DDIT4 licenses only healthy cells to proliferate during injury-induced metaplasia

Abstract

Background and Aims:

In stomach, metaplasia can arise from differentiated chief cells that become mitotic via paligenosis, a stepwise program. In paligenosis, mitosis initiation requires reactivation of the cellular energy hub mTORC1 after initial mTORC1 suppression by DNA damage induced transcript 4 (DDIT4 aka REDD1). Here, we use DDIT4-deficient mice and human cells to study how metaplasia increases tumorigenesis risk.

Methods:

A tissue microarray of human gastric tissue specimens was analyzed by immunohistochemistry for DDIT4. C57BL/6 mice were administered combinations of intraperitoneal injections of high-dose tamoxifen (TAM) to induce Spasmolytic Polypeptide Expressing Metaplasia (SPEM) and rapamycin to block mTORC1 activity, and N-methyl-N-nitrosourea (MNU) in drinking water to induce spontaneous gastric tumors. Stomachs were analyzed for proliferation, DNA damage, and tumor formation. CRISPR/Cas9-generated DDIT4^−/− and control human gastric cells were analyzed for growth in vitro and in xenografts with and without 5-fluorouracil (5-FU) treatment.

Results:

DDIT4 was expressed in normal gastric chief cells in mice and humans and decreased as chief cells became metaplastic. Paligenotic Ddit4^−/− chief cells maintained constitutively high mTORC1, causing increased mitosis of metaplastic cells despite DNA damage. Lower DDIT4 expression correlated with longer survival of gastric cancer patients. 5-FU-treated DDIT4^−/− human gastric epithelial cells had significantly increased cells entering mitosis despite DNA damage and increased proliferation in vitro and in xenografts. MNU-treated Ddit4^−/− mice had increased spontaneous tumorigenesis after multiple rounds of paligenosis induced by TAM.

Conclusions:

During injury-induced metaplastic proliferation, failure of licensing mTORC1 reactivation correlates with increased proliferation of cells harboring DNA damage, as well as increased tumor formation and growth in mice and humans.

Introduction

Tissue injury stimulates fully differentiated cells to proliferate and reenter the cell cycle to replace damaged cells and fuel regeneration^{1,2, 3}. Such injury-induced reprogramming of mature cells into progenitors is executed by an evolutionarily conserved, stepwise cellular program called paligenosis (Figure S1). The stages of paligenosis are: 1) the cellular master energetics-regulator mTORC1 is quenched as lysosomal and autophagic activity ramps up to dismantle mature cell features; 2) expression of progenitor or embryonic genes is induced; 3) mTORC1 is reactivated to initiate S-phase of mitosis and repair the tissue⁴. Understanding how paligenosis works may be critical for understanding tumorigenesis, as precancerous lesions like metaplasia can arise from mature cells and are often characterized by expression of embryonic and progenitor genes^5–7.

The idea that mature cells can retool their energetics to fuel regeneration and cause precancerous lesions is well over a century old. Pathologist George Adami thought injury invoked a program to retool mature cell energy usage from fueling physiological function to fueling proliferation^{8, 9}. Specifically in the stomach, the link between dedifferentiation of mature cells into mucus-secreting, more primitive cells has been considered a key initiating event in gastric cancer since at least the time of Ménétrier^{10, 11}. The lesion containing potentially precancerous mucous cells has been referred to as chronic atrophic gastritis (most commonly used by pathologists today), spasmolytic polypeptide expressing metaplasia (SPEM, based on the pattern of gene expression), and pseudopyloric metaplasia (based on histological similarity to antrum/pylorus)^{12, 13}.

By dissecting the mechanisms that govern paligenosis, we can focus on the core set of genes that evolved specifically to regulate entry into metaplasia and govern how metaplasia normally resolves without becoming chronic or progressing to cancer. A key early event in the paligenosis that generates SPEM cells from digestive-enzyme-secreting, differentiated chief cells is activation of DDIT4¹⁴. DDIT4 is an evolutionarily conserved, injury-induced scaffold that suppresses the cellular energy hub mTORC1 by blocking 14-3-3 proteins from interacting with the TSC1/2 complex, thereby inactivating Rheb, a small GTPase that activates mTORC1^15–18. Early after paligenotic injury, DDIT4 is induced to suppress mTORC1¹⁴. DDIT4 slowly degrades as paligenosis progresses, but mTORC1 remains suppressed as the injury activates p53, an mTORC1 inhibitor. Healthy cells, however, concomitantly accumulate IFRD1, which suppresses p53, derepressing mTORC1 and permitting its reactivation to fuel S-phase entry.

