Excess growth hormone suppresses DNA damage repair in epithelial cells

Growth hormone (GH) secreted by the pituitary gland acts through insulin-like growth factor–1 (IGF-1) to regulate tissue growth and metabolism (1–4). GH also acts independently of IGF-1 in regulating bone, muscle, and liver tissue functions (5–8). In adults, GH is only approved as a physiologic replacement for pituitary-deficient individuals or to improve AIDS-associated muscle wasting (9). The age associated somatopause characterized by physiologic decline of circulating GH occurs due to diminished pituitary GH production (10, 11). GH replacement in adults with proven GH deficiency results in loss of abdominal fat, increased muscle mass and energy, increased bone mineral density, and improved exercise capacity (12–14). Based on these GH effects reported in pituitary-deficient adults, recombinant human GH and GH-related products have been aggressively promoted in an attempt to reverse age-associated frailty in pituitary-replete healthy adults (15) and are also widely used to improperly enhance athletic performance (16). However, inappropriate treatment of otherwise healthy adults with GH produces modest changes in body composition with significant side effects, including arthritis, edema, insulin resistance, and heart disease (12, 16, 17). Furthermore, evidence from animal and human studies support a direct role for GH in the progression of neoplastic epithelial growth (18–20). Excess pituitary tumor GH secretion in acromegaly results in soft tissue overgrowth and increased adenoma formation in the colon, skin, thyroid, and prostate (21), and it is associated with increased risk for colorectal carcinoma (22, 23). GH triggers epithelial-to-mesenchymal transition, creating a proneoplastic mucosal environment (24–27). By contrast, abrogated GH signaling is associated with decreased cancer development in humans and in mice (20, 28–30). For example, short-stature humans harboring an inactivating mutation in the GH receptor (GHR) do not develop cancer (28), and GHR knockdown in human melanoma cells attenuates tumor progression (31).

In view of widespread GH abuse, it is important to further examine mechanisms for protumorigenic GH actions. We recently showed that GH is induced in nonpituitary human breast and colon cells in response to activation of the p53/p21 pathway (32), which plays an important role in DNA damage response (DDR). Since we also demonstrated that GH, in turn, suppresses colon p53/p21 (24), we sought to determine whether GH regulates DDR.

Oncogenic mutations may be associated with inefficient repair of damaged DNA, and surveillance mechanisms detect DNA damage and exert DNA repair to maintain genomic stability. The Ser/Thr protein kinase ataxia telangiectasia mutated (ATM) is a key component of signaling pathways activated by DNA damage (33). ATM is activated in response to DNA strand breaks by Ser¹⁹⁸¹ autophosphorylation (34). Activated ATM triggers DDR (33), phosphorylates checkpoint kinase 2 (Chk2) to arrest cell proliferation (35), and phosphorylates p53 at Ser¹⁵, resulting in its activation and stabilization (36). In turn, activated p53 drives cell cycle arrest, as well as promotion of DNA repair and programmed cell death (37). γ-Histone 2A variant (γH2AX) formed by ATM-induced Ser¹³⁹ phosphorylation in response to double- or single-strand DNA breaks (33, 38, 39) is required for DNA repair protein assembly, as well as for activation of cell cycle checkpoint proteins (37, 40).

We show here that GH decreases ATM kinase activity and suppresses ATM autophosphorylation, with subsequently decreased H2AX and p53 phosphorylation in both normal colon cells and in human 3-dimensional intestinal organoids. In in vitro and in vivo models, GH induced tripartite motif containing 29 (TRIM29) (41) and suppressed tat interacting protein 60 kDa (Tip60), leading to ATM destabilization (42). GH treatment led to increased unrepaired DNA both in vivo and in vitro, likely due to decreased nonhomologous end-joining (NHEJ), and enhanced anchorage-independent growth of normal colon cells. In athymic mice bearing GH-secreting human colon HCT116 xenografts, metastatic lesions were more prevalent, and the number of metastases correlates strongly with the levels of circulating GH. The results indicate that suppression of DNA repair by GH may guide colon cells toward transformation, and they highlight an adverse risk for inappropriate adult GH administration.

