Riboflavin-LSD1 axis participates in the in vivo tumor-associated macrophage morphology in human colorectal liver metastases

The study included fourteen patients who underwent surgery at the IRCCS Humanitas Research Hospital between 2020 and 2022 for CLM (Table 1). The study protocol, submitted to the international clinical trial registry (ClinicalTrials.gov registration number NCT03888638), was designed to identify macrophage-related morphological features associated with distinct clinical outcomes. Written informed consent was obtained from each patient included in the study. The study protocol was in accordance with the ethical guidelines established in the 1975 Declaration of Helsinki and compliant with the procedures of the local ethical committee of the IRCCS Humanitas Research Hospital (registration number 282/19).

Macrophages (TAMs) were FACS sorted from the peritumoral areas of surgically resected CLM of 14 patients. Single-cell suspensions were obtained by manually mincing tissue into small fragments and incubating them for 1 h at 37 °C in Hanks' Balanced Salt Solution (HBSS; Euroclone) with 1 mg/ml Type IV Collagenase (Sigma-Aldrich), 2% fetal bovine serum (FBS; Sigma-Aldrich), 50 μg/ml DNase I (Sigma-Aldrich), and 10 mM Hepes (Lonza). The resulting cell suspension was filtered through a 100-μm cell strainer and erythrocytes were lysed with 1X BD Lysing Buffer (BD Biosciences). Cells were then incubated with a blocking solution containing 1% human serum in saline solution and stained with the following fluorophore-conjugated primary antibodies: anti-CD45 (BD Biosciences; clone HI30), anti-CD11b (BD Biosciences; clone ICRF44), anti-CD16 (BioLegend; clone 3G8), anti-CD14 (BD Biosciences; clone M5E2), anti-CD66b (BioLegend; clone G10F5), and anti-CD163 (BD Biosciences; clone GHI/61). Fixable Viability Stain 700 fluorescent dye (BD Biosciences) was used for dead cell exclusion. SYTO 16 Green Fluorescent Nucleic Acid Stain (ThermoFisher) was used to identify nucleated cells. Large and small TAMs (20,000 cells/type) were FACS sorted on a FACSAria III (BD Biosciences) following the gating strategy as previously published [4]. The freshly sorted L- and S-TAMs were promptly subjected to snap-freezing with liquid nitrogen to maintain the metabolite profiles of the samples. Subsequently, they were stored at − 80 °C until the metabolomic analysis was carried out. The tumoral tissue of the same patients was digested using the MACS tumor dissociation kit (Miltenyi Biotec), according to manufacturer protocol. Briefly, the MACS tumor dissociation kit enzyme mix (300 μl) was added to each sample. Next, samples were put into the gentleMACS Dissociator and digested using the tumor program. The cell suspension was then applied to a 100-μm cell strainer. Cells were counted and cultured in RPMI 1640 (Euroclone), supplemented with 20% of FBS (Euroclone), 1% non-essential amino acids (NEAA, Lonza), 1% glutamine (Lonza), and 1% penicillin/streptomycin (Sigma–Aldrich) in a 37 °C, 5% CO2 incubator.

Buffy coats were obtained from anonymized healthy donors (Humanitas Hospital) approved for in-vitro research. Monocytes were isolated by density gradient centrifugation using pluriMate centrifuge tubes (Pluriselect), after incubation with anti-human CD14 M-pluriBead (Pluriselect), according to the manufacturer's instructions. Isolated monocytes were washed, resuspended in phosphate-buffered saline (PBS) 1x^−/− and seeded in 6-well plates for 1 h in an incomplete RPMI-1640 medium without FBS, at a concentration of 2 × 10⁶ cells for each well to promote their adhesion.

The culture medium of M2 macrophages was supplemented with tumor necrosis factor-alpha (TNFα; 20 ng/ml; PeproTech) and Transforming Growth Factor-beta (TGFβ; 10 ng/ml; PeproTech) for 18 h as previously reported [18].