Because p53 blocks only reactivation of mTORC1 after initial mTORC1 suppression by DDIT4, paligenotic cells in Ddit4^−/− mice would be predicted to bypass p53 to re-enter the cell cycle. Long-lived, post-mitotic cells like chief cells may be more likely to have accumulated oncogenic mutations that are unmasked when the cells begin to proliferate^{13, 19}. Thus, we reasoned that the aberrant paligenosis wherein DDIT4-deficient cells bypass a p53 checkpoint could be used as a tool to study the functional consequences of aberrant metaplasia.

Here, we focus on the failure of licensing in cells lacking DDIT4 during the paligenosis that causes SPEM. We show that DDIT4 is expressed specifically in normal mouse and human chief cells, and its expression decreases in patients with chronic metaplasia and dysplasia. Patients with low expression have poorer survival. In our acute, rapid, synchronous SPEM induction model using high-dose tamoxifen (TAM), we see early, brief induction of DDIT4 and later extinction. Loss of DDIT4, as expected, caused dysregulated, constitutively increased mTORC1 in paligenosis with a functional consequence of the failed checkpoint being increased cells proliferating despite DNA damage. In human gastric adenocarcinoma cells, loss of DDIT4 also increased mitosis despite DNA damage and despite chemotherapeutic agents in vitro and in xenografts into mice. Finally, we show that Ddit4^−/− mice had significantly increased MNU-catalyzed spontaneous tumorigenesis after multiple rounds of TAM-induced, paligenosis-fueled SPEM.

Materials and Methods

Regulatory Approval

All human gastric pathological tissue specimens included in this paper were approved by the Institutional Review Board Ethics Committee of Washington University School of Medicine or China Medical University. All experiments using animals followed protocols approved by the Washington University School of Medicine or China Medical University Animal Studies Committees.

Animal studies and associated reagents

SPEM was induced by Tamoxifen (5 mg/20 g body weight; Toronto Research Chemicals) dissolved in 10% ethanol and 90% sunflower oil, injected intraperitoneally (IP) daily with mice euthanized at 12–72h as described previously²⁰. Rapamycin (60 μg/20 g body weight; LC Laboratories) was injected IP in 0.25% Tween-20, 0.25% polyethylene glycol in PBS daily for 3 days pre-treatment and throughout the injury time course. 5-bromo-2’-deoxyuridine (BrdU; 120 mg/kg) and 5-fluoro-2’-deoxyuridine (12 mg/kg) in sterile water 90 min was injected IP before sacrifice for all BrdU labeling experiments.

Gastric cancer cell DDIT4 knockout

To generate a stable DDIT4 knockout MGC-803 cell line, a published protocol described previously²¹ was used with minor modifications. Three gRNA sets were designed and cloned into pSpCas9(BB)-2A-Puro plasmids (Addgene, PX459) to make double nicks on the DNA strand. These were applied to MGC-803 cells and selected for with puromycin (200 μg/ml) for two weeks. At least 20 single-cell derived clones were established and confirmed via PCR. The DDIT4 knockout MGC-803 cell lines (DDIT4^−/−) and negative control (NC) cell lines (DDIT4^+/+) were further confirmed via western blot against DDIT4.

Statistical analyses

All statistical analyses were performed using SPSS 19.0 software (IBM, only for Kaplan–Meier survival analysis) or GraphPad Prism 7 (La Jolla). Overall survival was determined using Kaplan–Meier curves, with an event defined as a cancer-related death. The log-rank test was used to assess differences between the survival curves in different groups. In histopathological univariate analysis, continuous data were analyzed by a two-tailed Student’s t-test, and categorical data were evaluated by a two-tailed χ² test. For normally distributed mouse or cell data, statistical analysis of significance was by t-test in the case of 2-sample comparisons or ANOVA with post-hoc inter-sample comparison to evaluate significance when there were ≥3. For 2-sample data whose distribution was not demonstrated or expected to be normal, the Mann-Whitney Test was used. All data were expressed as mean±standard deviation or standard error of mean (SEM), depending on experiment design. P<0.05 was considered statistically significant.

Results

DDIT4 is expressed in healthy chief cells and early in paligenosis but lost in chronic metaplasia and dysplasia

We examined DDIT4 expression in a human tissue microarray representing 218 patients with normal and precancerous lesions, as well as a collection of >50 biopsy and resection specimens harboring regions of transition from normal to SPEM (Table S1). Most gastric adenocarcinomas are thought to arise via a defined sequence of lesions with normal stomach undergoing first chronic atrophy with SPEM, progressing to intestinal metaplasia and finally to dysplasia^{22, 23}. DDIT4 expression progressively declined with each step (Figure 1A, B).

Figure 1. DDIT4 is expressed in human chief cells with expression decreasing in proliferative or dysplastic lesions.

