolic changes in complex media, thus uncovering metabolic conversions that could be biologically relevant.

Fig. 2.

Neighboring fibroblasts metabolize extracellular MTA, leading to the release of various byproducts. (A) Representative SERS spectra of fibroblast supernatants, where the main MTA vibrations are identified with color bars. Right panels present the deconvolution of peaks fitted to the selected wavenumber range (between 700–800 and 1,400–1,500 cm⁻¹). The plots display both the original data and the individual Gaussian components, showcasing the underlying peak structure within the analyzed spectrum. (B) Diagram of the intracellular pathway of MTA, highlighting its bifurcation after MTAP activity, toward either polyamine synthesis or degradation into uric acid (including the salvage pathway via hypoxanthine). The scheme in the Upper Inset depicts the extracellular fluctuations of purines in fibroblasts (C) LC–MS results under control conditions or 10 µM MTA, labeled as C or S, respectively. Levels of MTA, adenine, and hypoxanthine are plotted separately, for both MDA-MB-231 and HBF cell lines. (D) SERS spectra of fibroblasts (HBF) and MDA-MB-231 coculture supernatants 24 h upon MTA supplementation and control conditions, along with convolutional analysis within the 700–800 cm⁻¹ range. (E) LC–MS results upon 24 h of MTA supplementation (10 µM). Levels of MTA, adenine, and hypoxanthine were plotted separately, for MDA-MB-231, HBF, and the coculture of both lines. (F) Consumption of supplemented HX and adenine, both at 10 µM (Upper and Lower panels, respectively), over time. The ratio of the peaks at 725 cm⁻¹ (for HX) and 735 cm⁻¹ (for adenine) against the CTAC peak (760 cm⁻¹) were calculated over time, showing a decreasing profile for both metabolites as time progresses. (G) Schematic image representing the purine-based metabolic crosstalk between cancer cells (MDA-MB-231) and stromal fibroblasts (HBF). Due to MTAP deletion, cells release MTA into the extracellular milieu, from which fibroblasts can uptake it and convert it into adenine and hypoxanthine. These metabolites are then secreted and internalized by the tumor once again. (H) Two-dimensional representation of the spectra obtained under 10 µM of MTA for single cell cultures (HBF and MDA-MB-231) and the coculture modality, visualized using the statistical method “t-Distributed Stochastic Neighbor Embedding” (t-SNE). (I) Cytotoxicity assay assessing LDH release in HeLa, HBF, and MDA-MB-231 cells, treated with or without 5 µM Pemetrexed, for a duration of 72 h. The graph evaluates the impact of HBF (treated with 10 µM MTA for 24 h)-conditioned medium (HBF-CM) on MDA-MB-231 viability, with the results expressed as a percentage of cytotoxicity. Mean values (bars), SD (whiskers), each point represents an individual biological experiment. Asterisks indicate statistical significance in the ANOVA test with post hoc analysis: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Upon MTA supplementation, the media of fibroblasts was enriched with higher amounts of Ade and HX. In contrast, MDA-MB-231 cells only secreted and accumulated MTA (Fig. 2C). The apparent discrepancy in the HX signals between mass spectrometry (Fig. 2C) and SERS measurements (Fig. 2A) could be explained on the basis of the superior affinity of adenine for metallic surfaces, which hinders HX monitoring. This is particularly evident for fibroblasts with 10 µM MTA supplementation, where no HX could be detected in the SERS spectra. Experimental studies and theoretical calculations are shown in SI Appendix, Figs. S11 and S12, respectively. Overall, we interpret these results in terms of a metabolic conversion of MTA exclusive to fibroblasts (depicted in the illustration of Fig. 2B, confirmed for different fibroblast subtypes, SI Appendix, Fig. S13), which leads to the production and secretion of byproducts (adenine and hypoxanthine).

