Light-dark shift promotes colon carcinogenesis through accelerated colon aging

Summary

Colorectal cancer (CRC) is the third most common cancer worldwide, with rising prevalence among younger adults. Several lifestyle factors, particularly disruptions in circadian rhythms by light-dark (LD) shifts, are known to increase CRC risk. Epidemiological studies previously showed LD-shifts are associated with increased risk of CRC. To explore the mechanisms and interactions between LD-shift and intestinal aging, we investigated how the combination of LD-shifts and aging impacts colon carcinogenesis development. Our data showed that LD-shifts and aging increased colon tumorigenesis. Notably, LD-shift accelerated intestinal aging by altering aging-related pathways, such as intestinal barrier damage, accompanied by dysbiotic changes in the intestinal microbiota that negatively impacts barrier stability. The increased carcinogenesis and intestinal aging were preceded by enrichment in host-microbiome features that are strongly regulated by the circadian clock. Overall, our results suggest that LD-shifts, increasingly prevalent among young adults, contribute to both intestinal aging and the development of colon carcinogenesis.

Graphical abstract

Highlights

• •Light-dark (LD) shift impacts intestinal microbiota and barrier pathways
• •LD shift accelerates intestinal aging and colon carcinogenesis
• •Host-microbiome aging features are strongly regulated by the circadian clock

Cell biology; Cancer

Introduction

According to World Health Organization 2023 data, colorectal cancer (CRC) is the third leading cause of cancer mortality worldwide and is one of the most preventable cancers. Typically, CRC transforms from premalignant polyps to dysplasia that grow on the mucosal surface of the colorectum. The risk of CRC increases with age, with most cases (>90%) occurring in individuals older than 50 years.¹ Although the introduction of age-tailored CRC screening programs, such as stool tests and colonoscopies, for middle-aged and older adults has reduced CRC incidence and mortality in the general population, new cases of CRC in younger adults have been increasing at an alarming rate.^2,3 Understanding the factors and mechanisms that contribute to the increasing risk of this age-associated disease among younger individuals is key to developing effective interventions against the predicted burden of early onset CRC.

Several lifestyle-associated factors such as diet and alcohol consumption have been shown to increase the risk of the disease.⁴ Among the lifestyle factors that may impact the risk of CRC, disruption of circadian rhythms, which govern a wide spectrum of host physiology, has become increasingly common in modern society. Circadian rhythm is an approximately 24-h sleep/wake cycle regulated by light/dark cues in the central clock of the brain. Shifts in the light/dark cycle can increase the risk of CRC based on existing epidemiological data reviewed and endorsed by the International Agency for Research on Cancer (IARC).⁵ Data from our laboratory and others suggest that light/dark shifting (LD-shift) could have deleterious effects on intestinal homeostasis and gut integrity.^{6,7,8,9,10,11} The mechanisms by which the LD shift may promote CRC are not yet fully understood.

Here, we studied the effects of LD-shift and the underlying mechanisms of colon tumorigenesis in a colon-specific mouse model of polyposis across three age groups (early, middle, and late). We observed that an LD shift could aggravate the impact of age by promoting aging-related pathways in the colon and inducing intestinal barrier dysfunction. These changes were accompanied by changes in the intestinal microbiota, characterized by lower relative abundances of putative beneficial bacterial genera involved in barrier stability, paired with increased relative abundances in putative pro-inflammatory bacterial genera. To investigate the host-microbiome interaction in response to LD shift and aging on the risk of CRC, we compared the mucosal and luminal microbiota profiles using high-throughput sequencing of 16S ribosomal RNA (rRNA) gene amplicons and colonic transcriptome (RNA-seq) across time, with or without LD shift. By integrating these data, we identified unique host-microbe interaction networks involved in intestinal aging and colon carcinogenesis by LD shift. We verified that these networks are related to human CRC, respond to microbial interventions that ameliorate polyposis, and are associated with the circadian clock. The role of the molecular clock was further established by investigating clock-mutant animals and generating mice with clock-altered polyposis. These data suggest that an LD shift, an increasingly frequent habit among younger adults, promotes intestinal aging and colon carcinogenesis.

Results

Light-dark shift accelerates colon tumorigenesis by aging

To address the effect of a shift in the light/dark cycle (LD-shift) on colon carcinogenesis, we used our previously reported TS4Cre × APC^Δ468 (TS4/APC) mice as a preferred model of colon tumorigenesis,^12,13 rather than classical adenomatous polyposis coli (APC) mice, where tumors predominantly develop in the small bowel. These TS4/APC mice showed clear age-dependent colon tumorigenesis up to 28-week, when the tumor burden peaked, with increased total number of polyps (Figure 1A). These mice typically succumb to cancer between 28 and 32 weeks of age. To monitor changes in the colon over time, even before the tumor burden peaks, we chose three age groups within this survival window. Consequently, we designate the 12-, 18-, and 28-week-old mice in our study as the ‘early’, ‘middle’, and ‘late’ age groups, respectively.

Figure 1.

Shift interacts with age to promote polyposis in TS4Cre × APC^Δ468 mice

(A) The total number of polyps significantly increased after 18- and 28-week of shift and age.

(B) Large polyp burden significantly increased after 18- and 28-week of shift and age.

(C) The total number of advanced lesions significantly increased after 18- and 28-week of shift and age. Data are represented as mean ± SD. Significant difference: ∗p < 0.05.

(D) TS4Cre × APC^Δ468 mice 18-week of shift and age exhibited histological features consistent with high-grade dysplasia: (i) Low-Grade Dysplasia: The white arrow points to nuclear changes in a crypt, which affects only a portion of the mucosal thickness. This is distinct from the adjacent normal colon mucosa (black arrowhead). Scale bar: 500 μm. (ii) High-Grade Dysplasia/Carcinoma in situ: The lesion involves the full thickness of the mucosa but has not invaded the submucosa, as evidenced by the intact basement membrane (dashed black line). The white arrow highlights the gland extending to the bottom of the mucosa. Scale bar: 500 μm. (iii) The top panel shows a magnified view of the area in (ii), showing a part of the polyp surface that is eroded consistent with surface erosion (black arrow). Scale bar: 100 μm. (iv) Invasive Colon Cancer: The white arrow in this panel indicates a gland infiltrating the submucosa, a hallmark of invasive cancer. Scale bar: 500 μm. (v) Features of Invasive Carcinoma: A magnified view of (iv) demonstrates back-to-back and cribriforming glands. Scale bar: 150 μm.

(E) The number of differentially expressed genes in each pairwise comparison, including LD aging and LD shifting, for mice at 12- and 18-week of age. NS = Non-shifted, S = Shifted, wk = weeks.

Interestingly, the LD shift accelerated tumorigenesis with age. TS4/APC mice subjected to LD-shift by 18 weeks of age showed a comparable number of polyps to non-shifted animals at 28 weeks old, while TS4/APC mice subjected to LD shift experienced the highest tumor polyp burden by 28 weeks of age (35% more compared to 28 weeks non-shifted, Figure 1B). As the mice aged, those in the LD-shift group (at 18 and 28 weeks) not only had a higher tumor polyp burden compared to the non-shifted group but also exhibited a greater number of advanced lesions, as indicated by larger tumor sizes (Figure 1C). By 18 weeks of age, the LD-shift group exhibited histological features consistent with high-grade dysplasia (Figure 1D). These results indicate that the LD shift could mimic the effect of age in CRC-susceptible mice by exacerbating colon tumorigenesis by age.

Colonic mucosa gene expression re-organizes aging associated pathways upon LD-shift

To understand how colon tissue changes due to an LD shift as a function of age, we profiled the transcriptome-wide gene expression (RNA-Seq depth: >40 million reads) of the non-tumoral mucosa of the TS4/APC mouse intestine over time. High burden of polyps in the late group (28-week-old mice) makes it challenging to assess the expression profile of the non-tumoral mucosa. We focused on the early (12 weeks) and middle (18 weeks) age groups, before the development of such a high tumor load. This strategy enabled us to examine age-related changes in colon tissue expression patterns, as illustrated in Figure 1E. To examine differential gene expression (DEGs) within each condition, we used the negative binomial distribution test implemented in DESeq2.¹⁴ LD shift combined with age (18-week shifted mice) resulted in the highest number of changes in colonic gene expression among the four mice groups (1,074 DEGs) (Figure 1E).

Comparison of 12-week non-shifted and 18-week shifted mice identified 1,074 DEGs (p ≤ 0.05). Of these, 701 genes were downregulated and 373 were upregulated in 18-week shifted mice (p ≤ 0.05; fold-change ≥1.5) (Figure 2A). A full list of DEGs is provided (Data S1).

Figure 2.

Differential gene expression signatures for TS4Cre × APC^Δ468 mice LD aging and shifting in the colonic mucosa

(A) Volcano plot showing differential gene expression profiles with significantly (p < 0.05; fold change ≥1.5) upregulated (n = 373 genes) and downregulated (n = 701) genes in 18-week shifted mice when compared with 12-week non-shifted mice.

