Age-diet interactions significantly influence intratumoral gene expression, gut microbiome signature and tumor microenvironment in colorectal cancer.

Introduction
According to Cancer Statistics 2025, Colorectal cancer (CRC) is one of the leading causes of cancer-related deaths worldwide with the estimated number of new cases and deaths in United States due to CRC predicted to be 154,270 and 52,900, respectively. Of these estimated cases, approximately 87 % are expected to be late-onset (LO-CRC), ≥50 years old, with the remaining 13 % being early-onset (EO-CRC), <50 years old, at the time of diagnosis. Studies on LO-CRC have led to significant advances in its detection, understanding, and treatment resulting in an annual decrease of 1.3-1.5 % in incidence rates since 2011. Nevertheless, within the same period, EO-CRC observed an annual rise in cases of 1-2 %, highlighting the inefficacy of current treatments to address its unique attributes [1].
Compared to LO-CRC, EO-CRC exhibits distinct clinical and molecular characteristics suggesting that it may be a distinct disease entity. While hereditary conditions including Lynch syndrome, familial adenomatous polyposis, MUTYH-associated polyposis, and hamartomatous polyposis syndromes constitute 30 % known hereditary cancer syndromes or de novo hereditary cancer mutations, 20 % familial CRC, and 50 % are sporadic CRC [2]. The rise in EO-CRC incidence, specifically sporadic cases, may be attributed to generational differences in exogenous and endogenous factors such as environmental exposures, lifestyles, diets, gut microbiomes, and immune systems. Yet despite these known differences, the similar therapeutic recommendations for EO- and LO-CRC reflects a significant gap in the translation of our understanding of these factors into effective practices that address the distinct clinical and molecular features seen in both EO- and LO-CRC [3]. Therefore, dedicated efforts to investigate the molecular contexture and biology of cancer among different age groups are required to advance our current understanding of the mechanisms by which lifestyle, diet, and aging impact cancer onset, progression, metastasis, and therapeutic response.
CRC is a multifactorial disease caused by the interactions of various environmental, genetic, and epigenetic factors. Among potential etiologies, age and diet are two of the strongest risk factors for CRC incidence [4]. The correlation of advanced age with increased cases of CRC is one of the most well-established axioms, yet the risk of CRC varies greatly between similar-aged individuals due to the influence of additional intrinsic and extrinsic factors [5]. The Western lifestyle and its associated dietary habits have been shown to be strongly linked with increased risk of CRC, particularly in distal and rectal tumors [6]. Among its dietary components, fat intake has been notably associated with an elevated CRC risk, yet the underlying mechanism remains unknown [7]. The high fat (HF) diet has also been shown to be a driver of CRC tumorigenesis by inducing gut microbial dysbiosis [8]. In contrast, calorie restriction (CR) has been observed to reduce tumor growth rates in several cancers, while also being associated with reprogramming of gut microbiota structure and amelioration of gut microbial dysbiosis in CRC [9].
There is a consensus that gut microbiota composition changes with age [10]. The process by which bacterial species colonize the gut to create the microbiome begins in utero and continues differentiating at birth [11,12]. As individuals age, exposure to changing environments, including dietary patterns, increases individual microbial diversity [13]. This emphasizes the sensitive nature of microbiota population fluctuations in response to extrinsic influences. Emerging evidence over the past decade has suggested gut microbial dysbiosis as a major player in CRC development and a potential diagnostic and therapeutic biomarker. The gut microbiota interacts with the colonic epithelia and host’s immune cells, releasing metabolites, proteins, macromolecules, etc. which regulate CRC tumorigenesis. Consequently, gut microbiota has been shown to modify the therapeutic responses to chemotherapy and immunotherapy in CRC [14].
As age increases, somatic mutations in the mitochondrial DNA of human colorectal epithelium accumulate leading to defects in oxidative phosphorylation prevalent in colorectal tumors; however, their active role in tumorigenesis is largely unknown [15]. Mitochondrial DNA has recently been shown to express microproteins encoded within the larger protein coding and rRNA genes, including humanin (HN), Mitochondrial ORF of the 12S rRNA Type-C (MOTS-c), Small humanin like peptide 2 (SHLP2) and SHMOOSE. These peptides have been shown to have a variety of biological actions that can modulate tumor progression [[16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26]]. Despite their identification, the effect that age, diet, and the tumor microenvironment (TME) have on the expression of these microproteins in tumors has not been previously assessed.
In this study, we evaluated the effects of age on CRC tumor biology using syngeneic mice models with a focus on specific differences in tumor growth patterns, role of TME and mitochondrial proteins, and dietary interactions in young and old mice harboring CRC allografts. The tripartite association between age, diet and the gut microbiota composition in CRC has not been well studied. Therefore, we further aimed to determine the changes in bacterial composition and signaling pathway abundance of the gut microbiome in response to age-diet interaction in CRC mice model.

