Post-fast refeeding enhances intestinal stem cell-mediated regeneration and tumourigenesis through mTORC1-dependent polyamine synthesis

For more than a century, fasting regimens have improved health, lifespan, and tissue regeneration in diverse organisms, including humans. However, how fasting and post-fast refeeding impact adult stem cells and tumour formation has yet to be explored in depth. Here, we demonstrate that post-fast refeeding increases intestinal stem cell (ISC) proliferation and tumour formation: Post-fast refeeding augments the regenerative capacity of Lgr5+ intestinal stem cells (ISCs), and loss of the tumour suppressor Apc in ISCs under post-fast refeeding leads to a higher tumour incidence in the small intestine and colon than in the fasted or ad libitum (AL) fed states. This demonstrates that post-fast refeeding is a distinct state. Mechanistically, we discovered that robust induction of mTORC1 in post-fast-refed ISCs increases protein synthesis via polyamine metabolism to drive these changes, as inhibition of mTORC1, polyamine metabolite production, or protein synthesis abrogates the regenerative or tumourigenic effects of post-fast refeeding. Thus, fast-refeeding cycles must be carefully considered when planning diet-based strategies for regeneration without increasing cancer risk, as post-fast refeeding leads to a burst not only in stem cell-driven regeneration but also in tumourigenicity.

is available in the Supplementary Files section.
There is NO Competing Interest.

50
A critical question in regenerative medicine is whether long-term dietary strategies 51 can promote tissue regeneration without increasing cancer risk. For over a century, fasting 52 interventions, including short-term fasting, iterative fasting, and caloric restriction, have been 53 reported to extend life-span and enhance tissue regeneration in diverse species 1-6 , but whether  26 . In this model, tamoxifen administration activated tdTomato labelling in 6 ISCs and their progeny. We then quantified the tdTomato positive length along the crypt-villus 112 axis after 2 d of tamoxifen administration under fasting and refeeding regimens (Fig. 1c). We 113 found that 1 d refeeding increases the tdTomato-positive length by 60% in the small intestine 114 and 35% in the colon compared to that of the AL controls, indicating that post-fast refeeding 115 boosted ISC-mediated epithelial proliferation (Fig. 1d, Extended Data Fig. 1k). We also 116 assessed whether post-fast refeeding mediates ISC-driven repair after radiation-induced injury. 117 Using the aforementioned Lgr5 ISC lineage tracer model, we labeled ISCs 24 h prior to 118 radiation exposure with tdTomato, subjected mice to fasting/refeeding regimens, and then 119 tracked tdTomato-positive cells for 3 d (Fig. 1e). Like homeostasis, the 1 d post-fast refeeding 120 group showed the greatest effect on ISC-mediated repair after radiation-induced injury (Fig.   121 1f). Overall, our data support the conclusion that many benefits of fasting occur after refeeding 122 by stimulating ISCs to generate greater numbers of progeny in homeostasis and after injury. To better understand how nutrient-sensing signalling pathways mediate ISC 126 adaptation in response to fasting and refeeding, we assessed how these fasting and refeeding 127 influenced the insulin/PI3K and mTORC1 pathways, which are known to promote eukaryotic 128 cell growth and metabolism in response to environmental inputs, including nutrients and 129 growth factors 27 . We first confirmed that the fasting/refeeding cycle changes these signalling 7 responses: A 24-h fast caused blood glucose levels to decline by 50% compared to the AL state, 131 which returned to AL baseline 30 min after post-fast refeeding (Extended Data Fig. 2a). 132 Elevations in blood glucose levels stimulate insulin secretion, which in turn activates 133 intracellular PI3K-AKT signalling and downstream mTORC1 signalling 28,27 . Indeed, in 134 intestinal crypts, we observed increased levels of phosphorylated AKT (S473) protein, which 135 is a marker for AKT activation downstream of insulin-PI3K signalling, and increased levels of 136 phosphorylated 40S ribosomal protein S6 (pS6) and eukaryotic translation initiation factor 4E-137 binding protein (p4EBP1), which indicate that mTORC1 activity is also activated shortly after 138 refeeding (Extended Data Fig. 2b). To determine whether changes in mTORC1 activity occur 139 downstream of the insulin-PI3K pathway, we administered a dual inhibitor of the insulin 140 receptor (IR) and IGF-1 receptor (IGF-1R), OSI-906, and a selective PI3K inhibitor, BKM120, 141 prior to refeeding 29,30 . Intestinal crypt immunoblots confirmed that treatment with these 142 inhibitors abrogated the induction of pAKT, pS6, and p4EBP1 by refeeding (Extended Data 143 Fig. 2b), demonstrating that PI3K-AKT signalling is required to induce mTORC1 activation in 144 the refed state. 145 We next measured mTORC1 activity in intestinal crypts and sorted ISCs and early 146 progenitors at 1 or 3 days of refeeding. We identified that mTORC1 activity both at the early 147 (refed 1 h) and the later refeeding (refed 1 d) time points were greater than that of the AL state 148 (Fig. 2a). These changes were also noted at the crypt bottoms, including in Lgr5+ ISCs and 8 their early daughter progenitors, as we saw increased pS6 levels by immunohistochemistry 150 (IHC) and immunoblots in flow-sorted Lgr5-GFP high ISCs and Lgr5-GFP low progenitors ( Fig.   151 2b-c). To determine whether elevated mTORC1 activity was required for refeeding-induced 152 proliferation and regeneration, we administered the mTORC1 inhibitor rapamycin shortly 153 before or during the refeeding phase. As expected, rapamycin treatment abrogated pS6 level in 154 crypts (Extended Data Fig. 2c). Rapamycin treatment did not entirely block the refeeding-  fasted for 24 h (f), refed for 24 h (rf), and refed with rapamycin treatment (rf +rapa) (Fig. 3a). 9 We performed UMAP analysis to partition GFP+ cells into 17 clusters based on the expression 169 of established marker genes (Fig. 3b, Extended Data Fig. 3a) 31 , indicating that ISCs and 170 progenitor cells are composed of various heterogeneous states. We identified three ISC clusters 171 (clusters 5, 2, and 10) based on Lgr5 expression levels (Fig. 3c), which closely matched 172 previously identified ISC clusters by Biton and colleagues (termed Biton ISC-I, II, and III) 32 .  Focusing on these three clusters, we first observed that their relative proportions did not  dampened with rapamycin treatment (Fig. 3f). Protein analysis suggested a progressive 202 increase in OAT proteins over 24 h of refeeding (Fig. 3g). OAT is a mitochondrial enzyme that 203 is mainly expressed in the liver, intestine, brain, and kidneys. In the intestine, OAT converts 204 ornithine from pyrroline-5-carboxylate (P5C) that is produced from proline and glutamate.