(B) Enriched molecular pathways in 18-week shifted and 12-week non-shifted mice.

(C) Volcano plot showing differential gene expression profiles with significantly (p < 0.05; fold change ≥1.5) upregulated (n = 464 genes) and downregulated (n = 427) genes in 12-week shifted mice when compared with 12-week non-shifted mice.

(D) Enriched molecular pathways in 12-week shifted and 12-week non-shifted mice.

(E) Volcano plot showing differential gene expression profiles with significantly (p < 0.05; fold change ≥1.25) upregulated (n = 114 genes) and downregulated (n = 217) genes in 18-week shifted mice when compared with 18-week non-shifted mice.

(F) Enriched molecular pathways in 18-week shifted and 18-week non-shifted mice.

(G) Intersection of colorectal cancer (CRC) dysregulated genes (n = 4,566) with differentially regulated genes in shift-only (n = 111), age-only (n = 91), and combined shift-age conditions.

To capture gene networks that were functionally altered in response to age shift, we next assessed the enrichment of DEGs at the pathway level (p ≤ 0.05) (Figure 2B). The primary canonical pathways that were downregulated after an 18-week shift were associated with extracellular processes, including the maintenance of the extracellular matrix (ECM) and cell-cell communication (e.g., regulation of the actin cytoskeleton, ECM-receptor interaction, and focal adhesion). These downregulated DEGs related to extracellular processes are essential for maintaining the functional integrity of the intestinal barrier, which is known to be impacted by aging.¹⁵ Among the upregulated pathways, DNA damage-related processes are affected by aging and contribute to intestinal barrier damage and carcinogenesis.¹⁶ As an independent validation of DNA damage upon LD shift, we quantified the protein levels of H2AX, an essential DNA damage marker, in both shifted and non-shifted conditions at an early polyposis age (i.e., at 12 weeks before polyposis is accelerated by shifting). These data showed increased (p = 0.0005) levels of H2AX in the shifted colon compared to those in the non-shifted colon (Figure S1).

A similar pattern of pathway enrichment was observed when 12-week (downregulated: 427 genes, upregulated: 464 genes) (Figures 2C and 2D) or 18-week shifted (downregulated: 217 genes, upregulated: 114 genes) (Figures 2E and 2F), and non-shifted conditions were compared (effect of shift only). Pathway enrichment in middle and early age revealed a preponderance of downregulated extracellular processes (downregulated: 119 genes, upregulated: 153 genes; (Figures S2A–S2B), similar to changes induced by the LD shift in middle and early age (downregulated: 113 genes, upregulated 81 genes) (Figure S2C). These data suggest that intestinal changes resulting from an LD shift mimic the molecular changes that occur as animals age. Consistent with the age-mimicking effect of LD shift in the colon, the 12-week shifted, and 18-week non-shifted mice were the most similar groups in the colon gene expression patterns, with only three pathways differentially enriched between them (Figure S2D). Among these, there was enrichment for circadian rhythm, consistent with the expected early effect of LD shift on the circadian clock. Comparable enrichment patterns upon LD shift (at an early age) and with age (by middle age) suggested the occurrence of similar changes at the transcriptome level.

To examine the relevance of shift-age DEGs in our TS4/APC mouse model to human CRC, we turned our attention to The Cancer Genomic Atlas (TCGA) human data of CRC cases. Briefly, we obtained control-matched samples (n = 89) from TCGA and extracted the most significantly dysregulated CRC genes (p < 0.05, n = 4566 genes). We then compared the DEGs obtained from the age-shift (n = 1073 genes), age-only (n = 774 genes), and shift-only (n = 1198 genes) conditions with the CRC dysregulated genes extracted from TCGA. The shift-age condition showed the largest overlap with the dysregulated CRC genes (n = 178 genes), followed by the shift-only condition (n = 111 genes), indicating a more pronounced change in gene expression in the CRC profile upon the shift-age combination (Figure 2G). Taken together, these observations suggest that age and shift interact to reorganize colonic gene expression to enhance the risk of colon cancer.

Markers of intestinal aging and barrier integrity are impacted by the LD-shift

Considering the observed impact of the shift on aging-related pathways in our model, we intersected the DEGs (p < 0.05; 12-week non-shifted vs. 18-week shifted) with gene lists representing pathways associated with cellular hallmarks of aging. To identify these aging hallmarks, we compiled gene lists from eight well-documented aging-related processes focused on the colon: senescence (503 genes,¹⁷), inflammation (147 genes,¹⁸), autophagy (162 genes,¹⁹), oxidative stress (685 genes,²⁰), intestinal barrier (182 genes,²¹), and DNA damage response (47 genes,²²). By quantifying the maximal clique centrality (MCC,²³) of the genes belonging to these pathways, we identified key hubs for each process. Intersection of the hubs with DEGs revealed modulation of several essential genes involved in senescence (24 genes, p = 0.021), inflammation (15 genes, p = 0.012), autophagy (18 genes, p = 0.009), and the intestinal barrier (21 genes, p = 0.007). Among these were notable increases in the expression of senescence-associated genes such as UBE2G (Ubiquitin-Conjugating Enzyme E2 G1²⁴) and CENP-A (Centromere Protein A²⁵), inflammation-related genes like IRF8 (Interferon Regulatory Factor 8²⁶) and FUT2, as well as autophagy-related genes including LAMTOR2 and CHMP2B. Importantly, in accordance with the strong effect of shift age on barrier related processes (Figure 2), several genes involved in the integrity and function of the intestinal barrier were significantly downregulated in 18-week shifted mice (Figures 3A–3D). For instance, we identified MUC13 as one of the most downregulated genes in middle-aged mice (log₂FC = 1.3; p < 0.0009). MUC13 encodes mucin, which is expressed on the apical surface of the epithelium and contributes to the intestinal barrier.²⁷ Similarly, another significantly downregulated gene by age shift was CLDN4 (log2FC = 1.70; p < 0.05), a crucial component of the tight junction that interacts with cadherins to maintain intestinal barrier integrity.²⁸ Other barrier-associated genes that were affected by shift age were CTNND1, CLDN7 (both coding for tight junction proteins), CDH1 (epithelial cadherin or E-cadherin), KLF4 (maintenance of epithelial homeostasis), and TJP1 (ZO-1, Tight junction proteins-1).

Figure 3.

Differential gene expression-associated pathways for TS4Cre × APC^Δ468 mice LD 18-week shift and 12-week non-shift colonic mucosa

(A–D) Differentially expressed genes in selected aging-associated pathways: (A) senescence, (B) inflammation, (C) autophagy, and (D) intestinal barrier. (A-C) Orange: Genes upregulated in 18-week shifted conditions compared with 12-week non-shifted conditions; (D) Green: Genes downregulated in 18-week shifted conditions compared with 12-week non-shifted conditions. Significantly dysregulated genes were defined as p < 0.05, fold change ≥1.25.

(E) Average E-cadherin intensity in LD 12-week non-shifted and 18-week shifted mice colon. Scale bar: 100 μm.

(F) Average Goblet cell quantification in the LD 12-week non-shifted and 18-week shifted mice colon. Scale bar: 100 μm. Data are represented as mean ± SD. ∗p < 0.05; ∗∗∗∗p < 0.0001.

Considering the role of disrupted barriers in the early stages of colon carcinogenesis^29,30 and to validate our gene expression results, we investigated the tissue status of several barrier markers at the protein level in response to the LD shift in the early age groups. E-cadherin, a marker of adherens junctions important in colonocyte integrity, decreased (p = 0.027) in response to the LD shift (Figure 3E). This finding was consistent with the downregulation of CDH1 expression. In addition to colonocytes, mucin-secreting goblet cells that are differentiated from stem cells also play a key role in maintaining the epithelial lining of the colon. Given the downregulation of goblet cell differentiation-associated genes involved in mucus production in our RNA-seq (MUC2, MUC13, and AGR2), we quantified the abundance of goblet cells that reside in the colon surface epithelium. Shifted mice had a significantly reduced (p < 0.0001) number of goblet cells (mean = 0.48 ± s.d. = 0.06) compared to non-shifted group (mean = 0.74 ± s.d. = 0.05) (Figure 3F).

In summary, we observed that LD shift can affect the cellular hallmarks of aging in the colon, directly relevant to colon carcinogenesis, including those involved in maintaining intestinal barrier function.