Materials and methods

Animal experiments
Wildtype male C57BL/6J mice aged 2.5 months were purchased from Jackson Laboratories (RRID: IMSR_JAX:000664) and male C57BL/6JN mice aged 25-29 months were obtained from National Institute of Aging (NIA). Mice were housed under standard laboratory conditions with food and water provided ad libitum, maintained under a12-hour light-dark cycle at 22ºC in animal facility. Animal handling and procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Southern California (USC). All experiments were performed in accordance with approved protocols. 5×105 MC38 (Kerafast Cat# ENH204-FP, RRID: CVCL_B288) cells, suspended in 100ul 1X PBS, were subcutaneously injected into the right flanks of young (2.5 months; N=22) and old (25-29 months; N=19) C57BL/6J and C57BL/6JN mice, respectively. In both old and young cohorts, mice were randomized into three different diet groups: normal (NR; standard chow), high-fat (HF), and calorie-restricted (CR; 30 % reduction in total calories). Mice were housed individually to allow for daily food intake measurements. Tumor measurements were taken using calipers from day 3 until the survival endpoint (tumor size exceeding 1.5 cm length/breadth or visible morbidity symptoms) was reached. Tumor volumes were calculated using formula: Length X Breadth2/2. Once at endpoint, fecal samples were collected before mice were euthanized for isolation of blood and tumor samples.

RNA sequencing
The collected tumor samples were sent to the USC Molecular Genomics Core for RNA sequencing service. Using the Poly(A) RNA selection method, mRNA (eukaryotic) was selected for library preparation. TruSeq stranded total RNA kit (Illumina) was used to achieve cytoplasmic rRNA depletion while retaining mitochondrial rRNA. Prepared libraries were sequenced with Illumina Hiseq NGS platform and output reads were processed for differential gene expression analysis by trimming, alignment, normalization, and quantification. Raw sequencing reads were trimmed using trimmomatic v0.39, aligned to GRCm39 using STAR v2.7.10a, and quantified using featureCount v2.0.1. Initial analysis of differential gene expression was conducted using DESeq2 v1.34.0. Genes with an absolute log2 fold-change greater than one (|log2FoldChange|>1) and adjusted p-value below 0.05 (q<0.05) were classified as significantly differentially expressed genes (DEGs). Ingenuity Pathway Analysis was used to assess enriched pathways based on identified DEGs.

Analysis of mitochondrial-derived transcripts
We identified over 700 putative murine mitochondrial small open reading frames (smORFs) which were analyzed using FASTQ files generated for RNA sequencing analysis. Aligned sorted BAM files and a GTF file for potential mitochondrial smORFs were inputted into R (v4.2.2) to count the mitochondrial smORFs. Count matrices were generated using the summarizeOverlaps function of the GenomicAlignments (v1.34.1) package in R and used for the downstream analyses. Differentially expressed mitochondrial transcripts between each group were analyzed using the DESeq2 Package in R with a cut-off adjusted p-value of 0.05.

Measurement of Mouse pro-inflammatory biomarkers
Plasma levels of inflammatory biomarkers, including IFN-γ, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-10, IL-12p70, KC/GRO, and TNF-α, were measured by MSD proinflammatory panel I kit (Catalog no: K15048D-1, Rockville, Maryland) according to manufacturer’s instructions. Briefly, 50 µl of prepared standards and samples were added into each well and incubated for 2 hours on a shaker. After 3 PBS washes, 25µl of detection antibodies were added into each well and incubated for 2 hours on a shaker. The plate was washed 3 times, followed by the addition of read buffer into each well before being analyzed on the MSD Quick plex SQ120.