A. DDIT4 expression in human stomach mucosa tumor microarrays. Histological scores were: 0, no staining; 1, staining in ≤20% of cells; 2, moderate to strong staining in≥20% of cells; 3, strong staining in ≥50% of cells. Scale bar, 30 μm; boxed area insert, 15 μm.

B. Average histological score (upper plot) of DDIT4 expression scores in human stomach tissue microarrays of only regions with normal histology and precancerous lesions. Lower plot shows fractional distribution of each score within each lesion phenotype.

C. SPEM lesions categorized as Ki-67 positive (“proliferative”) and Ki-67 negative (“quiescent”) were examined in serial tissue sections for DDIT4 and phosphorylated-S6 (correlating with mTORC1 activity). Depicted are representative images of lesions with proliferative SPEM or quiescent SPEM in serial slides of a single core from the tissue microarray. Scale bar, 30 μm; boxed area, 15 μm.

D. Upper plot shows average DDIT4 score (0= no expression, 3=maximal) for the lesion type, and lower plot shows fraction of biopsy with a given DDIT score for each lesion.

E. Human gastric resection specimen in a region of normal chief cells, SPEM, and cells with hybrid, transitional phenotype. Green=GSII; Red=PGC; Blue=DAPI; Purple=DDIT4. Fluorescence channels are separated in inserts to show co-labeling patterns. Scale bar, 30 μm; boxed area insert, 15 μm.

Chronic metaplasia increases risk for tumorigenesis^{13, 24, 25}. In humans, most SPEM lesions are mitotically quiescent at any one time^{4, 26, 27}. Figure 1C and D show that quiescent SPEM resembled normal tissue in having higher DDIT4 expression, whereas proliferating SPEM had lower DDIT4. As expected, high DDIT4 in quiescent SPEM correlated with low mTORC1, whereas low DDIT4 in proliferating SPEM correlated with high mTORC1⁴. Normal chief cells in human stomach had abundant DDIT4, but in regions where cells had transitional chief cell-SPEM cell morphology (see references^{4, 28}) DDIT4 was decreased (Figure 1E).

Given that DDIT4 suppresses mTORC1^{29, 30}, and mTORC1 reactivation dictates which cells will proliferate during paligenosis and metaplasia¹⁴, we hypothesized we could use DDIT4 as a tool to dissect how metaplasia is regulated. We used high-dose tamoxifen (TAM), which causes SPEM and DNA damage across the whole gastric body of mice within 3 days, with resolution normally by 14 days^{20, 31}, to clarify a role for DDIT4 in progression of metaplasia to dysplasia/neoplasia. DDIT4 expression in the body of the stomach was confined to chief cells. During Stage 1 of paligenosis, cells undergo dramatic loss of mTORC1 and massive autophagy to downscale differentiated cell architecture, whereas Stages 2 and 3 of paligenosis are characterized by cells expressing metaplastic/progenitor markers and then re-inducing mTORC1 to enter the cell cycle (Figure S1). During TAM-induced paligenosis, DDIT4 increased somewhat in Stage 1 of paligenosis, as mTORC1 was suppressed, and then was largely lost from the stomach by Stages 2 and 3 when mTORC1 reactivated (Figure 2A, Figure S1).

Figure 2. DDIT4 is expressed early in chief cell paligenosis is required for mTORC1 decrease, degradation of secretory cell architecture, and to moderate paligenotic proliferation.

A. DDIT4 expression in stomach chief cells as they undergo paligenosis to SPEM after TAM. Scale bar, 30 μm; boxed area insert, 15 μm.

B. Immunofluorescence for mature chief cell secretory granule protein GIF and a proxy for mTORC1 activity S6 phosphorylated one residues 240/244 in Ddit4^+/+ and Ddit4^−/− mice in stomach glands at various time points after TAM. Green=GIF; Red=pS6; Blue=DAPI. Scale bar, 30 μm; boxed area insert, 15 μm.

C. Transmission electron microscopy of single chief cells (outlined in red) from wild type and Ddit4^−/− mice 24h post TAM (late Stage 1). Yellow outlines indicate autodegradative (autophagic/lysosomal) compartments engulfing organelles in reprogramming chief cells. Scale bar, 1 μm.

D. Immunofluorescence for GIF and autophagosome marker LC3B 24h after tamoxifen injury (late Stage 1). Yellow outlines single paligenotic chief cells. Green=LC3B; Red=GIF; Blue=DAPI. Scale bar, 30 μm.

E. H&E staining of Ddit4^+/+ and Ddit4^−/− mice 72h post TAM (Stage 3). Inserts highlight base of gland where chief cells have become cuboidocolumnar SPEM cells in controls but retain substantial cytoplasm in the absence of DDIT4. Scale bar, 30 μm; boxed area insert, 15 μm.