Accelerated polyposis precedes intestinal dysbiosis

The intestinal barrier restricts the passage of luminal toxic compounds such as bacterial byproducts. If these compounds pass through the barrier due to disruption, they may promote tumor growth.³¹ “Beneficial” microbiota on the other hand, may modulate colon neoplastic transformation, in part via contribution to the barrier function.³⁰ The intestinal microbiota is affected by disruption of circadian rhythm and aging.^32,33

We first analyzed the colonic content of the TS4/APC LD mice groups to study the microbiota that were in direct contact with the intestinal mucosa at the time of sacrifice. Beta-diversity analyses revealed that the majority of LD mice groups had no significant differences in microbial community structure with respect to Bray-Curtis dissimilarity in the colonic content, except for the shifted mice between ages 12 to 18 weeks (R = 0.270, p = 0.024) and 12 to 28 weeks (R = 0.227, p = 0.008) (Figures S3A–S3C). Differential abundance analysis indicated that the relative abundance of the beneficial genera Bifidobacterium and Lactobacillus decreased (p < 0.01) over time in 18- and 28-week shift mice compared to 12-week shift mice (Figures S4A and S4B). Alternatively, the relative abundances of putative proinflammatory genera, like Mucispirillum, Rikenella, Helicobacteraceae Unclassified, and Bilophila increased (p < 0.05) over time in 18- and 28-week shift mice compared to 12-week shift mice (Figures S4C–S4F).

To gain insight into the earlier effects of shift and age before high tumor burden occurs, we compared the colonic content microbial profiles of the shifted middle (18-week shifted) vs. non-shifted early (12-week non-shifted) age group mice. While we observed no significant overall difference in the microbial community structures between these two groups (R = 0.152, p = 0.11), the alpha-diversity indices of observed richness showed a significant decrease in 18-week shifted mice (p = 0.032) in the number of different genera represented in the colonic content microbial community. The Shannon, Simpson, and evenness indices did not reach statistical significance (p > 0.05). To identify important genera altering microbial richness, we performed a random forest analysis between shifted 18-week and non-shifted 12-week mice (Figure 4A). Shift by age resulted in higher trends of relative abundances of several putative pathobiont-producing genera, previously linked to CRC, including Bilophila, Mucispirillium, Odoribacter, and Helicobacter.³⁴ Additionally, the age shift revealed a decrease in the relative abundance of several beneficial short-chain fatty acid (SCFA)-producing genera that contribute to intestinal barrier stability, like Bifidobacterium, [Ruminococcus], and Dorea³⁵ (Figures S5A–S5H).

Figure 4.

Microbial dysbiosis with shift and age in TS4Cre × APC^Δ468 mice

(A) Top colonic content microbial genera defining the 12-week non-shift and 18-week shift conditions based on the random-forest algorithm.

(B) Correlation between differentially abundant microbial genera under the non-shift and shift conditions. Blue: negative correlation; red: positive correlation.

(C) Correlation of top differentially abundant genera with top differentially expressed genes when 18-week shifted and 12-week non-shifted conditions were compared. Blue: negative correlation; Red: positive correlation.

(D) Interaction network of the top bacterial genera and gene correlations. Colors represent differential expression of genes and differentially abundant genera. Green: down upon shift; red: up upon shift.

(E) Pairwise correlation of selected bacterial gene expression profiles depicting a significant (R² = 0.79, p = 0.039) positive correlation between Smad4 and genus Bifidobacterium, a significant (R² = −0.68, p = 0.042) negative correlation between Fgfr2 and the genus Ruminococcus.

(F) Intersection of colorectal cancer (CRC) dysregulated genes that interact with bacteria (n = 77) in our data and colorectal cancer (CRC) dysregulated genes that interact with bacteria (n = 48) from TCGA.

(G) Gene-microbe correlations of the selected host genes from TCGA samples.

To determine whether the microbial alterations preceded or followed colon tumorigenesis and to eliminate any potential heterogeneity among different TS4/APC LD mice groups, we also sampled feces of non-shifted and shifted mice as they aged at four time points: early time points (6- and 12-week) and later phases (18 and 28 weeks). We found a significant effect of age and shift versus non-shifted mice on fecal microbial community structures at 18 weeks of age (R = 0.272, p = 0.025), prior to colon tumorigenesis at 28-week of age (Figures S6A–S6B). Similar to colonic content, differential abundance analyses indicated no significant alterations in fecal microbial composition for the majority of highly abundant (>1%) genera among the different mouse groups (Figure S6C). To account for potential genera alternations of less abundant microbial taxa, we conducted random forest analysis of the fecal samples over time, which confirmed 80% of the previously mentioned age-shift-related changes in the colonic microbial profile (data not shown). Briefly, the relative abundances of genera Bifidobacterium, a known SCFA-producing bacterium, equol-producing Adlercreutzia, and hormone-producing Coriobacteriaceae Unclassified all significantly decreased with age upon LD shift (Figures S7A–S7C). Similarly, the relative abundances of six putative proinflammatory and opportunistic pathogenic genera significantly increased in shifted mice over time: Rikenellaceae Unclassified, Bacteroidales Unclassified, Odoribacter, Clostridiaceae Unclassified, and Desulfovibrionaceae Unclassified (Figures S7D–S7H). Overall, these findings suggest that changes in intestinal microbial composition occur both upon shifting and over time, similar to the pattern observed at the level of colon gene expression.

Given the disrupted barrier markers in the colonic tissue and lower relative abundances of beneficial bacteria in both colonic content and feces, we then measured the levels of targeted SCFA metabolites over time in the mouse fecal samples. We found no effect of the shift on the total or individual levels of the measured SCFA metabolites over time (p > 0.05) (Data S2). These results imply a more complex microbial interaction underlying shift-induced colon tumorigenesis. Given the close interaction of microbes in the intestine and the need to identify microbial populations that change together in response to shifts and aging, we used a systems approach to identify the most significant interactions among the altered microbes in our model (Figure 4B). Eight genera (Bilophila, Odoribacter, Bifidobacterium, Coprococcus, Ruminococcus, Bacteroides, Mucispirillum, and Helicobacter) were found to depict strongly correlated with their abundance values. Notably, we found a strong negative correlation (R² ≤ −0.50) between the barrier-related beneficial genera Bifidobacterium and Ruminococcus and the pathogenic genera Odoribacter and Helicobacter. On the other hand, the beneficial genera Ruminococcus and Bifidobacterium (R² = 0.64) and the pathogenic genera Odoribacter and Helicobacter (R² = 0.61) showed strong positive correlations. This suggests that microbes, rather than in isolation, and potentially via interacting pathways, modulate the impact of the shift on the intestine over time. Next, we hypothesized that microbe-gene networks drive the shift-induced aging phenotype in the colon.

Shift related “dysbiosis” coincides with “aging” phenotype in the colon

To assess the relationship between host genes and microbes responding to LD shift in the colon, we examined the interrelation of 77 differentially expressed hub genes from aging-related pathways and eight microbial genera that were found to be significantly altered upon shift. We visualized the link between taxon abundance and host gene expression (Figure 4C). All microbial genera showed strong correlations (R ≥ 0.5) with host genes, while several bacteria were found to correlate with multiple genes (e.g., Bilophila, Bifidobacterium, Ruminococcus, and Mucispirillum). By shortlisting the top correlated host-microbe (Spearman correlation between -R: −0.5 to 0.5) and microbe-microbe interactions, we created a sub-network representative of the shift-related dysbiosis pattern. The sub-network is shown (Figure 4D), where edges represent correlations (red: positive; blue: negative) and node colors indicate the differential expression of the gene or microbe (red: upregulated; green: downregulated).

The sub-network depicts correlated abundance changes of microbes or genes as a function of their modulation by LD-shift. For example, decreased abundance of Bifidobacterium upon shift correlated with downregulation of SMAD4, an essential CRC-associated gene. The abundance of the genus Bifidobacterium and expression of SMAD4 was highly correlated (R² = 0.79, p = 0.039) (Figure 4E), suggesting the existence of a co-regulatory mechanism. A decreased abundance of Bifidobacterium has been repeatedly reported in human CRC,³⁶ while its anticancer activity has been demonstrated in experimental models of CRC.³⁷ Decreased SMAD4 expression has been reported in approximately 20% of patients with CRC. SMAD4 is an intracellular signaling component of the transforming growth factor β (TGF-β) pathway, which is involved in transmitting signals from the cell surface to the nucleus.³⁸ While the direct link between Bifidobacterium and SMAD4 has not been studied, the TGF-β/SMAD4 pathway is heavily involved in the cross-talk between the intestinal microbiota and components of the intestinal barrier (reviewed in³⁹). Another host-microbe interaction of interest in our study was the negative correlation between genus Ruminococcus abundance and FGFR2 expression (R² = −0.68, p = 0.042) (Figure 4E). Both the decrease in relative abundance of the genus Ruminococcus and upregulation expression of FGFR2 were found to be related to CRC risk and progression in prior studies.⁴⁰

To evaluate the relevance of the observed microbe-gene networks found in our TS4/APC mouse model, we focused on microbe-gene correlations in human CRC. To this end, we extracted human CRC tumor data including microbiome, from TCMA, which is a collection of curated and decontaminated microbial compositions of gastrointestinal cancers, including colorectal tissues.⁴¹ In addition, we obtained the corresponding transcriptome data from TCGA database.⁴² To study microbe-gene interactions in human CRC, we correlated bacterial abundances greater than 0.001% in all TCMA samples with the expression levels of our list of aging hallmark genes (n = 76) (Table S1) from each TCGA sample (n = 112 CRC samples). All eight bacterial genera found in our mouse data were also present in these human CRC samples and were strongly linked to many of the same genes (n = 48/76 genes, 62.3%) identified in our analysis (Figures 4F and 4G). Strong correlations were observed between Bifidobacterium-SMAD4 (R² = 0.63) and Ruminococcus-FGFR2 (R² = 0.58), further substantiating our observations in our animal model.