Fecal microbiome analysis
Collected fecal samples were processed using the ZymoBIOMICS® Shotgun Metagenomic Sequencing Service (Zymo Research, Irvine, CA). Initial DNA extraction was performed with the ZymoBIOMICS®-96 MagBead DNA Kit (Catalogue no.: D4302 Zymo Research, Irvine, CA) based on sample type and volume, according to the manufacturer’s instructions. Sequencing libraries were prepared with the Nextera® DNA Flex Library Prep Kit (Catalogue no.: 20018704; Illumina, San Diego, CA) using up to 100 ng of DNA input tagged with internal dual-index 8 bp barcodes with Nextera® adapters (Illumina, San Diego, CA) based on the manufacturer’s protocol. All libraries were quantified using TapeStation® (Agilent Technologies, Santa Clara, CA) and pooled in equal abundance. The final pool was quantified using qPCR before being sequenced on the NovaSeq® (Illumina, San Diego, CA) sequencing platform. Raw sequence reads were trimmed to remove low quality fractions and adapters with Trimmomatic-0.33 [27]: quality trimming was done using a sliding window with a 6 bp window size, quality cutoff of 20, and minimum read size of 70 bp. Antimicrobial resistance and virulence factor gene identification was performed with the DIAMOND sequence aligner [28]. This study employed MetaPhlAn 4 (https://huttenhower.sph.harvard.edu/metaphlan/) for taxonomic profiling, HUMAnN3.6 (https://huttenhower.sph.harvard.edu/humann/) for functional pathway analysis, and QIIME 2 (https://qiime2.org/) for microbiome composition comparisons. Differential feature abundances were analyzed using ANCOM-BC (https://docs.qiime2.org/2024.10/plugins/available/composition/ancombc/), which involved log ratio transformation, bias correction, and statistical testing. Multiple comparisons were adjusted using the Holm method to control the false discovery rate. Beta diversity was assessed using Bray-Curtis and Jaccard dissimilarity measures with Pairwise PERMANOVA, while alpha diversity was evaluated using the Shannon Diversity Index with the Kruskal-Wallis test.

Statistical analysis
All statistical analyses were performed using Prism9 software (GraphPad Software, San Diego, CA). The value of P ≤ 0.05 was considered statistically significant, with *P < 0.05; **P < 0.01; and ***P < 0.001; ****P < 0.0001.

Results

Effect of age and diet on tumor growth, body weight and food intake
We implanted mismatch repair deficient (dMMR)/microsatellite instability-high (MSI-H) CRC MC38 cell line in C57BL/6J mice to create a syngeneic mouse model, to assess the effect of host age and diet on tumor growth. The young 2.5-month-old mice represented EO-CRC, while the older (25-29)-month-old mice represented an LO-CRC model. Within each cohort, mice were fed with one of three different diets: HF, CR, or NR (Fig. 1). Our findings revealed that tumor growth rates were influenced by both diet and chronological age. Overall tumor growth trends suggested that the CR diet was the most effective at reducing growth rates in both young and old mice cohorts. Young-CR mice displayed significantly slower tumor growth compared to both Young-NR (P < 0.01) and Young-HF (P < 0.001) groups (Fig. 1B). Similarly, Old-CR mice had significantly reduced growth compared to the Old-NR (P < 0.01) and Old-HF (P < 0.0001) group (Fig. 1C). There was no significant difference in food intake (Fig. 1G) or body weight (Fig. 1H) in correlation with age or diet throughout the study period.