F. Quantification of gland dropout for data as in panel (E). Each colored datapoint represents mean of all counts from a single mouse (>11 microscopic fields counted per mouse); each gray datapoint = % dropout in a single field; black line indicates mean from all mice, with significance estimated by unpaired, two-tailed t test.

G. Immunofluorescence for GIF, GSII (progenitor and, when co-expressed with GIF, SPEM cell marker) and BrdU (proliferation marker) 72h post TAM. White=BrdU; Green=GIF; Red= GSII; Blue=DAPI. Scale bar, 30 μm; boxed area insert, 15 μm.

H. Quantification of BrdU⁺GIF⁺GSII⁺ cells (proliferating SPEM cells). Each colored datapoint represents mean of >40 gastric units counted from a single mouse; each gray datapoint = BrdU+ cells in a single unit; black line indicates mean from all mice, with significance estimated by unpaired, two-tailed t test.

DDIT4 is required for the mTORC1 suppression early in paligenosis that allows only healthy cells to later reactivate mTORC1

We examined the requirement for DDIT4 in our TAM model (Figure 2A). In homeostatic conditions, Ddit4^–/– mice had a phenotype confined to chief cells with modest, but statistically significant, reduction in secretory granules (Figure S2A–E). Ddit4^–/– mice showed much more dramatic defects during paligenosis. Paligenotic cells did not turn off mTORC1 in Stage 1 (Figure 2B; note that phospho-S6 is a proxy for mTORC1 activity^{4, 14}). Lack of mTORC1 suppression correlated, as expected, with defective autophagy induction (Figure 2C, D). The defective autophagy led to preserved mature chief cell phenotype later on at Stage 3 (note retention of the mature chief cell granule marker GIF, Figure 2B, G). Stage 3 in Ddit4^−/− mice also differed as proliferation was increased despite the persistent differentiated cell phenotype. Additionally, TAM injury normally causes about 15% of gland bases harboring chief cells to dropout in wildtype mice (Figure 2E, F), but Ddit4^–/– mice had nearly no detectable gland loss. The loss of cells in control mice is due to p53-mediated apoptosis that occurs in Stage 2 to 3 transition as cells attempt to reactivate mTORC1 to drive proliferation¹⁴. Thus, the decreased cell loss and increased proliferation in Ddit4^–/– mice were consistent with inappropriate entry into Stage 3 due to lack of initial suppression of mTORC1 in Stage 1 (Figure 2G, H).

We next determined the consequences of Ddit4^–/– cells’ bypassing the licensing checkpoint for Stage 3 entry (Figure 3A). TAM causes DNA damage²⁰, but cells with DNA damage executing the fully functioning paligenosis program would be expected to first repair the damage before entering S-phase, or they would undergo apoptosis^{32, 33}. As anticipated, in Ddit4^+/+ mice BrdU⁺ cells (ie those that had entered S-phase) were largely negative for the DNA damage-repair marker γH2AX. However, in Ddit4^−/− mice, both the overall number of γH2AX⁺/BrdU⁺ cells and the fraction of γH2AX⁺ cells that were also BrdU+ were substantially and statistically increased (Figure 3B, C). The increase in both mitotic cells and in mitotic DNA-damage-harboring cells was mTORC1-dependent, as rapamycin blocked both phenotypes (Figure 3D, E). Overall, the results indicate that aberrant cell cycle entry in DDIT4-deficient cells is due to defective mTORC1 suppression and thus failed licensing in the Stage 2 to 3 transition.

Figure 3. Loss of DDIT4 increases entry of paligenotic cells into the cell cycle despite DNA damage, which is rescued by mTORC1 inhibition by rapamycin.

A. Immunofluorescence for BrdU and DNA damage marker γ-H2AX. Green, BrdU; Red, γ-H2AX; Blue, DAPI. BrdU⁺γ-H2AX⁻ cells (ie cells with DNA damage that have not entered into cell cycle) in one full gastric unit are marked with white arrows. Yellow dashed line outlines a representative gastric gland. Scale bar, 30 μm.

B. Data as in (A) quantified, focusing on BrdU⁺γ-H2AX⁺ cells 72 hours post TAM Colored datapoint: mean of ≥ 28 glands quantified per each mouse; gray datapoint: count from each gland; dark black lines = mean ± SEM across all mice; significance estimated by unpaired, two-tailed t test.