Counterintuitively to reported clinical observations, the secretion of MTA by MTAP-deficient cells could be perceived as a detrimental event that compromises tumor growth, as it reduces intracellular purine levels essential for nucleic acid synthesis and overall cellular function (26). However, the above-described findings regarding the conversion of MTA in fibroblasts may have uncovered a metabolic symbiotic pathway through which MTAP-deficient cells can replenish their purine pools through intercellular communication with fibroblasts. Due to the lack of MTAP, cancer cells cannot metabolize MTA through the methionine salvage pathway, which would compromise purine availability and nucleotide synthesis. However, tumor-resident fibroblasts would uptake and metabolize MTA-producing purine intermediate metabolites that could be made available for cancer cells through secretion. As a result, a symbiotic metabolic relationship between these two cell populations emerges as a plausible explanation for how tumor cells can overcome potential purine limitations in MTAP-deficient environments.

With this hypothesis in mind, we cocultured MDA-MB-231 and HBF (Human Breast Fibroblast) cells and recorded the SERS spectra of their derived supernatants in Fig. 2D. The resulting SERS spectra did not display intense vibrations at 735 cm⁻¹, confirming that neither MTA nor adenine accumulated in the coculture media. Notwithstanding, the convolutional panel shows a weak signal of HX for 5 and 10 µM MTA, albeit it completely disappeared after 24 h in monoculture experiments (SI Appendix, Fig. S14). The same experiments were replicated employing 3D spheroids (both mono and coculture), to provide more physiologically relevant conditions, as well as in situ measurements by applying PLGA-covered substrates to acquire multiple measurements within a single SERS substrate (SI Appendix, Figs. S15 and S16). These results were further validated by targeted LC–MS studies in Fig. 2E, highlighting the pronounced extracellular decay in Ade and MTA in coculture. This decay is particularly notable when compared with the consumption of the same metabolites in cancer cells alone. In Fig. 2F, we demonstrate that, despite an impaired MTA metabolism, MDA-MB-231 has intensively taken up Ade from media, and, to a lesser extent, consumed HX. As a complementary test, we decided to further examine this crosstalk by analyzing spectra at various time points along the course of the experiment. In brief, this assay included a 24-h HBF monoculture (treated with MTA), followed by subsequent MDA-MB-231 culture in the resulting conditioned media (namely as HBF-CM). This enabled us to visualize, in SI Appendix, Fig. S17, the changes from the initial supplemented media, with MTA, to the subsequent stages of the experiment. These experiments confirmed the interplay between cancer cells and fibroblasts via MTA, which contributes to the recycling of purine pools, and may facilitate tumor growth in the context of MTAP deletion.

Moreover, the dimensional reduction of SERS spectra by t-SNE analysis enabled the visualization of specific clusters (Fig. 2H). In this instance, we could distinguish three distinct clusters, representing three TME metabolic scenarios upon supplementation with MTA: high extracellular MTA in MTAP-deficient tumor cells (representing the inability of these cells to metabolize MTA); low extracellular MTA with high abundance of Ade and HX in fibroblasts (representing MTA uptake and conversion into Ade and HX) and low extracellular MTA with low abundance of Ade and HX in cocultures (representing the uptake of MTA by fibroblasts and the recycling of fibroblast-produced purines by tumor cells).

Finally, we explored whether this metabolic symbiosis could harbor therapeutic potential. Loss of MTAP in tumor cells reduces purine-recycling capacity. In turn, MTAP-deficient cells are highly dependent on the de novo purine synthesis (Fig. 2I, comparing MDA-MB-231 cells to HeLa), which is being exploited as a therapeutic option using drugs such as Pemetrexed (PEM) that inhibit this pathway (37). We predicted that the high sensitivity of MTAP-deficient MDA-MB-231 cells would be counteracted by the presence of fibroblasts, since they would provide a unique source of purines. Interestingly, when MDA-MB-231 cells were exposed to the HBF CM medium (supplemented with MTA), sensitivity to PEM was significantly reduced (Fig. 2I). This finding underscores the underlying implications of MTA-based metabolic symbiosis for tumor cell survival, serving as an illustrative example of how such interactions can eventually impact disease outcomes.