Gene expression analysis was performed on freshly sorted and in-vitro differentiated macrophages by qRT-PCR. Briefly, mRNA was extracted using a commercial kit (Total RNA Purification Kit, NORGEN) and quantified using NanoDrop™ 2000c spectrophotometer (Thermo Fisher Scientific). Subsequently, 500 ng of mRNA were reverse transcribed in cDNA using the High-Capacity RNA-to-cDNA™ Kit (Applied Biosystems™). The qRT-PCR was performed on selected genes for M1 and M2 macrophages, using SYBR green PCR master mix (Applied Biosystems™) and custom-designed primers (Sigma-Aldrich), listed in Table S1. The qRT-PCR was carried out using a 7900HT Fast Real-Time PCR System (Applied Biosystems™). The experiments were carried out in triplicate for each condition. The gene expression was normalized using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as housekeeping and determined using the 2^−ΔCT or 2^−ΔΔCT method.

Metabolites in L- and S-TAMs pellets (n = 20.000 cells/condition) were extracted by adding on the top of the cells pellets 25 µL cold MeOH (1:4), suspensions were incubated for 20 min at − 80 °C, and then centrifuged at 13,000 g × 15 min. The supernatant was stored at − 80 °C until the untargeted and targeted metabolomics analysis.

Flow Injection Analysis High-resolution mass spectrometry (FIA-HRMS) was used for untargeted metabolomics [19]. A portion of the metabolites extract (8 μL) was analyzed by -Orbitrap QExactive Mass Spectrometer (ThermoFisher Scientific) equipped with an electrospray source operated in negative and positive modes. Each run was carried out by injecting 8 μL of sample extract at a flow rate of 50 μL/min (Agilent 1200 Series) of mobile phase consisting of isopropanol/water (60:40, v/v) buffered with 5 mM ammonium at pH 9 for negative mode and methanol/water (60:40, v/v) with 0.1% formic acid at pH 3 for positive mode. Reference masses for internal calibration were used in continuous infusion during the analysis (m/z 210.1285 for positive and m/z 212.0750 for negative ionization). Mass spectra were recorded from m/z 50 to 1.000 with 60.000 resolutions. The source temperature was set to 240 °C with 25 L/min drying gas and a nebulizer pressure of 35 psig. MS/MS fragmentation pattern of the significant features was collected and used to confirm metabolite identity. All data processing and analysis were done with MATLAB R2016a (The Mathworks) using our in-house developed script [20].

The targeted approach investigated the abundance of 83 metabolites (32 amino acids and derivatives, 19 nucleic acid-related compounds, 13 vitamins, 4 sugars, and 18 intermediates of glycolysis and tricarboxylic acid cycle (TCA cycle) by using liquid chromatography coupled with a triple quadrupole mass spectrometry system (LCMS-8060, Shimadzu). One (1) μL of metabolites extract was injected into a Discovery HS F5-3 (2.1 mm I.D. × 150 mm, 3 µm) column (Sigma-Aldric) using a 20-min gradient from 0 to 95% B (Acetonitrile), A (10 mM NH₄HCO₂ pH 3.5) at 350 µL/min. The LCMS-8060 mass spectrometer was equipped with an ESI source operating in both positive and negative ion and selected reaction monitoring (SRM) modes. The transitions identified during the optimization of the method are reported in Table S2. The MS settings were as follows: nebulizing gas flow rate: 3.0 L/min; drying gas flow rate: 15.0 L/min; DL Temperature: 250 °C; block heater temperature: 400 °C. Peak areas were automatically integrated using LabSolution Insight LC–MS (Shimadzu). Retention times and area ratio between quantifier and qualifier ions of all metabolites of interest were validated using pure standards.

Lysine-specific demethylase 1 (LSD1) enzyme activity was evaluated using the Abcam KDM1/LSD1 Activity Quantification kit (ab113459, Abcam) according to the manufacturer’s protocol. Briefly, LSD1 protein was extracted from both the nuclei of S-L TAM pairs (8 subjects) and in-vitro M1 and M2 polarized monocytes. Nuclei were isolated using a Qproteome Nuclear Protein Kit (Qiagen) following the manufacturer’s protocol. Soluble protein extracts were then subjected to the KDM1/LSD1 activity kit.