Transcriptomic profile based on age and diet
Studies have shown that EO-CRC presents a distinct molecular profile when compared to LO-CRC [29,30]. Additionally, the nutrients present in various diets impact gene expression, transcription factor binding, and post-translational modifications throughout the body [29,30]. Utilizing this knowledge, we combined both the aging and dietary-related gene signatures to investigate their influence on CRC. To determine the concurrent effect of age and diet on transcriptional regulation in the tumor, we compared the expression profiles generated by bulk RNA sequencing in young and old mice fed with either NR, CR, or HF diets (Fig. 2, Fig. S1-S2). Comparison between Old- and Young-CR mice revealed upregulation of genes associated with immune escape, including Cd163, Fcrl6, Ctsg in old mice. Within the same comparison, gene ontology identified antigen processing and presentation of exogenous peptide antigen, MHC class II protein complex assembly, positive regulation of cell-cell adhesion, TNF superfamily cytokine production, and T cell activation as some of the top significant pathways. Interestingly, in the Old- vs. Young-HF top DEGs list, 15 oncogenes were upregulated in Old-HF as opposed to only two in Young-HF. Most of the upregulated genes (Sptb, Rarres2, Spib, etc.) identified in the Old-HF mice were associated with cell proliferation, invasion, and chemosensitivity. The top pathways/biological processes associated with the Old- vs. Young-HF comparison were endopeptidases, proteolysis, chemotaxis, ECM remodeling, and metastasis (Fig. S3-S6). Similarly, Old- vs. Young-NR identified most upregulated oncogenes within the Old-NR group. The top pathways/ biological processes in Old- vs. Young-NR included Meiotic cell cycle, H3-K9/H3-K14 acetylation, P granule/germplasm/poleplasm, chromocenter, and desmosome. However, there were less genes upregulated in old-NR compared to old-HF.
IPA analysis of DEGs showed that in both old and young CR subgroups, the canonical pathways related to immune response, especially T-cells, were activated (Fig. S7). The old vs young HF analysis showed activation of bile acid key regulator FXR/RXR and cytoplasmic tyrosine kinase FAK signaling involved in cell movement, migration of phagocytes in the HF subgroup. Briefly, diet-based comparisons revealed (i) Apolipoprotein E (Apoe), which plays key role in lipid metabolism, was found as the hub gene in Old-CR vs. Old-HF, (ii) predicted activation of genes Agt, Smarca 44, Ccr2, Tgfb11, Ikbkb and Il1B was found in Old-HF vs. Old-NR, (iii) predicted activation of Agt, Rictor, Il10, Bcl6 and Vip; and inhibition of Myc, Mycn, Cd40, App and Kzf1in Old-CR vs. Old-NR, (iv) predicted activation of Dusp1, Il10, Nr3c1 and Ar genes and inhibition of Jun, Tnf, App, Atf4, and others was found in Young-CR vs. Young-HF.

Comprehensive analysis of cytokines shows age-diet influence on tumor microenvironment
To study the relative contribution of age and diet on the tumor immune microenvironment in CRC, we measured the plasma levels of biomarkers including IFN-γ, IL-1β, IL-2, IL-5, IL-6, IL-10, IL-12p70, TNF-α, chemokine KC/GRO, mitochondrial proteins HN, and MOTS-c, associated with inflammatory response and immune system regulation (Fig. 3). TNF-α and IL-2, which play pro-tumorigenic roles in CRC, were significantly lower in Young-CR compared to Young-HF and Young-NR groups. TNF-α was also significantly higher in the Old-NR control group compared to the Young-NR control group. While not significant, TNF-α showed visual reduction in the Old-CR group compared to Old-HF and Old-NR. IL-5, a tumor suppressive cytokine in CRC, had no significant difference in young mice cohort based on diet comparison, but it was significantly lower in Old-CR group compared to Old-HF and Old-NR. Similarly, pro-tumorigenic and inflammatory IL-12p70 had no significant association with diet in the young mice cohort but was higher level in Old-CR compared to Old-HF. Proinflammatory chemokine KC/GRO was significantly lower in Young-CR compared to Young-NR mice. Mitochondrial proteins HN and MOTS-c were significantly higher in Young-NR compared to Young-HF and Old-NR mice. Additionally, cytokine comparison showed that pro-inflammatory and CRC tumor promoting cytokines IFN-γ, IL-2, IL-6, and IL-12p70 showed significantly higher levels in old mice compared to young mice. Overall, the levels of several pro-inflammatory cytokines associated with CRC tumor initiation and progression were higher in the old mice cohort.

Analysis of putative murine mitochondrial small open reading frames (smORFs)
We analyzed over 700 putative murine mitochondrial smORFs that potentially encode mitochondrial microproteins. In young mice, HF dramatically reduced multiple smORFs, while CR did not affect smORF expressions. In old mice, CR changed a few smORFs, however, there were no significant impacts on smORF expression levels. Interestingly, aging significantly decreased the expression levels of multiple mitochondrial smORFs in the normal diet while there was no notable difference in the CR and HF age groups (Fig. S8).