Overall, the previous results suggest a unique metabolic profile for fibroblasts, in response to extracellular MTA (confirmed for different fibroblast subtypes, SI Appendix, Fig. S13). Considering that fibroblasts typically play crucial roles in metabolic rewiring and signaling through the secretion of various metabolites and cytokines, we hypothesized that MTA could, in turn, influence their molecular contribution to TME remodeling (38, 39). Thus, in MTAP-deficient environments, the activation status of stromal fibroblasts could be influenced by the extracellular availability of MTA. To investigate this further, we examined potential changes in transcriptomics upon MTA supplementation, using RNA-sequencing analysis (refer to Materials and Methods for RNA extraction protocol). Hierarchical clustering and PCA (in SI Appendix, Fig. S18 and Fig. 3A) revealed the high reproducibility of the gene expression profiles from interrogated fibroblasts. This analysis identified over 950 genes that were differentially expressed between control and MTA-treated fibroblasts, with both up-regulated and down-regulated genes represented in the volcano plot (Fig. 3B). Manual curation of the up-regulated genes allowed us to spot some enriched gene sets implicated in critical biological pathways. These pathways include the release of paracrine soluble factors such as PGE2, HGF, and TGF-β, the Wnt Signaling Pathway (WNT5A), myofibroblast differentiation (COL4A6, MYO1B), and angiogenesis (OLR1, VCAN). The heatmap in Fig. 3C illustrates the upregulation of representative genes, selected for their involvement in these and similar processes. Furthermore, the Gene Set Enrichment Analysis (GSEA) results shown in Fig. 3 D and E (Materials and Methods) indicated that exogenous MTA significantly upregulates molecular programs connected to the extracellular compartment (e.g., basement membrane and collagen-containing extracellular matrix) and to inflammatory signatures, which might suggest a downstream influence on other neighboring cell populations. This analysis also identified with the highest normalized enrichment score (NES) (40), a variety of functional signaling pathways associated with immune responses such as TNF-$\alpha$ signaling via NFB, IFN-$\gamma$ response, and IL-6/JAK-STAT3. Along with these outputs, the GSEA highlighted other additional MTA-associated changes, including those related to Epithelial to Mesenchymal transition, cytokine and chemokine production, and cancer pathways: MAPK, mTOR or Wnt/β-catenin (SI Appendix, Table S1). Arguably, the observed response to extracellular MTA implies a considerable impact on the transcriptional program of treated fibroblasts, which is subsequently accompanied by changes in their signaling programs.

Fig. 3.

Extracellular MTA can rewire fibroblast phenotypes. (A) Principal component analysis (PCA) score plot of the first two principal components (PC1—80.5%, PC2—19.4%), showing the relative separation of both control fibroblasts and MTA-treated fibroblasts. (B) Analysis of the transcriptomics results, which compare gene expression in control and MTA treatment, the volcano plot shows those genes with the most substantial fold changes between the two conditions. (C) A heat map of RNA-seq expression data reveals selected genes associated with cancer pathways, mesenchymal transition (red), and immune response (orange), with gene expression illustrated using Z-scores. (D) A dotplot representation of the most enriched gene sets (determined based on NES and P-values) by the GSEA. In this plot, gene ontology (GO) cellular compartment terms and Hallmark pathways gene sets are shown. The size of each dot corresponds to the degree of enrichment in the respective gene set, providing a visually intuitive illustration of the results. (E) Results from the GSEA using the “Inflammatory Response” gene set from MSigDB Hallmark 2020. The enrichment score (ES) plot displays the extent to which the gene set is overrepresented at either the top or bottom of the ranked list of genes. The ranked list metric shown below reflects the correlation between each condition and the phenotype of interest. The obtained ES indicates a significant association between MTA exposure and the inflammatory response.

The Role of Fibroblasts as Modulators of the Immune Response to MTA.