Formalin-fixed and paraffin-embedded specimens (2-μm thick tissue section) of CLM and peritumor tissue were deparaffinized and rehydrated in PBS. Antigen retrieval was performed by heat treatment using an ethylenediaminetetraacetic acid (EDTA) buffer (Dako; 0.25 mM, pH 8) in a water bath at 98 °C for 20 min. After washing with PBS, endogenous peroxidases were blocked via incubation with the Peroxidase Blocking Solution (Dako) for 15 min at room temperature (RT), and subsequently to block nonspecific binding, the slides were incubated with Background Sniper (Biocare Medical) for 20 min. After washing with PBS, the sections were then incubated with the primary antibody anti-human CD163 (Leica Biosystems, 10D6 clone, diluted 1:200) overnight at RT. The slides were washed twice with PBS and then incubated with the detection system EnVision + System HRP-labelled anti-mouse (Dako) for 1 h at RT. Following this, diaminobenzidine tetrahydrochloride (Dako) was used as a chromogen to visualize the positive cells. Nuclei were lightly counterstained with a Haematoxylin solution (Dako). The sections were dehydrated and mounted with a mounting medium (Eukitt).

The sorting strategy described by Donadon et al. was applied in this work to isolate large and small macrophages from the human peritumoral liver tissue of patients surgically resected for CLM [4]. The morphological assessment confirmed that S-, and L-TAMs were characterized by a different expression of CD163 (Fig. S1A), moreover, the distinct flow cytometry scatter plots of S-and L-TAMs characteristics resemble their morphologic aspect as previously reported for CLM patients (Fig. S1B), where L-TAM correlated to worse prognosis [4, 9].

To identify TAM features associated with the two distinct macrophage subpopulations, we conducted a wide metabolic profiling (untargeted and targeted mass spectrometry-based metabolomics analyses) on freshly isolated L- and S-TAM pairs derived from the 14 CLM patients. This analysis was aimed at elucidating the possible metabolic alterations underlying the in vivo TAM polarization and morphology in CLM. Due to difficulties in macrophage isolation from freshly resected peritumoral tissue, 20.000 cells were identified as the adequate number to perform a complete metabolic analysis. After cell sorting, the S- and L- TAMs were analyzed by untargeted profiling using the HMDB database. We identified 1380 and 270 m/z features in positive and negative ion mode respectively of whom 14 were statistically different (Wilcoxon Mann–Whitney test, p values < 0.05) in their abundance between the two macrophage groups (Fig. 1A). Metabolites identity by MS/MS analysis was confirmed for 7 of them (Table S3): one amine (Histamine), three amino acids and derivatives (guanidoacetic acid, alpha-aminobutyric acid, and glutathione), one phosphate ester (O-Phosphoethanolamine), one carboximidic acid (N8-Acetylspermidine) and one glycerophosphocholine. Furthermore, the 7 statistically significant different m/z features, corresponding to lipid species, the identification could not have been established, given the existence of isomeric species with indistinguishable fragmentation spectra and the lack of reference standards. For these molecules, it was only possible to attribute the metabolic class to which they belong and not the exact chemical identity. The identified classes are two diacylglycerols (DG), two ceramide phosphates (CerP), and two phosphatidylcholines (PC) (Fig. 1A, Table S3). We further investigated, by using a targeted metabolomics approach, alteration in metabolites belonging to central cellular pathways such as glycolysis, the TCA cycle, the pentose phosphate pathway (PPP), and the urea cycle. The targeted approach quantified 48 out of the 83 measurable metabolites (Table S4) and indicated that L- and S-TAMs populations had a comparable central cellular metabolism with no substantial difference in the presence, an abundance of metabolites related to glycolysis, TCA cycle, urea cycle. We observed that only 4 metabolites, two purine nucleosides (Adenosine, Guanosine), a water-soluble vitamin (Riboflavin), and a non-proteinogenic amino acid (Ornithine) statistically discriminated (Wilcoxon Mann–Whitney test, p values < 0.05) between L- and S- TAMs populations (Fig. 1B, Table S4). No significant enrichment of specific biochemical pathways was found by using MetaboAnalyst (a web server for metabolomics data interpretation) on merged untargeted and targeted significantly different metabolites.

Comprehensively, metabolomics profiling indicates that L-TAMs were characterized by an increased abundance of the majority of deregulated metabolites except a few lipid species decreasing relative to the S-TAMs (Fig. 1C). To note, the arginine catabolism, used to generate ornithine and guanidoacetic acid, was the only metabolic circuit representative of the L-TAMs relative to the S- counterparts, thus supporting the role of such metabolic phenotype already ascribed to TAMs metabolism (Fig. 1D) [21].

According to these results, we hypothesized that the L- and S-TAM subpopulations have a similar central metabolic landscape with few discriminating metabolic points that only partially recapitulate the metabolic observation obtained in the in vitro induced TAM populations.