Microbial intervention corrects dysregulated microbe-gene networks of polyposis mice

To investigate the causality between the microbiome and host response in the intestine, we investigated whether restructuring the microbial populations could rescue the alterations in the host-microbe profiles observed above. To this end, we used a fiber prebiotic that was previously shown to promote beneficial gut bacteria and ameliorate tumorigenesis in TS4Cre×cAPC^l°^x468 mice.⁴³ After feeding TS4Cre×cAPC^l°^x468 mice a prebiotic fiber versus a standard rodent chow (control) diet for 12-week as previously described,⁴³ we performed 16S rRNA sequencing of fecal microbial profiles and colonic mucosa RNA-Seq gene expression. Analysis of these microbes and genes between the two conditions revealed 101 bacterial genera (25 enriched, 76 depleted) and 1,701 genes (418 upregulated and 1,283 downregulated) differentially regulated under prebiotic conditions (Data S3).

Among these data, seven out of eight (87%) bacteria and 43/56 (77%) of the genes were found to be altered upon age shift (Figure 5A). Furthermore, differentially abundant bacteria in the prebiotic condition showed an inverse relationship with their fold changes in response to shift age, suggesting that microbial intervention partially reversed microbial dysbiosis associated with shift age (R² = −0.73, p < 0.001) (Figure 5B). In addition, we observed a similar correlation pattern between Fgfr2-Ruminococcus (R² = −0.34, p < 0.05) and Smad4-Bifidobacterium (R² = 0.56, p < 0.01), as observed in the shifted-age condition (Figure 5C). This further strengthens our observation of strong gene-microbe co-expression profiles. In addition, the expression pattern of host genes was reversed when the effects of shift age and prebiotics were compared (Figure 5D). This suggested that most of the colon gene dysregulation patterns linked to dysbiosis from the effect of shift-age responded favorably to microbial restructuring upon prebiotic fiber treatment (Figure 5D). Overall, these data support the causal role of microbiota in the reorganization of intestinal gene expression.

Figure 5.

Prebiotics rescue the effect of shift age in mice

(A) Top: Intersection of shift-mediated bacteria (n = 8) with prebiotic-mediated bacteria (n = 7). Bottom: Intersection of shift-mediated genes (n = 56) with prebiotic-mediated genes (n = 43).

(B) Scatterplot of the differentially abundant bacterial genera changing in 18-week shifted mice compared with 12-week non-shifted mice, plus bacterial alterations in prebiotic mice compared to healthy littermates.

(C) Pairwise correlation of selected bacterial gene expression profiles depicting a positive correlation between Smad4 and genus Bifidobacterium and a negative correlation between Fgfr2 and genus Ruminococcus.

(D) Differential expression levels of top genes changing in 18-week shifted mice compared with 12-week non-shifted mice and prebiotic mice compared with healthy littermates. Blue: downregulated genes; Yellow: no change in gene expression; Red: upregulated genes. Pairwise correlation between top DEGs (n = 55) and differentially enriched bacteria (n = 8). Red: positive correlation; blue: negative correlation.

Hallmarks of intestinal aging are coupled with an altered molecular circadian clock in response to LD-shift

So far, we have observed that LD shift alters the microbiota and promotes aging in the intestine of CRC-prone mice. LD shift is an established model of circadian disruption known to affect the molecular clock.^6,7 To evaluate the hypothesis that the intestinal changes in the aging hallmarks are coupled with the circadian clock, we examined correlations between the transcriptional expression of aging hallmark genes and circadian clock associated genes (CCGs) (Figure 6A). The CCGs (n = 15) were defined as the genes that control the transcriptional/translational feedback loop of the circadian clock.⁴⁴ We observed that 64 of 76 (84%) aging hallmark genes were significantly correlated (R² ≥ 0.3, p < 0.05) with at least one clock gene, with several aging-related genes strongly correlated with multiple CCGs (52 of 76, 68%). Among the aging hallmark genes, cadherin-encoding genes that are involved in barrier integrity (CDH1, CDH5, and CDH10) correlated with the maximum number of CCGs (n ≥ 5). In contrast, almost all clock genes correlated with multiple aging hallmark genes (Figure 6A).

Figure 6.

Correlation between circadian clock genes and aging hallmark genes

(A) Pairwise correlation of circadian clock genes (n = 14) and aging hallmark genes that were differentially expressed in 18-week non-shifted mice compared with 12-week shifted mice. Red: positive correlation, Blue: negative correlation.

(B) Interaction network of top clock-host gene correlations. Blue: negative correlation; red: positive correlation. Green: Clock genes.

(C) Enriched pathways for the top host genes that strongly correlate with circadian clock genes.

(D and E) Euclidean distance of clock correlation distance matrices for (D) shift-age conditions and (E) prebiotic vs. 18-week shifted conditions. Gray circles represent randomized pairwise correlation measures. ∗∗p < 0.01.

Overall, we identified 155 strong correlations (R² > 0.6, p < 0.01) between the aging hallmark genes and CCGs. These interactions were also observed in the interaction network (Figure 6B). Furthermore, pathway enrichment analysis revealed that biological processes that correlated with CCGs (R² ≥ 0.6, 53 of 76 aging hallmark genes) mostly belonged to aging-related pathways that are involved in the intestinal barrier, including cell communication and adhesion, host-microbe interaction, autophagy and apoptosis, and stress response pathways (Figure 6C). These data suggest that changes in the expression of intestinal aging markers are associated with changes in the circadian clock.

As a potential mechanism for mediating the impact of LD shift on host molecular processes, we examined whether LD shift results in intestinal clock dysfunction. To this end, we estimated the molecular clock progression by calculating the clock correlation distance (CCD), which measures the relative expression of the CCGs as described previously⁴⁵, where a weakened correlation among CCGs implies clock dysfunction. We observed a dampened profile of pairwise correlations between clock genes in 18-week shifted mice when compared with the 12-week shifted condition (Figures S8A and S8B). A similar, though weaker, difference can be seen when 12- and 18-week non-shifted mice were compared (Figures S8C and S8D), consistent with changes in the circadian clock progression as animals grew from early to middle age. Euclidean distance analysis revealed that the shift was the main condition that significantly (p < 0.01) altered CCD when compared to non-shifted mice (Figure 6D). Overall, these results suggest that circadian clock progression in the colon is significantly affected by this shift.

Since restructuring the microbial populations rescued the alterations in the host-microbe profiles, we next asked whether fiber prebiotics mediate these changes through alteration of intestinal clock progression. Measurement of the clock correlation distance (CCD) for prebiotic conditions showed a marked co-expression pattern of clock genes (Figure S8E), signifying (p < 0.01) a comparatively functional clock in reference to the shift-age condition (Figure 6E). Furthermore, aging hallmark genes were strongly correlated (R² > ± 0.5) with clock genes (Figure S8F). These data further confirm that changes in the expression of aging hallmark genes are coupled with the circadian clock, and prebiotic exposure alleviates the shift-age-mediated dampening of the circadian clock.

Circadian clock disruption mimics the impact of an age shift on colon carcinogenesis.

Considering the observed effects of LD shift on the colon circadian clock, which were strongly coupled with alterations in the aging genes in the colon, and the strong coupling of the alterations of the aging genes with those of the clock genes in the colon, we investigated whether the aging-mimicking features are mechanistically linked to changes in the molecular clock. To evaluate the involvement of a clock-dependent mechanism, we generated APC^Δ468Clock^Δ19/Δ19 (APC/ClockMut) mice, as described in the STAR Methods (Figures 7A and 7B). APC/ClockMut mice showed accelerated carcinogenesis when compared to APC^Δ468 (APC) mice with ∼3-fold more polyps (p < 0.001) (Figure 7C). Enhanced carcinogenesis was further supported by the presence of larger polyps in APC/ClockMut mice (Figure 7D). In addition, APC/ClockMut mice had hyperproliferative crypts (p < 0.0001) compared to their APC littermates (Figures 7E and 7F). Similar to the observed intestinal aging in LD-shift mice, APC/ClockMut mice showed early aging features when compared to APC mice, as characterized by decreased barrier markers (E-cadherin, p = 0.040) and higher levels of apoptosis markers (Cleaved caspase-3, p = 0.022) (Figures 7G and 7H).

Figure 7.