205
Ornithine is a non-proteogenic amino acid used for either citrulline (urea cycle) or polyamine 11 synthesis ( Fig. 3h) 33 . To understand how refeeding alters ornithine metabolism, we measured 207 the levels of proline, glutamate, and ornithine in the intestinal tissue of mice that were subject 208 to AL, fasted, refed for 4 h, and refed for 24 h. Notably, fasting decreased the levels of proline, 209 glutamate, and ornithine in the intestine, while post-fast refeeding significantly increased their 210 levels, starting from 4 h of feeding (Fig. 3i). Taken together, we uncovered that refeeding       Fig. 6d).

305
Tsc1 co-deletion with Apc in ISCs resulted in a greater tumour burden (Extended Data Fig. 6e), 306 supporting the notion that refeeding via mTORC1 stimulation promotes the tumourigenicity of 307 Lgr5+ ISCs. 308 We also addressed the necessity of polyamines and protein synthesis in actuating the 309 increased tumour-initiating capacity of the refed state. To inhibit polyamine production, we 310 administered the ODC inhibitor DFMO to AL or refed Apc loxp/loxp : Villin-CreERT2 mice (Fig.   311 6c) and found that ODC inhibition attenuated tumour formation in the refed intestine and colon 312 with no effect on the AL cohort (Fig. 6d, Extended Data Fig. 6f). Also, given that mTORC1 which persists for at least 1 d after refeeding (Fig. 2). Thirdly, post-fast refeeding stimulates 335 global protein translation, which is coordinated, in part, by increased mTORC1 activity and 336 polyamine metabolite synthesis (Fig. 6). Together, these findings support a model in which 337 fasting primes (or permits the survival of) ISCs for a robust regenerative response upon 338 nutritional stimulation (i.e., post-fast refeeding). 18 We also uncovered that polyamine metabolites mediate many aspects of post-fast

CONTACT FOR REAGENT AND RESOURCE SHARING 416
Further information and requests for resources and reagents should be directed to and will be fulfilled Med Chem Express) was suspended in 5% DMSO, 40% PEG-300, 5% Tween-80, and 50% ddH2O at 456 a concentration of 6.25 mg/ml, then administered by oral gavage at 160 µl per 20g (50 mg/kg).  and Parallel Computing toolboxes as described previously 56 . The resulting data included the intensity 583 of each mass peak in each analyzed sample. Peak picking was done for each sample once on the total 584 profile spectrum obtained by summing all single scans recorded over time, and using wavelet 585 decomposition as provided by the Bioinformatic toolbox. In this procedure, a cutoff was applied to filter 586 peaks of less than 5,000 ion counts (in the summed spectrum) to avoid the detection of features that are 587 too low to deliver meaningful insights. Centroid lists from samples were then merged to a single matrix 588 by binning the accurate centroid masses within the tolerance given by the instrument resolution. Starting Germany). LC-MS grade acetic acid, formic acid, and ammonium formate were purchased from Fisher 602 Scientific (Schwerte, Germany). PBS (1x) was purchased from CLS (Eppelheim, Germany). 603 Dansylchloride, spermidine, spermine, putrescine, N1-Ac-putrescine, N1-Ac-spermine, spermidine-604 2 H6, spermine-butan-2 H8 were from Sigma Aldrich (Steinheim, Germany). Ultrapure and desalted water 605 with a resistivity of 18.2 M Ω/cm was generated by a Sartorius Stedim water purification system 606 (Sartorius, Goettingen, Germany). All other chemicals were purchased by local distributor in the highest 607 possible grade. 608 for 2 minutes and subject to the same freezing-thawing cycle described before. The thaw samples were 619 again vortex and sonicated for 2 minutes, respectively. Next, the samples were centrifuged at 3000g for 620 5 minutes and the supernatant is collected. The combined supernatant was dried in a vacuum-centrifuge 621 for 45 min until complete dryness. The dried extract is resuspended in 200 µL acetonitrile/water (50/50; 622 v/v) by sonicating and vortexing for 2 minutes, respectively. 623