We demonstrated that MTA could function as an active paracrine messenger, modulating the activity of neighboring fibroblasts. Similar to the study in fibroblasts, we aimed at interrogating macrophages upon exogenous MTA addition to the local milieu. We thereby exposed the RAW 264.7 cell (a widely used cell model for preliminary macrophage studies) to increasing concentrations of MTA up to 10 µM and examined the secreted signatures using SERS. Based on the collected data (Fig. 4A), we can assert that RAW 264.7 cells were able to consume MTA without secreting additional purine derivative metabolites during the monitoring time lapse—at least within the sensitive range of the technology, ≈1 µM. It should be noted that this finding differs from what we observed above with fibroblasts and the release of purine derivatives. LC–MS experiments were also conducted to validate the SERS screening (Fig. 4B and SI Appendix, Fig. S19).

Fig. 4.

Extracellular MTA modulates macrophage polarization. (A) SERS spectra of RAW 264.7 cell supernatants with varying MTA concentrations. (B) LC–MS results (MTA, Ade and HX) of RAW 264.7 cells after being supplemented with 10 µM MTA. (C) NOS2-normalized gene expression on RAW 264.7 cells after supplementation with metabolites 10 µM of MTA, 10 µM adenine, 10 µM hypoxanthine, and 25 µM of MTA, individually. GAPDH was used as housekeeping gene. (D) Schematic representation of the CM polarization studies on RAW 264.7 cells. (E) NOS2-normalized gene expression in RAW 264.7 cells (proinflammatory marker) following treatment with conditioned media from HeLa, MDA-MB-231, and fibroblasts (Left), and media derived from coculture (Right). All cultures were supplemented with MTA 10 µM. GAPDH was used as housekeeping gene. (F) Arg-1 normalized gene expression in RAW 264.7 cells (antiinflammatory marker) after exposure to the same conditions of HeLa, MDA-MB-231, and fibroblasts, and coculture derived media, Left and Right panels, respectively. GAPDH was used as housekeeping gene. Mean values (bars), SD (whiskers), each point represents an individual biological experiment. Asterisks indicate statistical significance in the one sample t test: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 compared to the control group.

Furthermore, in the specific case of macrophages, we investigated whether the availability and uptake of MTA could influence its phenotype and biological activity. To this end, we monitored whether MTA could influence macrophage polarization towards pro- or anti-inflammatory phenotypes (41, 42). Previous clinical reports have related the loss of MTAP to an impaired immune activity against tumor cells, identifying high concentrations of MTA as a potential driver of immune evasion (43).

In this context, to assess macrophage polarization, we conducted specific qPCR analyses to measure the expression level of well-established polarization markers: higher Arginase 1 (Arg1) levels are commonly found in antiinflammatory macrophages, whereas Nitric Oxide Synthase 2 (NOS2) expression is activated in proinflammatory phenotypes (see Materials and Methods and SI Appendix, Fig. S20 for further details). To draw a comparison, we evaluated the impact on polarization under supplementation with the principal purine derivatives identified thus far, namely MTA, adenine and hypoxanthine (all at 10 µM). Using these conditions (Fig. 4C), NOS2 expression increased significantly with both MTA and adenine incubation. Combined with a negligible effect on Arg 1 levels (SI Appendix, Fig. S21), these results suggest that proinflammatory polarization may occur in response to exogenous MTA. This response was observed consistent with varying concentrations of MTA (Fig. 4C).

Our results show that MTA and adenine can influence macrophage function when supplemented in vitro. However, as we described above, these metabolites can exert a significant effect in other stroma cell types (fibroblasts) leading to a strong molecular reprograming that includes the production of secreted factors. Based on these data, we hypothesized that studying the effect of these metabolites alone in macrophages would not allow us to understand the biological responses in a multicellular TME. To address this experimental constraint, we generated conditioned media from tumor cells and fibroblast cocultures exposed to MTA and assayed the effect of these conditioned media on macrophage function, as depicted in Fig. 4D.