Circadian clock is altered with shift-age

(A and B) (A) Generation of the Apc^Δ468Clock^Δ19/Δ19 mouse. Representative adenomatous polyps in the jejunum of 5.5 months old male (Left) Apc^Δ468 and (Right) Apc^Δ468Clock^Δ19/Δ19 mice. Scale bar: 10 mm. (B) Histology of the small bowel from the representative Apc^Δ468 and Apc^Δ468Clock^Δ19/Δ19 mice. Adenomatous polyps are indicated by arrows with enlarged, boxed areas. H&E-stained jelly roll sections of intestine, (1.6x) from representative (i) Apc^Δ468 and (ii) Apc^Δ468Clock^Δ19/Δ19 mice. Scale bar: 1,000um. (iii-iv) and (v-vi) show respective adenomatous polyps at different magnifications (5× and 10x, respectively); Scale bars: 400 μm or 200 μm. (vii-viii) enlarged image of the same at 20x. Scale bar: 100 μm.

(C) The number of polyps in APC-and APC/Clock-mutant mice.

(D) Size of polyps in APC-and APC/Clock-mutant mice. Each bar displays the average from several independent experiments, with error bars representing the standard deviation of those experiments. ∗∗p < 0.01, ∗∗∗p < 0.001.

(E) Ki-67 staining of the healthy colon region from the APC and APC/Clock mutants, revealing a significant increase in Ki-67 immunoreactivity in the APC/Clock mutant. Scale bar: 100 μm. The quantitative analysis in (F) enumerates Ki-67 positive cells in APC and APC/Clock mutant mice at 20x, showing highly significant differences in the colon. ∗∗∗∗p < 0.00001.

(G) E-cadherin and (H) cleaved caspase-3 levels in APC and APC/Clock mutant mice. Representative crypts with low (top) and (high). Scale bars: 50 μm. (G) E-cadherin and (H) cleaved caspase-3 expressions are shown. Data are represented as mean ± SD. ∗p < 0.05.

Stool analysis of alpha-diversity indices showed a significant increase in both Shannon Index (APC: 2.78 ± 0.04, APC/ClockMut: 2.98 ± 0.05; p = 0.005) and Pielou’s Evenness (APC: 0.62 ± 0.01, APC/ClockMut: 0.66 ± 0.01; p = 0.016) within APC/ClockMut mice compared to APC mice (Figure S9A). Beta-diversity indicated that the overall microbial community structures between these two mice groups, at the genus level, were significantly different (ANOSIM: R = 0.45, p = 0.002). In comparison to APC mice, the microbial features of APC/ClockMut mice mimicked a similar microbial fecal environment of dysbiosis observed in LD-shift mice. For instance, differential abundances analysis indicated the genera Adlercreutzia, Coriobacteriaceae Unclassified, Lachnospiraceae Unclassified, Allobaculum, and [Ruminococcus] were lower, while putative pathobiont genera Turicibacter, Rikenellaceae Unclassified, Bacteroidales Unclassified, Clostridiaceae Unclassified, and Desulfovibrionaceae Unclassified relative abundances were higher in APC/ClockMut genotype (Figure S9B).

To expand our findings beyond the polyposis model and to further establish the role of the molecular clock on the microbiota, we compared the fecal microbiota compositions of clock mutation (Clock^Δ19) mice to their age-matched wild-type littermates. We observed a microbial dysbiosis pattern similar to the LD-shift polyposis mice, where out of eight differentially abundant microbes identified by the LD shift, seven were significantly altered in the Clock^Δ19 mice versus the wild-type (Figure S10A). The Clock^Δ19 mice had significantly lower abundances of the beneficial genera Bifidobacterium and Ruminococcus, with increased abundances of the pathogenic genera Helicobacter and Odoribacter, in comparison to the wild-type mice (Figure S10B).

Taken together, these data show that disruption of the molecular clock could result in a similar dysbiosis pattern observed from the environmental mode of circadian disruption by LD shift in CRC-prone mice, where dysbiosis and altered clock accompany age-mimicking changes in the intestinal tissue and accelerated tumorigenesis.

Discussion

Disruption of circadian rhythms is common in modern society and is increasingly recognized as a risk factor for many chronic diseases, including cancer. Shifts in LD have been linked to an increased risk of CRC. Using our established CRC model, we observed that an LD shift accelerated intestinal polyposis in aged animals. These changes in the intestine were preceded by shift-induced enrichment of pathways that represent cellular hallmarks of aging, including the downregulation of intestinal barrier markers. Concomitant changes in the intestinal microbiota were noted, linked to intestinal aging, and verified in human CRC. Our analysis revealed host-microbiome networks that respond to LD shifts and changes in the molecular clock.

Age is a major risk factor for CRC.⁴⁶ Several previous studies have implicated host gene regulation in CRC.^47,48 However, to the best of our knowledge, our work is the first to address the role of LD shift in CRC across different ages. We identified several essential genes and pathways that were differentially expressed due to age shift. One of the most important processes enriched in our analysis was dampening of intestinal barrier markers. Loss of the barrier could allow bacterial penetration into the mucosa and colon epithelium, which is an early key event in mucosal inflammation and polyp formation.⁴⁹ The intestinal barrier is tightly regulated by the interaction of the gut lining, which is composed of intestinal epithelial cells and mucin-secreting goblet cells, with the gut contents, including the microbiota and their products. Our transcriptomic, histological, and microbial analyses indicate a multi-level effect of LD shift on intestinal barrier components starting at early ages. While multiple tissue markers of gut lining integrity were negatively affected by the LD shift, we also noted a decrease in microbial populations that are involved in barrier stability, such as the genera Bifidobacterium and Lactobacillus, which are known to promote the production of SCFAs, a key barrier stabilizer in the colon.^50,51 However, the levels of SCFAs alone could not explain the shift-induced polyposis phenotypes in our model. In addition to the reduction of beneficial prebiotic bacteria, we found that both colonic content and feces of the 18-week and 28-week LD shift aging mouse groups, compared to 12-week shifted mice, showed increased relative abundances of putative pro-inflammatory bacteria such as Bilophila, Helicobacteraceae, Rikenella, Odoribacter, Desulfovibrionaceae, and Bacteroidales that are linked to CRC.^52,53,54 These findings suggest that alterations in intestinal microbial composition occur both with shifting and age, mirroring the patterns seen in colon gene expression.

We concentrated on the intestinal barrier because shift age impacts barrier-related processes (Figure 2) and shows the greatest differential expression of barrier pathway genes (Figure 3). This focus provided us with mechanistic insights into how shift age affects colon microbial-host interactions. However, our data also suggests that other aging-related processes, such as senescence, inflammation, and autophagy, may be involved. These areas merit further investigation to enhance our understanding of how shift age contributes to colon carcinogenesis.

The single disease–single pathogen or gene relationship, though important in the etiology of infectious diseases, is not expected to explain a complex pathology, such as CRC. Co-culture and obligatory cross-feeding studies have suggested that the effects of bacteria can be modulated by other microbial members. Microbe-microbe correlations in our data revealed several correlative microbial abundances upon shift-age, pointing toward an integrated mechanism of action of multiple microbes toward host gene expression and regulation. Since there is a complex interaction between the microbial community and host genes in the gut, we performed a combined network analysis to find dysbiosis patterns by shifting our polyposis model across different ages. By integrating the mucosal microbiome with host gene expression, we found several strong microbe-gene networks.

Furthermore, we found strong microbial-host features, belonging to aging-related genes and pathways, implicated in human CRC pathogenesis. The favorable response of host mechanisms to microbial interventions in our study introduces new targets for CRC prevention in high-risk individuals. However, the contribution of LD shifts to early onset CRCs in humans by inducing intestinal aging needs to be studied separately.

Our study identified an interesting connection between colonic circadian clock dysfunction caused by a shift in the LD cycle, as well as genetic alterations in intestinal molecular clock dysfunction and intestinal aging changes, both in the tissue and microbiota. We found that molecular disruption of the colon clock induces changes in intestinal aging markers and microbiota, which mirror the alterations observed with the LD cycle shift. This suggests that the shift effect is at least partly mediated by molecular clock dysfunction, which is supported by our data indicating the strongest molecular clock dysfunction in the colon of shifted mice. The age-accelerating tumorigenic effect of the LD shift was characterized by strong correlations between the molecular clock and aging-related genes. Among these, significant correlations were observed with key molecular clock genes CLOCK, CRY2, and PER2, each demonstrating distinct co-expression patterns with various aging hallmark genes (Figure 6) including ACTN2, MAPK10, CDH10, and LAMTOR2.⁵⁵ These findings align with the reported molecular connections between circadian clocks and aging-related processes.^56,57 We found half of the included aging hallmark genes in our data to have clock binding in their gene promoters (38/76 genes, Table S1). In addition to transcriptome alterations, molecular clocks can regulate cellular processes via other mechanisms, such as epigenetic alterations or other post-transcriptional regulatory mechanisms, which were not studied here.^56,58,59