C. Data as in (A) quantified, focusing on the ratio of BrdU⁺γ-H2AX⁺ cells to H2AX⁺ cells (ie fraction of cycling cells with DNA damage to total cells with DNA damage) per gastric gland. Datapoints and statistics as in (B)

D. Immunofluorescence for BrdU and γ-H2AX staining of Ddit4^+/+ and Ddit4^−/− mouse stomachs treated with rapamycin. Green, BrdU; Red, γ-H2AX; Blue, DAPI. BrdU⁺γ-H2AX⁻ cells in one full gastric unit (outlined in yellow) are marked with white arrows; note outlined Ddit4^−/− unit had no BrdU⁺γ-H2AX⁻ cells, meaning all dividing cells had DNA damage. Scale bar, 30 μm.

E. Data as in (D), quantified for BrdU⁺γ-H2AX⁺ cells per gastric gland. Colored datapoint: mean of ≥ 42 glands quantified per each mouse; gray datapoint: count from each gland; dark black lines = mean ± SEM across all mice; significance estimated by one-way ANOVA with Tukey post hoc test.

Loss of DDIT4 causes cell-autonomous increased growth of human gastric cancer cells despite chemotherapy and DNA damage in vitro and in xenografts

To explore the cell-autonomous effects of DDIT4 on cell cycle entry and in the context of tumorigenesis, we returned to analysis of human patients and cells. We examined another annotated tissue microarray on which we had compiled 509 patients with gastric cancer. We saw statistically significant correlation of high DDIT4 expression with increased patient survival and more differentiated tumors, especially in patients who had received chemotherapy, consistent with DDIT4 slowing proliferation in response to DNA damage (Figure 4A–C; Table S3). To determine if loss of DDIT4 causes cell-autonomous effects on proliferation, we deleted DDIT4 from human MGC803 gastric cancer cells and treated them with the DNA damage-inducing, chemotherapeutic agent 5-FU (Figure S3A, B). 5-FU in control cells increased DDIT4 expression without initial effect on mTORC1 activity (Figure 4D). Contrastingly, DDIT4^−/− cells had elevated mTORC1 at baseline that persisted even after 5-FU (Figure 4D); they continued to enter G2 and M-phases of the cell cycle and proliferate, whereas control cells were largely arrested (Figure 4E; Figure S3C, D).

Figure 4. In humans, loss of DDIT4 correlates with worse gastric cancer survival and causes cell-autonomous increased mTORC1 and proliferation despite chemotherapy.

A. DDIT4 expression in representative cores scored as “negative” or “positive” from a human gastric cancer tissue microarray. Scale bar, 30 μm.

B. Overall survival determined via Kaplan-Meier curves separated by DDIT4 expression, with an event defined as a cancer-related death. The log-rank test was used to assess differences between the survival curves in different patient groups.

C. Overall survival determined using Kaplan-Meier curves in chemotherapy subgroups.

D. Western blot of DDIT4, pS6 Ser240/244 (proxy for mTORC1 activity) with β-tubulin loading control in CRIPSR-generated DDIT4 knockout (gRNA3-transfected cell line had the most effective knockout and was the one used for all experiments; see also Figure S3B) DDIT4^−/−) and negative control (DDIT4^+/+) MGC-803 gastric cancer cell lines at various time points after 5-FU treatment..

E. DDIT4^−/− and DDIT4^+/+ MGC-803 cells were treated with 5-FU, and proliferation measured via CCK-8 assay. Datapoint: mean ± SEM from 3 independent experiments with significance estimated by unpaired, two-tailed t test of the area under the curve (to avoid repeated measurement bias) of each individual experiment up until 24h and 48h.

F. GSEA of global mRNA expression profiling using all gene sets in Broad “Hallmark” compilation. All statistically significant gene sets (false discovery rate [FDR] and p value < 0.001) are shown with normalized enrichment score (NES)= 2.0 indicated by dashed red line.

G. DDIT4^−/− and DDIT4^+/+ cells were xenografted into the flanks of nude BALB/c mice and the mice treated with 5-FU according to the scheme at top. The growth of DDIT4^−/− tumors (n=8) was less inhibited by 5-FU than those from DDIT4^+/+ cells (n=8).

H. Tumor weight from (G) quantified. Each datapoint is a single tumor with mean of all tumors ±SD indicated and significance estimated by unpaired, two-tailed t test.

I. Tumor volume from (G) quantified. Tumor length (L) and width (W) were measured with calipers and volume by (L × W²) π/6. Statistics and plotting as for (H).