Alterations in gut microbiota induced by age and diet in colorectal cancer progression
Fecal microbiome analysis suggested higher alpha diversity in old mice compared to young mice. Though not significant (q-value: 0.07), the visual trend is notable and may have been limited due to small sample size (Fig. 4B). In comparing diets, no statistically significant differences were found in microbiome diversities (Fig. 4C). However, the Bray-Curtis emperor plot presented noticeable clustering of HF group (blue) as well as separate clustering between the microbiomes of young samples (solid) and old samples (ring) (Fig. 4D). Beta-diversity analysis using pairwise PERMANOVA showed statistically significant differences in the microbiomes of old vs. young (q-value: 0.001), CR vs. HF (q-value: 0.0315), and HF vs. NR groups (q-value: 0.0315) (Fig. 4C). ANCOM-BC analysis was used to determine the differentiating features and potential microbial functional pathways in each comparison group, with relative abundance differences over 100-fold (Supplementary Table S1-S4). Specific species of interest included Bacteroides thetaiotaomicron (P value: 1.43E-39) enriched in young mice and Parabacteroides goldsteinii (P value: 2.20E-23) enriched in old mice (Fig. 4F). Additionally, there was significant enrichment of Akkermansia muciniphila (P value: 0.008) in NR compared to CR (Fig. 4H), Lactococcus lactis (P value: 9.94E-160) in HF compared to NR, and Lachnospiraceae bacterium A4 (P value: 4.95E-10) in NR compared to HF (Fig. 4I). Looking at functional pathways, CMP-legionaminate biosynthesis I (P value: 2.38E-13), Thiamine diphosphate salvage II (P value: 3.52E-05)) and thiamine phosphate formation from pyrithiamine and oxythiamine (P value: 1.42E-04) were the most significantly enriched in young, while L-arginine biosynthesis IV (archaebacteria) (P value: 4.77E-11), L-citrulline biosynthesis (P value: 1.18E-29) and urea cycle (P value: 1.58E-19) were the most significantly enriched in old group, respectively (Fig. 4G, Table S3). Based on diet, the menaquinol biosynthesis pathway was found to be highly enriched in the HF group, and Clostridium acetobutylicum acidogenic fermentation pathway was enriched in CR group, compared to the NR group, respectively.