In the intestine, where microbes are crucial for regulating host responses, changes in the microbiota due to clock dysfunction in our model significantly influenced the molecular processes involved in intestinal aging and cancer development.^60,61 This aligns with the growing body of evidence that highlights the gut microbiota’s role as a mediator of environmental signals in the host and its influence on various age-related diseases.^62,63,64 Correction of the key host-microbe networks in our model upon prebiotic intervention holds promise that targeting microbial dysbiosis may reverse some of the altered microbial features and eventually promote healthy intestinal aging in at-risk individuals (i.e., shift workers). Additional investigations using conditional or other circadian clock gene knockout (KO) mice are needed to obtain clock-gene specific mechanistic insights. Specifically, conditional KO models can help isolate the effects of the intestinal clock genes independent from the central clock, thereby providing more precise insights into the role of the intestinal clock in intestinal aging homeostasis. We acknowledge both the technical challenges and compensatory mechanisms associated with tissue-specific models and recognize that many circadian clock genes have roles beyond the circadian system, which can complicate the interpretation of phenotypes.⁶⁵

In summary, we found that disruption of circadian rhythms leads to intestinal clock alterations, which, in interaction with aging signals, accelerate colon carcinogenesis in animals. This suggests that a functional circadian system is essential for healthy aging and enhances our understanding of why long-term desynchronization between environmental cues and the internal circadian clock, as seen in shift workers, increases the risk of aging-related diseases in both animal models and humans.^66,67,68

Limitations of the study

Limitations of our study include the use of 16S rRNA sequencing for the analysis of microbiota communities, which does not provide the resolution of metagenomic sequencing for identifying specific bacterial species/strains or functional genes and pathways. While barrier-related processes provided us with mechanistic insights into how shift and age interacted, other aging-related processes, such as senescence and inflammation may also be involved and merit further investigation in their contributions to CRC pathogenesis from circadian disruption. Our study exclusively examined male mice. Finally, while public datasets allowed us to test the clinical implications of the microbial-host features of accelerated intestinal aging derived from our experimental models in human CRCs, these need to be directly tested and demonstrated in future human studies.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Faraz Bishehsari (faraz.bishehsari@uth.tmc.edu).

Materials availability

The study did not generate new unique reagents.

Data and code availability

• •All sequencing data generated in this study have been deposited in the National Center for Biotechnology Information (NCBI) BioProject database. Accession numbers are listed in the key resources table.
• •The software and algorithms performed are listed in the key resources table.
• •Histological and immunofluorescent staining images and/or raw data files are available upon reasonable request from the lead contact.

Acknowledgments

This work was supported by the following National Institutes of Health (NIH) grants: [RO1-CA264048 and RO1-AI108682 to K.K.], [RO1-AA023417 and R24-AA026801 to A.K.], and [RO1-CA279487 and UO1-AG086145 to F.B.]. The authors are grateful to Dr. Fred Turek for providing breeding pairs for Clockmutant mice under award number R01AA020216. We thank the research volunteers and technical staff at RUSH Center for Integrated Microbiome and Chronobiology Research (CIMCR), the RUSH Genomics and Microbiome Core Facility (GMCF), RUSH Research Bioinformatics Core Facility (RRBC), Sequencing Core of the University of Illinois of Chicago (SQC), and all research and technical staff at the RUSH animal facility. Also, this research was supported in part by philanthropic funding from Mr. and Mrs. Larry Field, Mr. and Mrs. Glass, Mrs. Marcia, and Mr. Silas Keehn, the Sklar Family, the Johnson Family, and Mr. Harlan Berk.

Author contributions

Conceptualization: F.B.; Methodology: F.B., K.K., A.K., B.H., R.M.V., C.B.F., L.C., M.A.R., M.S., and A.O.; Investigation: F.B., D.S., P.A.E., K.K., A.K., B.H., R.M.V., C.B.F., L.C., M.A.R., D.A., and A.O.; Data analysis: D.S., P.A.E., A.N., S.J.G., L.C., M.S., D.A.; Visualization: D.S., D.A., P.A.E., A.O.; Funding acquisition: F.B., A.K.I., and K.K.; Project administration: F.B. and A.K.; Supervision: F.B.; Writing – original draft: F.B. and D.S.; Writing – review and editing: F.B., D.S., and P.A.E.

Declaration of interests

D.S. is the founder of Bainom Inc. None of the other authors have conflicts to declare.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Ki-67	Abcam	Abcam Cat# ab15580; RRID:AB_443209
Goat Anti-Rabbit IgG Antibody (H + L), Biotinylated	Vector Laboratories	Vector Laboratories Cat# BA-1000; RRID:AB_2313606
Anti-gamma H2A.X (phospho S139) antibody	Abcam	Abcam Cat# ab11174; RRID:AB_297813
E-Cadherin (24E10) Rabbit mAb	Cell Signaling	Cell Signaling Technology Cat# 3195; RRID:AB_2291471
Cleaved Caspase-3 (Asp175) (5A1E) Rabbit mAb	Cell Signaling	Cell Signaling Technology Cat# 9664; RRID:AB_2070042
Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 555 - Red	Thermo Fisher Scientific	Thermo Fisher Scientific Cat# A-31572; RRID:AB_162543
Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ Plus 488 - Green	Thermo Fisher Scientific	Thermo Fisher Scientific Cat# A32790; RRID:AB_2762833
Biological samples
Colonic Mucosa - hematoxylin and eosin	Processed at Rush University Medical Center - CIMCR	Obtained in this study
Colonic Mucosa - immunohistochemistry	Processed at Rush University Medical Center - CIMCR	Obtained in this study
Colonic Mucosa – RNASeq	Processed at RUSH Genomics and Microbiome Core Facility	Obtained in this study
Feces – Microbiota	Processed at RUSH Genomics and Microbiome Core Facility	Obtained in this study
Colonic Contents - Microbiota	Processed at RUSH Genomics and Microbiome Core Facility	Obtained in this study
Feces - Metabolomics	Processed at Colorado State University ARC-BIO	Obtained in this study
Chemicals, peptides, and recombinant proteins
Antigen Deloaker, 10×	Biocare Medical	Cat# CB910M
Target Retrieval Solution, 10×	Agilent Dako	Cat# S1699
Dual Endogenous Enzyme Block	Agilent Dako	Cat# S2003
Background Sniper	BioCare Medical	Cat# BS966L
Avidin/Biotin Blocking Kit	Vector Laboratories	Cat# SP-2001
Normal Donkey Serum	Jackson ImmunoResearch	Jackson ImmunoResearch Labs Cat# 017-000-121; RRID:AB_2337258
Peroxidase Streptavidin	Jackson ImmunoResearch	Jackson ImmunoResearch Labs Cat# 016-030-084; RRID:AB_2337238
Liquid DAB+ Substrate Chromogen System	Agilent Dako	Cat# K3468
Vector® TrueVIEW® Autofluorescence Quenching Kit	Vector Laboratories	Cat# SP-8400-15
NovaUltra Alcian blue Stain Kit	IHC World LLC	Cat# IW-3000
Critical commercial assays
FastDNA bead-beating Spin Kit for Soil	MP Biomedicals	Cat# 116560300
RNeasy Mini Kit	Qiagen	Cat# 74106
NEXTFLEX Rapid Directional RNA-Seq Kit 2.0	Revvity	Cat# NOVA-5198-03
NEXTFLEX UDI Barcodes (10NT, 1-1,536)	Revvity	Cat# NOVA-534100
Deposited data
16S rRNA, TS4/APC (TS4Cre × APCΔ468) mice LD shift and age colonic content and feces.	National Center for Biotechnology Information	[Database]: [PRJNA627170]
RNA-Seq, TS4/APC (TS4Cre × APCΔ468) mice LD shift and age colonic mucosa.	National Center for Biotechnology Information	[Database]: [PRJNA627202]
16S rRNA, TS4/APC (TS4Cre × cAPClox468) mice prebiotic feces.	National Center for Biotechnology Information	[Database]: [PRJNA523141]
RNASeq, TS4/APC (TS4Cre × cAPClox468) mice prebiotic colonic mucosa.	National Center for Biotechnology Information	[Database]: [PRJNA1086923]
16S rRNA, APC/ClockMutant (ApcΔ468ClockΔ19/Δ19) feces.	National Center for Biotechnology Information	[Database]: [PRJNA529438]
16S rRNA, ClockMutant (Clock Δ19) mice feces.	National Center for Biotechnology Information	[Database]: [PRJNA1165214]
Experimental models: Organisms/strains
TS4Cre × APCΔ468. 12, 18 and 28 weeks of age mice. Constant 12:12 LD cycle. Male. Colonic Content Sample Sizes: 12Wk Non-Shift (n = 7), 12Wk Shift (n = 7), 18Wk Non-Shift (n = 5), 18Wk Shift (n = 5), 28Wk Non-Shift (n = 6), and 28Wk Shift (n = 7). Feces Sample Sizes: 12Wk Non-Shift (n = 7), 12Wk Shift (n = 7), 18Wk Non-Shift (n = 5), 18Wk Shift (n = 7), 28Wk Non-Shift (n = 6), and 28Wk Shift (n = 7).	Laboratory animal model and laboratory experiments of F.B., Rush University Medical Center - CIMCR.	Feces and Colonic Content – 16S rRNA microbiota. Mucosa – RNASeq. Intestines – histology and tissue staining.
TS4Cre × cAPClox468. 18 weeks of age mice (6 weeks, followed by 12 weeks of prebiotic treatment). Male. Sample size n = 5.	Data provided and analyzed from Bishehsari et al.43	Feces – 16S rRNA microbiota. Mucosa – RNASeq.
ApcΔ468. 18 weeks of age mice (8 weeks, followed by 10 weeks of standard chow on 12:12 LD cycle.). Male. Sample size n = 7.	Laboratory animal model of K.K., Mayo Clinic. Laboratory experiments of F.B., Rush University Medical Center – CIMCR.	Feces – 16S rRNA microbiota. Intestines – histology and tissue staining.
ApcΔ468ClockΔ19/Δ19. 18 weeks of age mice (8 weeks, followed by 10 weeks of standard chow on 12:12 LD cycle.). Male. Sample size n = 6.	Laboratory animal model of K.K., Mayo Clinic. Laboratory experiments of F.B., Rush University Medical Center – CIMCR.	Feces – 16S rRNA microbiota. Intestines – histology and tissue staining.
Clock Δ19. 18 weeks of age mice (8 weeks, followed by 10 weeks of standard chow on 12:12 LD cycle.). Male. Sample size n = 7.	Data generously provided and analyzed from Voigt et al.69	Feces – 16S rRNA microbiota.
Wild-Type littermates. 18 weeks of age mice (8 weeks, followed by 10 weeks of standard chow on 12:12 LD cycle.). Male. Sample size n = 8.	Data generously provided and analyzed from Voigt et al.69	Feces – 16S rRNA microbiota.
Oligonucleotides
515F - GTGCCAGCMGCCGCGGTAA	Integrated DNA Technologies	Earth Microbiome Project
806R - GGACTACHVGGGTWTCTAAT	Integrated DNA Technologies	Earth Microbiome Project
Software and algorithms
STAR	Dobin et al.70	https://github.com/alexdobin/STAR
PicardTools	Broad Institute	https://broadinstitute.github.io/picard/
HTSeq	This paper	https://htseq.readthedocs.io/en/latest/
DESeq2	Love et al.14	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
Primer7	https://learninghub.primer-e.com/books/primer-v7-user-manual-tutorial	https://www.primer-e.com/software
Reactome	Jassal et al.71	https://reactome.org/
GraphPad Prism 10.0	This paper	https://www.graphpad.com/
ImageJ	National Institutes of Health	https://imagej.net/
SparCC	This paper	https://github.com/zdk123/SpiecEasi/blob/master/R/spaRcc.R
Cytoscape 3.10.0	Shannon et al.72	https://cytoscape.org/
MicrobiomeAnalyst 2.0	Lu et al.73	https://www.microbiomeanalyst.ca/MicrobiomeAnalyst/home.xhtml