To identify pathways governed by DDIT4, we did a transcriptomic analysis of DDIT4^+/+ and DDIT4^−/− cells treated with 5-FU at 24h. An unbiased gene set enrichment analysis (GSEA) revealed increased proliferation-associated transcripts (mitotic spindle, G2M checkpoint, KRAS signaling), mTORC1 and biosynthetic/energetic transcripts (protein secretion, mTORC1 signaling, glycolysis), and transcripts associated with the known role of DDIT4 in regulating hypoxia and the role we are elucidating here that it may play in apoptosis of cells with DNA damage (Figure 4F). Unbiased analysis for Gene Ontology terms enriched in DDIT4^−/− cells treated with 5-FU showed similar enrichment in proliferation-associated and bioenergetics-associated processes, molecular functions, and cellular compartments (Figure S3E). The increased proliferation of DDIT4^−/− cells persisted even when cells were xenografted into the flanks of nude BALB/c mice that were then administered 5-FU (Figure 4G–I). As expected, both in vitro and in vivo, the effects of DDIT4 were mTORC1-dependent, as rapamycin normalized DDIT4^−/− pS6 levels and cell proliferation rate (Figure 5A–F).

Figure 5. Rapamycin rescues the phenotype of human DDIT4^−/− gastric cancer cells indicating effects of DDIT4 loss on proliferation are via mTORC1.

A. DDIT4^−/− and DDIT4^+/+ MGC-803 gastric cancer cell lines were treated with 5-FU or 5-FU with rapamycin, and proliferation measured by CCK-8 assay. Datapoint: mean ± SEM of 3 independent experiments with significance estimated by unpaired, two-tailed t test of the area under the curve (to avoid repeated measurement bias) of each individual experiment up until 24h and 48h.

B. Confocal immunofluorescence for Ki-67 (proliferation) and γ-H2AX (DNA damage) of MGC-803 cells treated with 5-FU or 5-FU with Rapamycin. Green, γ-H2AX; Red, Ki67; Blue, DAPI. Scale bar, 5μm.

C. Quantification of Ki67⁺γ-H2AX⁺ cells (ie proliferating cells with DNA damage) per field in (B). Data point: mean ± SD from 10 independent fields with significance estimated by one-way ANOVA with Tukey post hoc test.

D. DDIT4^−/− and DDIT4^+/+ cells were xenografted into the flanks of nude BALB/c mice which were then treated with 5-FU or 5-FU with Rapamycin (see Figure 4G for treatment scheme). n=8 for each group.

E. Tumor weight in (D) quantified. Each datapoint represents a single tumor with mean ± SD indicated and significance estimated by one-way ANOVA with Tukey post hoc test.

F. Quantification of (D) for tumor volume estimated from tumor length (L) and width (W) as: (L × W²) π/6. Statistics and plotting as for (E).

Loss of DDIT4 increases risk for metaplasia advancing to invasive gastric adenocarcinomas

Finally, we tested the functional consequences of unlicensed paligenosis for tumorigenesis. SPEM is a precursor lesion for gastric cancer in humans, but it does not efficiently progress to tumors in mice. The rate of spontaneous tumor formation can be increased by mutagens like N-methyl-N-nitrosourea (MNU)^{34, 35}, which, we reasoned, would afford us a route to determine if lack of paligenosis licensing in the absence of DDIT4 increased risk for tumorigenesis (Figure 6A). In pilot experiments (not shown), we determined that treating mice both with multiple rounds of metaplasia-inducing TAM and MNU caused dysplastic, polypoid gastric lesions within 36 weeks, earlier than reported for MNU alone in C57/B6 mice. We next performed three independent experiments treating Ddit4^−/− and control mice with cycles of TAM and MNU (experiment #1: n=10 mice per genotype; #2: n= 15 mice; #3: n= 15 mice, Figure 6A; Figure S5A, B). Overall, 24 of 30 surviving Ddit4^−/− mice developed grossly-identifiable gastric tumors vs. 16 of 31 surviving controls (Figure 6B; Figure S5A, B). Wildtype tumors were mostly flat, raised lesions with focal cellular and glandular atypia; they were located in the corpus or overlapped corpus and antrum (Figure 6C). There were no truly invasive carcinomas in Ddit4^+/+ mice (Figure S6A).

Figure 6. Loss of DDIT4 increases spontaneous tumorigenesis after multiple rounds of paligenosis in a mouse model.

A. Scheme for TAM and MNU administration (top) and example tumors (bottom) in stomachs from 40 mice per genotype (30 Ddit4^−/−, 31 Ddit4^+/+ mice survived until endpoint of observation) in three independent cohorts. First number is cohort #, second is individual mouse number from that cohort. Tumor area outlined by red dashed line.

B. Total area per mouse occupied by tumors as measured under dissecting microscope from stomachs opened as for (B) quantified from all mice surviving until the endpoint of observation. Each datapoint: area of single mouse tumor area, log₂ scale; significance estimated by Mann-Whitney test.