Dansylation of Polyamines 624
Dansylation was carried out by using 80 µL of the obtained extracted of a standard solution with a 625 known concentration. The solution was diluted with 80 µL of water, 40 µL of PBS (1x) and 20 µL of 626 1M NaOH. The reaction was started by adding 80 µL of a dansylchloride solution (10 mg/mL in 627 acetone). The mixture was incubated at 55 °C for 30 minutes. Afterwards, the solution was dried in a 35 vacuum centrifuge to complete dryness for 45 minutes. The residue was dissolved in 80 µL of 629 acetonitrile/water (50/50; v/v) and subsequently analyzed by LC-MS. 630

LC-MS analysis and quantification 631
For polyamine analysis, an Agilent 1290 Infinity LC system coupled to an Agilent 6470 QqQ-MS was 632 used (Agilent Technologies Inc., Waldbronn, Germany). Liquid chromatographic separation was carried 633 out using a Zorbax Eclipse Plus C18 RRHD (50 x 2.1 mm, 1.8 µm; Agilent Technologies Inc., 634 Waldbronn, Germany). The elution was carried out by utilizing a gradient at a flow rate of 400 μL/min 635 with water (10 mM ammonium formate pH 3.5) as solvent A and acetonitrile/water (90/10; v/v; 10 mM 636 ammonium formate pH 3.5) as solvent B. The linear gradient was: 0 min, 15% B; 2 min, 15% B; 7 min, 637 100% B; 10 min, 100% B followed by 2 min at initial condition for re-equilibration. Column 638 temperature was 45 °C, injection volume 10 µL. Ionization was carried out in ESI positive mode by 639 using the Agilent jet stream source. The following MS parameters were used: capillary 4500 V, nozzle 640 voltage 1000 V, gas temp. 275 °C, gas flow 10 L/min, nebulizer gas pressure 25 psi, sheath gas temp. 641 350 °C, sheath gas flow 10 L/min. Detection was carried out in selected reaction monitoring (SRM) 642 using the following optimized transitions for the transitions and parameters shown in table 1. Cycle 643 time was 500 ms. Quantification of N1-Ac-putrescine, N1-Ac-spermine, putrescine, spermine and 644 spermidine in crypt cells was performed by an external calibration in a range from 0.1 µM to 90 µM 645 using 2 deuterated internal standards (spermidine-2 H6 and spermine-butan-2 H8) at a concentration of 10 646 36 µM. For calibration, the analyte to internal standard area ratios were linearly fitted to the corresponding 647 concentration ratios and compared to the area ratios detected in the samples. The cellranger (version 3.1.0, 10x Genomics) application mkfastq was used to deconvolute raw fastq 663 sequences and the application count was used to collapse UMIs, map to the mm10 mouse genome 664 assembly and prepare count matrices. The cellranger filtered_feature_bc_matrix output data were 37 imported into Seurat (version 3.2.0.) 57 and genes detected in less than 3 cells per sample were excluded 666 as were cells with fewer than 200 detected genes. No additional filtering was done based on 667 mitochondrial gene expression. These criteria resulted in 4760, 4282, 4552 and 4467 cells for al, f, rf 668 and rf.rapa respectively. The samples were merged into a single Seurat object and log normalized with 669 a scaling factor of 10000. The vst method was used to identify the top 2000 most variable gene and 670 dimensionality reduction was done using UMAP using 30 principal components. Initial examination of 671 UMAP plots identified a subtle systemic difference between samples that were accounted for by dataset 672 integration. Cell clusters were identified using a resolution value of 0.8 and clusters were assigned to 673 cell type by examining the expression of individual genes or ssGSEA projections. Differential 674 expression testing between clusters was done with the wilcox test implemented in the FindMarkers 675 function. Pre-ranked GSEA was run using GSEA version 4.1 58 using custom gene sets derived from 676 Biton's paper 32 with log2 fold change from the differential expression testing as a ranking metric. To 677 assess the cell cycle status in each cluster, the human g2m and s cell cycle gene lists provided by the 678 Seurat package were mapped to mouse orthologs using Mouse Genome Informatics orthology data. The 679 Percent Feature Set function was used to annotate each cell for the expression level of these sets. A (b) Organoid-forming assay for FACS-sorted ISCs from AL, Fasted, Refed 1d, and Refed 3d mice. Quanti cation (left) and representative images of day 3 organoids (right). n = 5 mice per group. Scale bar,