Discussion
Age has long been established as a major contributing risk factor influencing CRC incidence and mortality. However, recent reports of increasing case rates of EO-CRC in younger individuals have highlighted the importance of potential interactions between age with other risk factors [31]. Accumulating evidence has shown that LO- CRC and EO-CRC present as two molecularly distinct diseases in terms of differences in oncogenic mutation frequency and DNA methylation profile [29,32]. EO-CRC tends to be more distally located [33], histologically mucinous, and have poorly differentiated signet [34]. With reduced age, other lifestyle factors such as unhealthy diet, metabolic disorders and obesity are often much more involved as risk factors for EO-CRC [35]. Studies have shown that EO-CRC patients have higher triglycerides and lower high-density lipoproteins cholesterol levels when compared to LO-CRC [36]. HF diets have been connected to increasing the risk of EO-CRC while CR induced weight loss regulated the CRC predisposition in mice, in settings equivalent to human physiological ones [37]. With consideration to the existing pool of literature, this study aimed to decipher the mechanism of age-diet interactions and their potential influence on gut microbiota in CRC mice models. Our approach focused on CR and HF primarily due to CR being the most common approach for weight loss and HF being associated with western lifestyle diet. In our study, CR diet-fed mice showed an overall lower body weight relative to other diet groups. Additionally, CR-fed mice also showed lower tumor volume and slower growth rates compared to other groups, irrespective of age, supporting the antitumorigenic role of a healthy diet in cancer. The young mice (2.5 months), equivalent to young humans <50 years-old, showed increased tumor growth rates when compared to their old (25-29 months, >50 human years) counterparts in the HF and NR matched comparisons. This supports previous findings that EO-CRC presents a more aggressive form of CRC and suggests that EO-CRC is more influenced by HF diets than LO-CRC to promote tumor growth. These differences may be attributed to the bile acid metabolism or gut microbiota [38], which differ by age, pointing to a need for further investigation. Multiple studies have shown that diagnosed obesity at a young age is correlated with metabolic switches that support long chain fatty acids in colon cells responsible for stem cell like proliferation, which can lead to accumulation of oncogenic mutations and thus increasing the number of cells with tumorigenic potential over the years [39]. Our old-HF mice cohort showed higher expression of oncogenes compared to other groups, which is consistent with previous studies suggesting that obesity within aged individuals is associated with decreased expression of tumor suppressor genes and negative regulators of pro-tumorigenic signaling pathways [40].
Nuclear transcriptomic studies showed that mice in the CR group exhibited inhibition of the Myc target pathway. Studies have shown that Myc is activated by secondary bile acids associated with HF diet and leading to higher CRC cell proliferation [41]. Our interactome analysis identified FXR (farnesoid X receptor) /RXR (retinoid X receptors) signaling activation within the HF diet. This is consistent with previous reports [30,42] of a potential association between FXR signaling and gut –liver crosstalk playing a role in colonic crypts proliferation induced by a HF diet. The trend in our data suggests that the HF-related molecular and pathophysiological changes in colonic epithelium may prevented, or reversed, through CR. Earlier implementation of CR, along with the long-term weight loss effect, could rewire the signal transduction network to avoid the malignant colonic epithelial cells transformation [40].
Apolipoproteins, involved in lipid trafficking and metabolism, have been shown in recent studies to play a role in tumorigenesis [43]. Our study identified Apolipoprotein E (Apoe) as a hub gene in gene network analysis between the Old-CR and Old-HF groups. Apoe has previously been suggested to play a role in the gut microbiota-brain axis, wherein Apoe deletion showed aggravated cognitive dysfunction and reduction in the gut microbiome makeup in aging mice [44]. This may lead to Apoe being a driving gene in the trending difference in young and old mice. APOE has also been shown to have an important role in tumorigenesis and cancer progression, cell proliferation, angiogenesis, and metastasis. APOE-overexpression in HCT116 and HCT8 CRC cell lines had been shown to promote migration and invasion through jun-APOE-LRP1 axis, thus confirming its metastasis promoting function [45]. APOE overexpression has been associated with poor prognosis in CRC patients [46]. Our previous study also showed that in CRC patients enrolled in the CALGB/SWOG 80405 trial, APOE-high expression showed significantly shorter progression free survival (logrank P = 0.0021) and overall survival (P = 0.0017) compared to APOE-median and -low expression, respectively [47]. Based on our network analysis, the decrease in Apoe within the Old-CR may be indirectly driving the inhibition of angiogenesis and cell migration, but further investigation is needed.
Systemic cytokine profiles are regularly reported as important markers in CRC progression and prognosis. Our study assessed the effect of age-diet on cytokine profiling in CRC mice models. Studies have indicated that the unique immune microenvironment in EO-CRC may be involved with greater CRC progression compared to LO-CRC. Furthermore, immune aging, especially inflammaging, is accelerated in EO-CRC relative to what is seen in older populations [48]. Diets, such as CR, have been considered to reform immune response to cancer [49] and may have potential to delay immunological senescence by regulating gut microbiome through reduction of effector memory CD8+ T cells and memory B cells in mice [9,50]. Consistent with previous work, our data also showed that multiple top 30 DEGs in CR-fed Old vs. Young groups were associated with roles in the tumor immune microenvironment. Our cytokine profiling showed that tumor promoting TNF-α was significantly reduced in mice fed with CR diet when compared to HF. Diets including those high in intake of red and processed meats had been reported to interact with TLRs and NF-κB pathways encoding genes leading to development of CRC and can regulate TNF treatment response [51,52]. Previous studies have shown that various diets have a significant effect on anti-TNF drug efficacy through regulation by the microbiome composition and host genetic makeup to modify immune response [53]. Thus, based on previous literature and our current findings, a CR diet may hold potential as a new strategy to improve anti-TNF treatment response. Furthermore, as mentioned above, in our study, IPA of DEGs showed that canonical pathways related to immune response were activated in CR diet cohorts irrespective of the age of the host. Since the influence of gut microbiota on tumor immune microenvironment is well known [54], we speculate that CR diet related changes in gut microbiome may regulate the immunotherapy response, which needs to be further investigated.