We thus repeated the experiments in RAW cells, but in this case, rather than using pure MTA solutions, we chose to expose macrophages directly to the supernatants obtained after 24 h of incubating fibroblasts, HeLa (MTAP+), or MDA-MB-231 (MTAP−/−) tumor cells supplemented with 10 μM MTA, as conditioned media (CM). Molecular macrophage analysis (Fig. 4E) revealed that only the CM generated under coculture with MTAP-deficient tumor cells (HBF and MDA-MB-231) was able to induce an antiinflammatory phenotype, in contrast to the monocultures and the previous results in Fig. 4C. Besides, the coculture media induced a significant switch to antiinflammatory polarization, illustrated by the increase in Arg1 and the reduction in NOS2 mRNA expression (Fig. 4F). These data not only reinforce the former idea that the effects of secreted MTA can strongly depend on the environmental conditions, but also emphasize the fibroblast-tumor cell crosstalk as a potential actor behind the impaired immune activation in MTAP-deficient malignancies, as reported in clinical subjects. Furthermore, CM from MTAP-positive cells (namely, HeLa cells) were found to trigger activities analogous to those by their depleted counterparts when enriched with MTA, prior to fibroblast supplementation. This finding further underscores the relevant role of MTA over other cytokines present in the conditioned media of tumor cells, to reprogram the stroma (SI Appendix, Fig. S22).

Finally, we transitioned from the conditioned media approach to a multicoculture system using Transwell® setups, as depicted in Fig. 5A. To reinforce our results and transition to the entire human TME, we employed the human THP-1 cell line instead of the previously used murine RAW 264.7 cell line. As in the previous section, we recorded the SERS spectra of supernatants after 24 h of MTA supplementation under each condition. As shown in SI Appendix, Fig. S23, we could observe that THP-1 in isolation yielded a similar pattern as RAW 264.7 for exogenous MTA consumption. Moreover, the data in Fig. 5B allowed us to easily track fluctuations in relevant purine derivative metabolites (along with changes in other vibrational features), based on the specific THP-1 regimen in the Transwell® setups—monoculture or coculture with one or two different cell types. Notably, a pronounced peak at 735 cm⁻¹ persisted in the supernatant of the THP-1 coculture with fibroblasts, indicating underlying adenine accumulation. In Fig. 5C, we could also identify distinct signatures for each condition using t-SNE dimensional representation, which, as highlighted throughout this work, could also imply varying impacts on macrophage behavior.

Fig. 5.

Fibroblasts actively participate as mediators of the immune response to MTA. (A) Schematic representation of the Transwell®-based design for THP-1 monocytes culture with different cell populations. (B) SERS spectra of supernatants after 24-h MTA supplementation (10 μM) in a multi-co-culture system using Transwell® setups, comparing THP-1 human cell line (monoculture) to cocultures with one or two different cell types (HBF and MDA-MB-231). In the Right panel, the convolution analysis for the separation of the principal vibrations within 700–800 cm⁻¹ is displayed. (C) t-SNE bidimensional plot of SERS spectra collected from Transwell® different cell culture conditions. (D) TNF (proinflammatory marker) and CD163 (anti-inflammatory) normalized genes expression in THP-1 upon supplementation with the conditioned media as follows: control, HBF, MDA-MB-231, and their simultaneous coculture. RPS9 was used as housekeeping gene. (E) Illustration of macrophage polarization in an MTAP-deficient TME, highlighting the role of fibroblasts in facilitating metabolite exchange, promoting tumor growth, and supporting immune suppressive macrophage phenotypes. In contrast with the proinflammatory polarization when MDA-MB-231 cells, or MTA solely, were present in isolation without fibroblasts. Mean values (bars), SD (whiskers), each point represents an individual biological experiment. Asterisks indicate statistical significance in the one sample t test: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 compared to the control group.