Experimental model and study participant details

Animals

Our study exclusively examined male mice. However, it is unknown whether these findings are relevant to female mice. We previously reported male mouse models of polyposis with constitutive (APC^Δ468),^74,75 and conditional (APC ^lox468) truncation of exons 11 and 12 of the adenomatous polyposis coli (APC) gene.^76,77 TS4Cre mice were previously described.⁷⁸

Light/dark shifting and aging animals

To address the effect of a shift in the LD cycle on colon carcinogenesis, we used TS4Cre × APC^Δ468 (TS4/APC) mice as a preferred model of colon tumorigenesis,¹² in comparison to classical adenomatous polyposis coli (APC) mice, where tumors predominantly develop in the small bowel. TS4/APC mice spontaneously developed polyposis upon loss of heterozygosity, starting from three months of age and growing until the mice became cachectic by seven months of age. Age-matched four-six-week-old male mice were housed individually in cages within ventilated light-tight cabinets. Mice were acclimated to housing for one-week prior to initiating the study and maintained in a constant 12 h of light followed by 12 h of darkness (12L:12D) in light- and temperature-controlled chambers and had continuous access to regular chow (ad libitum) and fresh water. Mice were placed in empty non-bedded cages for 24-h fecal collection at 6, 12, 18, and 28-week of age (n = 5–7 per group/age) for microbial analysis. The mouse feces were stored at −80°C for microbiota analysis (i.e., flash frozen, no preservative). The mice were sacrificed by decapitation at 12, 18, and 28-week of age. The colonic content and mucosa of the mice were collected and stored at −80°C for microbiota (i.e., flash frozen, no preservative) and RNA-Seq (i.e., RNAlater preservative) analyses. The intestine was fixed in 10% buffered formalin for 24 h for polyp characterization, histology, and tissue staining. Animal experiments were conducted at the Rush University Medical Center, Chicago, IL, USA, after approval of the animal protocol by the Institutional Animal Care and Use Committee (IACUC # 14-008).

Prebiotic intervention animals

We previously reported the effects of prebiotic Teklad Envigo #TD160445 high-fiber diet in ameliorating polyposis in TS4/APC (TS4Cre × cAPC^lox468) polyposis mice.⁴³ TS4/APC male mice received a standard rodent chow diet vs. Teklad Envigo #TD160445 for 12 weeks (n = 5 per group). Upon sacrifice at 18 weeks-of age, colonic mucosa and feces were used for RNA-Seq and 16S microbiota analyses, respectively.

Circadian clock disrupted animals

Clock mutant (Clock^Δ19/Δ19) mice have been previously described.⁶⁹ Briefly, ∼8-week-old Clock mutant (n = 7) and wild type male mice (n = 8) were individually housed on a 12:12 LD cycle and fed a standard rodent chow diet. After 10 weeks, stool samples were collected and used for microbiota analyses at 18-week of age.

Homozygous colonies of Clock^Δ19/Δ19 (ClockMut) and APC^Δ468 (APC) were interbred to produce APC^Δ468Clock^Δ19/Δ19 mice. Male APC^Δ468 (APC) and APC^Δ468Clock^Δ19/Δ19 (APC/ClockMut) mice colonic mucosa (histology and staining) and feces (microbiota) (n = 6–7 per group) were collected at 18-week of age and examined for this study.

Method details

Histology

The colonic mucosa was dissected, flushed with phosphate-buffered saline (PBS), and linearized longitudinally. Tissues were incubated in 10% formalin for 12–18 h and fixed overnight at room temperature. Swiss-rolling, paraffin embedding, tissue sectioning, and hematoxylin and eosin (H&E) staining were performed in the laboratory.

For H&E and immunohistochemistry staining, 5 μm thick sections were dewaxed and then hydrated using xylene and alcohol/water. For immunohistochemistry, antigen retrieval was performed using Antigen Deloaker (Cat# CB910M, Biocare Medical) and Target Retrieval Solution (Cat# S1699, Agilent Dako). Following antigen retrieval, tissues were washed with PBS and nonspecific background staining was blocked using a dual endogenous enzyme block (Cat# S2003; Agilent Dako), Fc-block (2.4G2, Antibody Hybridoma Core, Mayo Clinic; kindly provided by T. Beito), and Background Sniper (Cat# BS966L; BioCare Medical). Nonspecific avidin/biotin was blocked (Cat# SP-2001; Vector Laboratories). For Ki-67 staining, anti-Ki-67 (dilution: 1/300; Cat# ab15580, Abcam) as the primary antibody overnight and biotinylated goat anti-rabbit (Cat# BA-1000, Vector Laboratories) as the secondary antibody were applied to the sections for 45 min, followed by streptavidin (HRP conjugate, Cat# 016-030-084, Jackson Laboratories) for 30 min. Counterstaining was performed using Chromogen DAB + Substrate (Cat# K3468, Agilent Dako), followed by hematoxylin counterstaining. A Leica light microscope mounted with a Zeiss Axiocam 503 camera was used for imaging of IHC staining.