C. Representative H&E and PAS (for mucins) of one tumor from each mouse genotype. Tumors in wildtype mice were all raised, hyperplastic lesions with focal cellular and glandular atypia but no invasive features (red and green boxes show increased magnification). In the Ddit4^−/− tumor depicted (1 of 6 invasive tumors, #1–7), a massive, poorly differentiated tumor with both irregular glandular and signet ring (PAS-staining positive) morphologies is seen invading through muscle to serosa. Invasive signet ring cells are marked by yellow arrows. Scale bar, 500 μm; green arrows = perineural infiltration, scale bar 50 μm; red box shows lymph-vascular space invasion, 25 μm.

Ddit4^−/− lesions had similar location and morphology, but the size distribution was significantly shifted towards larger lesions (Figure 6A, B; Figure S5A, B). Moreover, 6 mice had invasive lesions, including poorly differentiated adenocarcinoma with “signet-ring” mucinous cells traveling alone or in nests and extending through the serosal surface, into duodenum and pancreas and penetrating lymph-vascular space (Figure S6A, B). The others were broadly invasive through muscle layers as irregular glands and as nests of cells (eg, see Figure S6A, B). Tumors in Ddit4^+/+ mice were distributed similarly to those of wildtype mice and similar to previous reports³⁶. The large and/or invasive tumors in the Ddit4^−/− mice were found in a zone of transition between the corpus and antrum (eg note parietal cells present in glands at the tumor margin, Figure S6A, B) consistent with loss of DDIT4 specifically in chief cells (which do not occur in mouse antrum) increasing the risk of large, invasive tumors after multiple rounds of TAM-induced paligenosis.

Discussion

Here, we show that DDIT4 suppresses mTORC1 to license mature cells to enter the cell cycle during the gastric metaplastic response. Absence of DDIT4 leads to inappropriately elevated mTORC1 and aberrant, increased cell cycle re-entry of chief cells harboring DNA mutations. The phenotypes caused by loss of DDIT4 culminate in increased tumorigenesis. Our experiments in human tissue and in human gastric cancer cell lines in vitro and in xenografts corroborate our mouse results. DDIT4 is expressed specifically in differentiated chief cells, and DDIT4 phenotypes are limited to differentiated chief cells before and after injury. Moreover, our experiments with gastric cell lines indicate DDIT4 has cell-autonomous effects on mTORC1 and proliferation. Thus, we highlight a role for mature cells as cells of origin for metaplasia and a mechanism by which metaplasia arising from those cells carries risk for progression to cancer.

DDIT4 seems to have an evolutionarily conserved role in licensing during injury-induced proliferation. In Drosophila, for example, DDIT4 orthologues suppress TOR to prevent stress-induced recruitment of proliferative cells³⁷. DDIT4 also plays a role in governing whether keratinocytes are activated in wound healing and in whether quiescent stem cells are recruited to become actively proliferating hematopoietic stem cells^{38, 39}.

Dynamic mTORC1 regulation governs injury-induced recruitment of mature cells into the cell cycle (ie paligenosis) with cells necessarily turning mTORC1 off early and then reactivating it later (see model in Figure 7, Supplementary Figure 1). This mTORC1 pattern features in paligenosis-like regeneration programs such as: neuronal axon regrowth^{14, 40}, liver and kidney repair⁴, satellite stem cell reactivation in muscle⁴¹. A critical role for mTORC1 in regeneration has also been shown across a broad range of species from hydra and sea anemone to humans⁴². We have shown that the mTORC1 fluctuation occurs because in differentiated cells, elevated mTORC1 maintains protein synthesis, which is needed in chief cells, for example, to secrete digestive enzymes. After paligenosis-inducing injury, DDIT4 is activated to suppress mTORC1 and shut off protein synthesis as well as to induce autophagy to recycle the elaborate rER and secretory granules into building blocks for subsequent cell division. The mTORC1 reactivation that later drives that cell division doesn’t happen until both DDIT4 is later degraded, and the p53 activated by injury is suppressed by IFRD1. We have shown recently that cells lacking IFRD1 fail mTORC1 reactivation and tend to undergo apoptosis instead of completing paligenosis¹⁴. Our results showing DDIT4 expression correlating with less proliferation in precancerous lesions and better survival of patients with gastric cancer confirm and extend on those of a previous report⁴³. Similarly, in other tissues it has been shown that DDIT4 safeguards against cells with DNA damage entering mitosis by mediating DNA repair and inducing apoptosis^{44, 45}. Additionally, dysregulated mTORC1 activity (eg as caused by loss of DDIT4) can cause genome instability in turn causing increased DNA damage⁴⁶. Overall, thus, it seems constitutive mTORC1 may help tumor cells divide in the face of acute chemotherapeutic stress.

Figure 7. A model for how DDIT4 causes failure of mTORC1 reactivation checkpoint and increased tumorigenesis.