This observation led to subsequent polarization experiments of THP-1 (Fig. 5D). TNF-$\alpha$   and CD163 are used here similarly to NOS2 and ARG1 in the RAW 264.7 cells; in other words, TNF-$\alpha$   serves as an indicator for proinflammatory polarization, while CD163 is associated with anti-inflammatory polarization (see SI Appendix, Fig. S24 for the validation test of such markers). As displayed in the Lower panels, higher levels of the CD163 marker were detected when fibroblasts were present (either in single or coculture) and supplemented with 10 μM of MTA, whereas a significant decrease in the expression was produced by the presence of either pure MTA or MDA-MB-231 cells. In contrast, higher TNF-$\alpha$   expression was observed with MDA-MB-231 cells and pure MTA, significantly decreasing when fibroblasts were incorporated to the culture. As reported in Fig. 5E, the metabolic interactions between fibroblasts and tumor cells lacking MTAP not only enable the latter to maintain their essential purine pools for proliferation, but also contribute to the development of an immunosuppressive niche. This, in turn, aids the tumor in the process to evade immune system surveillance and may explain the worse disease prognosis in cancer patients with such mutations.

Conclusions

The successful application of SERS in this study demonstrates that this technology could streamline our ability to rapidly capture metabolic interactions within complex environments. Indeed, the simple and rapid signal acquisition in SERS, along with its high sensitivity, meets the requirements to be a front-line tool that may subsequently orientate more targeted analyses. By following up with in-depth complementary techniques such as LC–MS, RNA-seq, and polarization studies, a comprehensive view of the TME metabolic state can be achieved. Importantly, we demonstrated the effective synergy between SERS and such other analytical methods. Together, they enabled the discovery of MTAP-driven metabolic interactions between tumor and stroma cells. In this manner, the role of fibroblasts in MTAP-deficient environments has been highlighted, showing their pivotal function in maintaining purine pools, potentially necessary for tumor cell proliferation, as well as contributing to an immunosuppressive TME. It is worth mentioning that, unraveling the complexity of such interactions in cancer patients could, in turn, pave the way toward novel therapeutic approaches, such as blocking the metabolic exchange between fibroblasts and tumor cells or modulate the MTA-mediated immune rewiring on case-by-case, or personalized basis.

Materials and Methods

Synthesis CTAC–Au Nanoparticles and SERS Measurements.

CTAC-capped gold nanoparticles were synthesized using a seeded growth technique, starting with gold seeds and carefully controlling the addition of growth solutions (44). The nanoparticles were then used to create plasmonic SERS substrates through drop casting. SERS measurements were performed on cell-derived liquid samples using these substrates and with no need for preliminary concentration or other pretreatment processes. The collected spectra underwent processing steps for data analysis, including removal of cosmic rays, baseline subtraction, normalization, and standardization. The fully processed datasets were then visualized and analyzed using t-distributed Stochastic Neighbor Embedding (t-SNE), partial least squares discriminant analysis (PLS-DA), and peak deconvolution.

Cell Culture Experiments.

This study utilized a diverse range of cell culture models, comprising cancer cells such as HeLa (cervical cancer), PC3 (prostate cancer), MDA-MB-231 (breast cancer), and U87 (glioblastoma), in addition to human fibroblasts (HDF) and macrophages (RAW 264.7 and THP-1). Experiments were performed under both isolated and coculture conditions, with different cell types being cocultured to study their interactive behavior. For macrophages, the effects of the treatments on gene expression were investigated using quantitative real-time PCR (qRT-PCR). Gene expression assays for Arginase 1, NOS2, TNF, and CD163 were specifically conducted to evaluate the macrophages’ polarization responses to different metabolites and culture conditions.

Detailed information about the fabrication of SERS substrates with CTAC-capped gold nanoparticles, the performance of cell experiments including the production of conditioned media, SERS measurements and equipment used, and complementary analysis by LC–MS, RNA-seq, and qPCR can be found in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All raw data from analytical techniques (SERS, RT-Q-PCR, LC-MS) for both Main and Supplementary figures have been deposited in ZENODO (https://doi.org/10.5281/zenodo.8130216) (45). RNA Sequencing source data isaccessible at the Gene Expression Omnibus (reference: GSE249082) (46).