Immunofluorescence staining

For immunofluorescence staining, 4-6 μm-thick sections were deparaffinized and then rehydrated using a series of alcohol dilutions in distilled water (100%, 95%, 90%, 70%, and 0%). Then, we conducted antigen retrieval by subjecting the slides to high heat in a pressure cooker for 10 min while immersing them in Target Retrieval Solution (Cat# S1699, Agilent Dako). The sections were subsequently blocked using a solution containing 2% bovine serum albumin (BSA) and 10% donkey serum from Jackson ImmunoResearch (Cat# 017-000-121) and incubated for 1 h. After three washes, the primary antibodies were added. We used a 1:500 dilution of Anti-gamma H2AX (phospho S139) rabbit antibody (Cat# ab11174, Abcam) for H2AX, a 1:500 dilution of E-cadherin (24E10) rabbit mAb (Cat# 3195, Cell Signaling) for E-cadherin, and a 1:200 dilution of Cleaved Caspase-3 (Asp175) (5A1E) rabbit mAb (Cat# 9664, Cell Signaling) for Cleaved Caspase 3. Primary antibodies were incubated overnight. The next day, after six washes, the slides were treated with 1:250 dilutions of the respective secondary antibodies (donkey anti-rabbit IgG Alexa FluorTM 555 (red, Cat# A-31572, Thermo Fisher Scientific) and Alexa FluorTM 488 (green, Cat#, A32790, Thermo Fisher Scientific)) for 45 min. This was followed by another six washes. We then applied DAPI to stain the nucleus. Finally, we applied a quenching reagent (Vector TrueVIEW Autofluorescence Quenching Kit, Cat# SP-8400-15, Vector Laboratories) to reduce unwanted autofluorescence, according to the manufacturer’s instructions, before mounting with a fluoromount aqueous medium. Immunofluorescent pictures were obtained with an Inverted Axio Observer 7 fluorescent microscope from Zeiss (491917-0001-000Z). The unprocessed CZI photos were converted to TIF format and transmitted to a computer. Images were analyzed using ImageJ software. Three crypts from three random locations in the colon rolls were chosen and the Corrected Total Cell Fluorescence (CTCF) of H2AX in the selected enterocytes was measured. The apical layer of the colon was chosen for E-cadherin analysis followed by CTCF quantification. We measured epithelial cells expressing Cleaved Caspase 3 in their cytoplasm in at least three sections per mouse across the field of view (FOV) for Cleaved Caspase 3 analysis.

We used Alcian blue dye for goblet cell staining according to the manufacturer’s instructions ( NovaUltraTM Alcian blue Stain Kit, Cat# IW 3000A, IHC World LLC.). We quantified the number of goblet cells and epithelial cells per 8 crypts in three randomly selected regions of the colon. We calculated the ratio of goblet cells to epithelial cells (goblet/epithelial) for each crypt.

Sample preparation of total RNA sequencing

RNA was isolated from the colonic mucosa of both LD aging and shifting mice, plus TS4Cre×cAPCl^°x468 mice with prebiotic fibers, using the TRIzol extraction procedure. Briefly, frozen tissues were lightly ground in a mortar and pestle and were constantly submerged in liquid nitrogen. Frozen tissue–10-100 mg was placed in RNase-free tubes containing TRIzol and RNase-free stainless-steel beads (NextAdvance, USA). Tissues were homogenized using a bullet blender at 4°C. Chloroform was used to separate the phases, and the RNA-containing aqueous phase was subjected to column purification using the RNeasy Mini RNA extraction kit (Qiagen, USA). RNA was treated with DNase, and purified RNA was checked for integrity using an Agilent Bioanalyzer, all samples had an RNA integrity number above 8.0. RNA-seq libraries were prepared using Revvity NEXTFLEX Rapid Directional RNAseq kit 2.0 (Cat# NOVA-5198-03) with 1536 UDI barcodes (Cat# NOVA-534100). Libraries were pooled to equal molarity and sequenced on an Illumina NovaSeq (2 × 100 bp) to achieve a minimum of 40M reads per sample at the Genomics and Microbiome Core Facility (GMCF) at Rush University Medical Center. The Rush Research Bioinformatics Core (RRBC) downloaded FastQ files to compute clusters for data processing.

RNA-seq data pre-processing

Reads from the RNA-sequencing data were aligned to the Mus musculus genome assembly GRCm38 (mm10) using STAR software.⁷⁰ Duplicated aligned reads were marked and removed using Picardtools software. Gene expression count data were extracted using HTseq software. Raw count data were normalized followed by log2 transformation. We filtered out genes with mean read counts (<6). All data preprocessing was performed using the R software.

Differential gene expression analysis

Genes with low counts were removed, and “expressed genes” were defined as those with at least six counts in six samples. VST was used to normalize RNA-seq across the 12 samples. Principal component analysis (PCA) plots were used as an unbiased approach to visualize the distribution of clusters by age. DESeq2 was performed to address age-related changes in overall levels of gene expression in young vs. aged mice, and DEGs significance defined as p-value <0.05 with a fold-change cutoff ≥1.5.¹⁴ Venn diagrams, heatmaps, enhanced volcano plots, and Log2 fold-change graphs were generated using the R programming language.

Aging hallmark genes

Aging hallmark genes are genes known to be involved in aging-associated processes. Aging hallmark genes were compiled from references described in the results section.

Pathway enrichment

Pathway enrichment analysis for differentially expressed and rhythmic genes was performed with Reactome⁷¹ using a hypergeometric statistical test and Benjamini-Hochberg (BH) false discovery rate (FDR) correction.⁷⁹ The redundant pathway terms were merged to create a parent term.

DNA extraction and microbiota analysis

Total DNA was extracted from mouse feces or colonic contents using a FastDNA bead-beating Spin Kit for Soil (MP Biomedicals, Solon, OH, USA). High-throughput amplicon sequencing was conducted using primers (515F/806R) targeting variable region 4 (V4) of the 16S ribosomal RNA (rRNA) genes⁸⁰ using a modified two-step targeted amplicon sequencing approach, as previously reported.⁸¹ Sequencing of all mice was performed using Illumina MiSeq (Illumina, San Diego, CA, USA), as previously described.⁷ All mouse studies had similar sequence processing, quality assessment, clustering, and biological observation matrix analyses.^7,43,69 Diversities including alpha-diversity and beta-diversity metrics are calculated after sample rarefaction depths of 34,000 (LD Shift + Aging: feces and colonic content) or 44,000 (Prebiotic: feces).

Alpha- and beta-diversity metrics were used to examine changes in microbial community structure between fecal and colonic mucosa samples. Alpha-diversity indices (Shannon, Simpson, richness, and evenness) were generated using the package ‘vegan’ implemented in the R programming language.⁸² To examine differences in community composition between samples, a pairwise Bray–Curtis dissimilarity (non-phylogenetic) metric was generated using the Primer7 software package and used to perform analysis of similarity (ANOSIM) calculations. ANOSIM was performed at the taxonomic level of the genus using square-root transformed data employing 999 permutations. Differences in the relative abundance of individual bacterial genera were assessed as previously described.^7,43,69 R-package random-forest algorithm (Boruta) was employed (iterations = 1,000) to detect genera of importance.⁸³ Statistical significance p < 0.05. Differential abundance and stacked histogram graphical visualizations were conducted using GraphPad Prism 10.0 (GraphPad Software, San Diego, California, USA).

Fecal metabolites

SCFA metabolites were measured over time in mouse fecal samples. Feces from the LD aging and shifting mice groups were collected after 6-, 12-, 18- and 28-week followed by homogenization in carbonate–phosphate buffer (one-part feces with three-part buffer). The samples were then centrifuged at 13,000 rpm for 10 min. The supernatant (400 μL) was combined with 100 μL of internal standard (5% phosphoric acidcontaining 50 mM of 4-methylvaleric acid and 8% of copper sulfate) and SCFA analysis was performed using a gas chromatograph equipped with a fused silica capillary column (NukonTM, Supelco No: 40369-03A, Bellefonte, PA, USA) and a flame ionization detector (GC-FID 7890A, Agilent Technologies, Inc., Santa Clara, CA, USA), as previously described.⁸⁴ SCFA quantification was assessed by measuring the peak areas of acetate, propionate, and butyrate relative to that of 4-methyl valeric acid.

Quantification and statistical analysis

The Shapiro-Wilk test was used to assess whether the data were normally distributed. In the case of normal distribution, Student’s t test (e.g., two comparisons) or one-way analysis of variance (ANOVA) (e.g., three or more comparisons) was used to assess differences between mice groups. If the data were not normally distributed, the non-parametric Mann-Whitney U test (e.g., two comparisons) or Kruskal-Wallis test (e.g., three or more comparisons) was used to assess differences between mice groups. The BH method was used to correct for multiple tests. Spearman correlations were used to assess the top host-microbe and microbe-microbe interactions. Only significantly strong positive and negative correlations (R < −0.5, R > 0.5; p < 0.05) generated from SparCC were exported and visualized as network plots within the open-source platform Cytoscape 3.10.0.⁷² Within the Cytoscape network plots, the thickness of the nodes or individual features was calculated using the cytoHubba module within Cytoscape 3.10.0. The nodes within cytoHubba were ranked using the Maximal Clique Centrality (MCC) method.^23,85 Statistical significance p < 0.05. The mean and standard deviation values are shown. Statistical analyses and graphical visualizations were, unless stated otherwise, conducted using GraphPad Prism 10.0 (GraphPad Software, San Diego, California, USA) or MicrobiomeAnalyst 2.0.⁷³

Additional resources

Animal experiments were conducted at the Rush University Medical Center, Chicago, IL, USA, after approval of the animal protocol by the Institutional Animal Care and Use Committee (IACUC # 14-008).

Published: December 9, 2024