Schematic summarizing the molecular-genetic dynamics corresponding to the cellular changes over the timecourse of paligenosis with relative expression of each marker or phenotype in Ddit4^+/+ (blue) vs Ddit4^−/− (green) mice.

Paradoxically, however, studies across multiple organs have shown disparate results for DDIT4 in tumorigenesis with some studies showing higher DDIT4 decreases cancer survival^{44, 47–50}. The discrepancies may arise because mTORC1’s role in tumorigenesis and in tumor cells is complex, and constitutive loss of DDIT4, one of the most important mechanisms for stress-induced mTORC1 suppression, may handicap cells’ ability to regulate their metabolic activity to adapt to different conditions dynamically (eg chemotherapy). Even in the current study, we see a statistically significant but not dramatic difference in patient survival, based on DDIT4 expression, suggesting constitutively increased, dysregulated mTORC1 may not be uniformly advantageous for the long-term survival and expansion of tumor cells.

The increased tumorigenesis in the stomachs of MNU-treated Ddit4^−/− mice is intriguing, because, to our knowledge, this is the first documentation that a specific aberration in the cellular program leading to metaplasia increases incidence of progression of metaplasia to tumors. There have been prior studies that have shown genes involved in progression of metaplasia to dysplasia and neoplasia; for example, Goldenring and colleagues have studied how TROP2 increases specifically in human lesions progressing between metaplasia to dysplasia⁵¹, which could eventually be developed as a tool or biomarker to follow metaplastic progression. But our studies here suggest a requirement for DDIT4 specifically in decreasing risk for subsequent tumorigenesis. And we note that in the Ddit4^−/− mice, the 6 invasive tumors that were induced were heterogeneous in phenotype from a signet-ring-type diffuse carcinoma to various adenocarcinomas, invading through the musculature and exhibiting more glandular features. Thus, in this relatively small sample size, the MNU Ddit4^−/− tumors resembled the invasiveness and morphological diversity that occur in human gastric cancer, which is not usually the case in mouse models of gastric cancer²⁵.

Our results are consistent with the “Cyclical Hit Model” of tumorigenesis^{13, 19} in which mature cells undergoing paligenosis are uniquely situated to accumulate somatic mutations that put an organ at risk for tumorigenesis. Active stem cells, may not typically accumulate mutations because they proliferate rapidly, frequently generating differentiated progeny that are eventually shed from the tissue⁵², thereby ridding the tissue of mutations. On the other hand, most mature cells that undergo paligenosis to repair damage subsequently return, eventually, to being long-lived, functional cells that can accumulate mutations⁵³. Any tumor-promoting mutations might eventually be unmasked in subsequent paligenosis events months or years later.

There is a limitation to the MNU plus TAM experiment in that we cannot confirm that chief cells were the cells of origin for the tumors, though we favor that as an interpretation. DDIT4 expression was restricted to chief cells (homeostatic and paligenotic), and the phenotype of DDIT4 loss in paligenosis was thus confined to chief cells. Moreover, in gastric cancer cells, loss of DDIT4 conferred increased tumor growth despite DNA damage, indicating DDIT4 works cell-autonomously. Additionally, the cells harboring DNA damage after TAM and loss of DDIT4 were the paligenotic chief cells. In the MNU experiments, the multiple rounds of paligenosis would put those cells at greatly increased risk for MNU-induced mutation (Figure 7). The location of the hyperplastic, non-invasive, raised lesions that appeared as tumors were relatively scattered throughout the stomach, but the invasive tumors were located in the transition between antrum and corpus, as many human tumors are¹³.

The loss of DDIT4 in later stages of paligenosis is not unexpected, as DDIT4 is controlled largely by protein turnover^{54, 55}. The early inducing event that activates DDIT4 isn’t known, but ongoing work suggests the culprit may be reactive oxygen species (ROS), which are early and important upstream stressors in gastric paligenosis⁵⁶ and are known to activate DDIT4.

Paligenosis likely evolved so multicellular organisms could recruit differentiated cells – the vast majority of cells in an adult – as regenerative progenitors. However, allowing long-lived cells to cyclically proliferate and then redifferentiate is dangerous, because such cells can accumulate mutations over the lifetime of the organism. Paligenosis thus must also involve strict licensing to ensure that damaged cells are not allowed to propagate. DDIT4 regulation of mTORC1 appears to be a central regulator of paligenosis that may help explain part of the mechanism of how cells progress from metaplasia to cancer.

Abbreviations:

DDIT4

DNA Damage Inducible Transcript 4

TAM

high-dose tamoxifen

SPEM

Spasmolytic polypeptide-expressing metaplasia

MNU

mutagen N-methyl-N-nitrosourea

5-FU

5-Fluorouracil