Traditionally, mitochondrial DNA has been believed to encode 13 protein-coding genes. However, recent studies have provided a series of evidence supporting the existence of novel small-sized proteins known as mitochondrial microproteins encoded in the mitochondrial smORFs [18,25,26]. These microproteins have various biological functions such as extending lifespan [23,24] and preventing neurodegenerative diseases [17,[19], [20], [21], [22]], metabolic and skeletal muscle disorders [16,[55], [56], [57], [58]]. Expectedly, our study found multiple mitochondrial smORFs to be differentially regulated by age and HF diet in tumors. However, since these smORFs are putative and not yet annotated, future studies that examine their detailed biological functions, specifically in the intestines, of those smORFs are necessary prior to interpreting our findings. Both diet and the age of the mouse uniquely modulate the mitochondrial transcriptome, independent from the effects of these same factors on the nuclear transcriptome of the tumor. These observations point to host factors that influence tumor progression through modulation of both nuclear and mitochondrial small and large proteins, which may prove to have specific roles in tumor biology.
Our group previously showed the clinical relevance of MOTSc as a biomarker, with higher MOTS-c levels associated with lower prostate cancer risk in European American men but not in African American men [59]. Mitokine MOTS-c has been known to be a systemic modulator of diet, aging, and obesity related metabolic function, acting as an exercise mimetic which promotes oxidation of glucose and fat, and decreases inflammation [60]. MOTSc holds a therapeutic potential in many diseases, and its antitumor effect on ovarian cancer has been reported [61,62] but its role in CRC has not been explored. The levels of MOTSc naturally decrease with age. Humanin (HN) is a mitochondrial polypeptide known to have anti-apoptotic and anti-oxidative properties [63]. Our results identified that MOTSc and HN significantly decreased in HF diet in young cohort, suggesting accelerated aging in this group, and potentially pointing towards the role of MDPs as potential therapeutic targets in EO-CRC. Humanin and MOTSc’s role in cancer is still in a nascent stage of investigation, especially CRC, and our group previously showed that both have anti-tumor roles [59,64]. To the best of our knowledge, this is the first study showing tripartite relationship between these mitokines and CRC and the effect of other factors such as chronological age of the host and diet.
The gut microbiota is an essential player during CRC tumorigenesis, as it is known to modulate the anti-cancer immune response and the tumor microenvironment. Its dysbiosis has been associated with CRC progression, making the gut microbiome a very attractive modality for improving CRC treatment outcomes [65,66]. Diet and chronological age are factors that influence an individual’s microbiome. However, how the modulation of gut microbiome by diet and age impacts the CRC progression remains elusive. Our results indicate that the microbiomes of the old individuals were significantly different and had higher alpha-diversity compared to our young group in the CRC mice model. Within the two groups, there were statistically significant compositional differences seen through the clustering of groups. Overall, in the young mice, we saw an enrichment of bacterial species that could potentially inhibit CRC tumor growth. The most elevated populations were B. thetaiotaomicron, a gram-negative bacterium reported to restrict CRC-associated liver metastasis [67], and Akkermansia muciniphila, which is a promising probiotic agent that could prevent and/or reduce CRC development [68]. A. muciniphila was also the most differentially abundant species between NR and CR-fed mice, being enriched in the NR-fed mice. Another interesting observation was that Lachnospiraceae bacterial species were more abundant in the CR-fed mice, except for Lachnospiraceae bacterium 28-4 which was found to be more enriched in the HF-fed mice. Functionally, the old group had more enrichment of the biosynthesis pathways such as the L-arginine, L-citrulline, L-methionine, putrescine, which correlates with the higher microbial diversity observed in this group in our study. Increased arginine uptake and hyperactive arginine metabolism is a hallmark of CRC [69], and thus understanding argininie metabolism pathway and its potential to be a diagnostic marker or therapeutic intervention needs to be explored further. To the best of our knowledge, our age stratified study is the first to show that arginine biosynthesis pathway is associated with gut microbiome [70]. Also, our analysis for the first time showed the connection between CMP-legionaminate biosynthesis pathway and CRC in young mice, suggesting further investigation of its role in EO-CRC [70]. Diet stratification showed enrichment of menaquinol (vitamin K2) biosynthesis pathway in HF cohort. Studies have shown that this pathway has increased activity in advance CRC and known to have a protecting role for gut microbial flora after chemotherapy [71]. We speculate that the increased activity of menaquinol pathway in high fat diet may either due to microbial dysbiosis or protective role against high fat consumption, which needs to be further investigated.
In a larger cohort including 1284 mCRC patients enrolled in the CALGB/SWOG 80405 trial, we have shown previously that plant-based diets are associated with better survival among mCRC patients [72]. In the same cohort, including 1149 mCRC patients, we also showed that replacement of animal fat with vegetable fats in the diet could lead to longer progression free survival in patients with mCRC [73]. This current study is preclinical proof of concept and along with other ongoing inter-related studies in patient cohorts by our group as mentioned above is a path forward towards real-world application. However, this study had several limitations and warrants for further investigation to address these gaps. Firstly, considering sex as a major variable factor, we used male hosts for our current study and planned to expand same study design in female hosts as a next step. Since MC38 cell line shows recipient sex-based differences (it grows faster in male host than females), to avoid experimental biases and considering sex as a biological variable, only sex matched results will be compared (male host results can’t be compared to female host). Secondly, mitochondrial small open reading frames (smORFs) that code MDPs in mice have not yet been annotated limiting their functional correlations. We also hope that future studies will provide insights into those smORFs and help in interpreting our findings more clearly. Thirdly, subcutaneous models don’t give complete picture of the tumor microenvironment but based on current objective of this pilot study without any treatments involved, subcutaneous models provided us with the overall picture of tumor intrinsic and TME gene expression changes. For comprehensive TME role, treatment efficacy studies and to complement our subcutaneous model findings, we will be using orthotopic models for future studies. Lastly, though in the current study we used MSI-H MC38-C57BL/6J as a model system, we plan to investigate effects of diet-age and microbiome in the MSS molecular subtype of CRC. Also, the sample size and period of treatment were substantially reduced because of limited survival of the old mice due to age-related complications, and further studies are warranted with larger sample sizes.