The correlation analysis between TAM morphologies and the statistically significant metabolites (from untargeted + targeted metabolomics) pointed out that riboflavin was the only metabolite with the strongest positive correlation with L-TAM morphology (Spearman r: 0.7732; p value: 0.0022), while the remaining metabolites showed moderate significant correlation (Spearman r: 0.5–0.6) values (Table 2). Histamine, O-phosphoethanolamine, glutathione, and L-alpha-aminobutyric acid were excluded from correlation analysis because their abundance obtained with the untargeted approach was not confirmed through targeted analysis. Consistently, Riboflavin was also found to be the metabolite with the strongest association with TAM morphologies (linear regression; standardized beta: 0.721; t = 3.752; p value = 0.002).

Riboflavin, also known as Vitamin B2, has an important role in cell metabolism linked to the generation of flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD) cofactors. In literature, riboflavin has been described to be able to directly interact with the flavoprotein LSD1 (lysine-specific histone demethylase) by itself or FAD (Fig. 2A), thereby influencing macrophages polarization toward the M2-like phenotype [22, 23]. In agreement with the increased riboflavin abundance, L-TAMs showed higher activity of LSD1 protein compared to S-TAMs (Fig. 2B). To corroborate this data, a qRT-PCR was performed on S-TAMs and L-TAMs sorted from the peritumoral samples (n = 4). The LSD1 expression was found to be significantly increased in the L-TAMs, compared to S-TAMs (Fig. 2C).

To understand the interplay between the Riboflavin-LSD1 axis and the different activation states of macrophages, the monocytes were differentiated in vitro into M1 and M2 macrophages. We observed a slightly higher expression and activity of LSD1 in the M2 macrophages, despite not being statistically significant, compared to the M1 macrophages (Fig. 2D). Overall, our data supported the involvement of the Riboflavin-LSD1 axis in driving the in vivo L-TAMs morphologies.

To investigate how the interaction between tumor cells and macrophages may impact LSD1 expression and TAM morphologies, we analyzed the expression of inflammatory cytokines TNFα and TGFβ, which are released by tumor cells within the tumor microenvironment [24].

According to the immunohistochemistry criteria previously reported [4, 9], we then selected 3 patients exhibiting high density of L-TAMs and 3 patients exhibiting low density of L-TAMs and high density of S-TAMs to isolate the corresponding primary tumor cells (Fig. 3A). Of note, the baseline demographic and tumor characteristics of these two groups of patients were statistically similar (Table S5), while we found an increase of TNFα and TGFβ in those patients with increased density of L-TAMs (Fig. 3B) with no differences in the LSD1 expression among these primary tumor cells (Fig. 3C).

To elucidate whether the TNFα and TGFβ released from primary tumor cells could be involved in the determination of L-TAMs morphology, the culture medium of the in vitro-polarized M2 macrophages was supplemented with TNFα and TGFβ for 18 h. As shown in Fig. 4A, M2 macrophages cultured in the presence of TNFα displayed a significantly higher LSD1 expression, compared to those exposed to TGFβ, suggesting the crucial role of TNFα in the upregulation of LSD1 in L-TAMs. Furthermore, when exposed to TNFα, M2-induced macrophages exhibited a significant upregulation of LSD1 expression compared to M1 macrophages, mirroring the observed differences in the LSD1 expression between S- and L-TAMs (Fig. 4B).

To gain insight into the biochemical link between riboflavin and LSD1 in the presence of inflammatory stimuli, we evaluated the expression of the two riboflavin membrane transporters, namely Solute Carrier Family 52 Member 2 (SLC52A2) and SLC52A3.

In agreement with the LSD1 upregulation, the M2 macrophages exposed to TNFα showed increased expression of SLC52A3 and SLC2A2 (although the latter did not reach statistical significance) compared to M2 macrophages cultured with TGFβ (Fig. 4C). Additionally, similar to the findings observed for LSD1 expression, TNFα-treated M2 macrophages exhibited a significantly increased expression of SLC52A3, compared to M1 macrophages (Fig. 4D).

These results suggest a collaborative role of Riboflavin and LSD1 in shaping the morphology of the L-TAMs within an inflammatory microenvironment influenced by tumor cells. Moreover, it is suggested that SLC52A3, one of the riboflavin membrane transporters, plays a direct role in the intracellular uptake of riboflavin, thereby contributing to this mechanism.