Conclusion
Based on transcriptomic profiles and comprehensive cytokine analysis, our data shows that age-diet interactions significantly influence the host TME leading to differential gene expression in the tumor. Metagenomic analysis revealed the mediating role of the gut microbiome in age- and diet-induced CRC progression. Collectively, our findings will advance current understanding of the mechanisms by which aging, diet and gut microbiota impact CRC onset, progression, metastasis, outcome and therapeutic response but warrants further investigation.

Ethics approval and consent to participate
Animal handling and procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Southern California (USC).

Consent for publication
Not Applicable.

Grant Support
Research reported in this publication was partly supported by the National Cancer Institute of the National Institutes of Health under Award Numbers P30CA014089 [to H-JL], the 10.13039/100019757Gloria Borges WunderGlo Foundation, 10.13039/100006508Dhont Family Foundation, Victoria and Philip Wilson Research Fund, San Pedro Peninsula Cancer Guild, Ming Hsieh Research award [to H-JL], Daniel Butler Research Fund and Julia Samwer Research Fund, R01AG069698, P30CA014089-45S3, HF-AGE-23-1273964-51 (to P.C.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Disclosures
H-JL reports receiving honoraria from consultant/advisory board membership for Merck Serono, Bayer, 3T bioscience, Fulgent, Oncocyte, Adagene and Genentech. All other authors declare no conflict of interests. Preliminary results of our study have been presented, in part, at the ASCO GI 2025, San Francisco, January 23-25, 2025 (received the Conquer Cancer Merit Award) and AACR Annual Meeting, Orlando, April 14-19, 2023 (in-person).

Data transparency statement
No accession codes, unique identifiers, or web links for publicly available datasets were used in this study. Data has been included in the main manuscript and supplementary material. No clinical datasets or third-party data was used.

CRediT authorship contribution statement
Shivani Soni: Conceptualization, Formal analysis, Investigation, Validation, Visualization, Writing – original draft. Pooja Mittal: Data curation, Formal analysis, Investigation, Validation, Visualization, Writing – original draft. Jae Ho Lo: Formal analysis, Investigation, Visualization, Writing – review & editing. Yan Yang: Formal analysis, Investigation, Methodology, Visualization, Writing – review & editing. Goar Smbatyan: Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – review & editing. Keehoon Lee: Formal analysis, Investigation, Methodology, Visualization, Writing – review & editing. Junxiang Wan: Formal analysis, Methodology, Visualization, Writing – review & editing. Hiroshi Kumagai: Data curation, Formal analysis, Investigation, Visualization, Writing – review & editing. Kelvin Yen: Formal analysis, Investigation, Visualization, Writing – review & editing. Hemal H. Mehta: Formal analysis, Investigation, Visualization, Writing – review & editing. Brendan Miller: Formal analysis, Software, Writing – review & editing. Lesly Torres-Gonzalez: Data curation, Investigation, Visualization, Writing – review & editing. Francesca Battaglin: Investigation, Visualization, Writing – review & editing. Unnati Hemant Shah: Investigation, Visualization, Writing – review & editing. Michela Bartolini: Investigation, Visualization, Writing – review & editing. Wu Zhang: Investigation, Visualization, Writing – review & editing. David W. Craig: Investigation, Supervision, Writing – review & editing. Josh Millstein: Data curation, Formal analysis, Investigation, Methodology, Software, Writing – review & editing. Pinchas Cohen: Conceptualization, Formal analysis, Funding acquisition, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing. Heinz-Josef Lenz: Conceptualization, Formal analysis, Funding acquisition, Project administration, Resources, Software, Supervision, Writing – original draft, Writing – review & editing.

Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: H-J.L. reports receiving honoraria from consultant/advisory board membership from Bayer, Genentech, Roche, Merck, Merck KG, Oncocyte, Fulgent, G1 Therapeutics, 3T Biosciences, Jazz Therapeutics, Protagonist.