GH suppresses etoposide-induced DDR. To examine whether increased GH affects DDR, we employed topoisomerase II inhibitor etoposide to produce single- and double-stranded DNA breaks (43). We treated human nontumorous colon cells (hNCC) with 500 ng/ml GH for 6 hours and then induced DNA damage with 5 μM etoposide. After 24 hours, total ATM expression was seen to be unchanged, but levels of pATM were lower in cells treated with GH and etoposide, compared with cells treated with etoposide alone. Decreased ATM phosphorylation, in turn, resulted in decreased phosphorylation of its target protein p53. ATM also regulates DNA repair via Rad50 phosphorylation (44), and we found that levels of pRad50 were markedly lower after GH treatment (Figure 1A and Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.125762DS1). Together, these results suggest that GH suppresses DDR by decreasing ATM phosphorylation. Similar results were observed in nontumorous human mammary cells (MCF12A) and in colon adenocarcinoma HCT116 cells (Supplemental Figure 2, A and B).

Figure 1

GH suppresses etoposide-induced DNA damage response (DDR) and increases unrepaired DNA damage. In all experiments, hNCC were pretreated with 500 ng/ml GH for 6 h and then treated with 5 μM etoposide (Etop). (A) Western blot of hNCC harvested 24 hours after etoposide treatment. (B) Western blot of hNCC harvested 1 and 3 hours after etoposide treatment. Shown are representative results from 3–5 independent experiments. For A and B, quantification of protein expression is depicted in Supplemental Figure 1, A and B. (C) Aggregate measurements of γH2AX foci in hNCC harvested 24 hours after etoposide treatment. One dot represents measurements in a single nucleus; 20–30 nuclei per image and 5 images per group were analyzed. Differences were assessed by nonparametric Wilcoxon rank sum test.

At a later time point, 96 hours after etoposide treatment, we found comparable decreased pATM levels in GH-treated hNCC cells (Supplemental Figure 1D).

We also tested effects of etoposide in hNCC at earlier time points. In cells pretreated with GH for 6 hours and then treated with etoposide for 1 or 3 hours, we observed decreased ATM and p53 phosphorylation compared with cells treated solely with etoposide (Figure 1B and Supplemental Figure 1B). Chk2 phosphorylation was unchanged in response to GH 24 hours after etoposide treatment, but GH reduced Chk2 phosphorylation 3 hours after etoposide (Figure 1B and Supplemental Figure 1B), reflecting an earlier reduction in ATM phosphorylation.

Phosphorylation of H2AX is an early event in the double-strand break (DSB) response. Activated ATM phosphorylates H2AX, and γH2AX marks damaged DNA sites to activate DDR and promote DNA repair (45). We found decreased γH2AX in GH-treated hNCC cells 1, 3, 24, and 96 hours after etoposide (Figure 1, A and B, and Supplemental Figure 1, A, B, and D), with similar findings in GH-treated MCF12A and HCT116 cells 24 hours after etoposide (Supplemental Figure 2, A and B). Cotreating hNCC cells with etoposide and GH resulted in decreased aggregate measurements of γH2AX observed on immunocytochemistry slides, such as the number of γH2AX foci per nucleus and their intensity (Figure 1C and Supplemental Figure 3, A and B).

Decreased γH2Ax in GH-treated cells may result from rapid repair of DNA damage or from DDR not fully activated in response to etoposide. We interpreted our results as indicating that GH decreases ATM phosphorylation (activation), which in turn leads to decreased phosphorylation of H2AX, p53, and Chk2 target proteins. In support of this interpretation, we observed that endogenous ATM kinase activity measured in hNCC exposed to GH was reduced by approximately 40% in response to etoposide (Figure 2, A and B, and Supplemental Figure 1C), indicating that DDR was not fully activated, thus allowing for decreased DNA damage repair and accumulated DNA damage. Indeed, employing the comet assay to assess DNA damage in nontumorous colon hNCC and mammary MCF12A cells, we observed that cells treated with both etoposide and GH exhibit more unrepaired DNA damage vs. cells treated solely with etoposide (Figure 2C and Supplemental Figure 2C). By contrast, in cancerous HCT116 cells, the extent of DNA damage did not differ significantly between cells challenged with etoposide in the presence or absence of GH (Supplemental Figure 2C).

Figure 2

GH suppresses ATM kinase activity. (A) ATM kinase assay. hNCC were pretreated with GH and treated with etoposide for 3 hours. Cell extracts were immunoprecipitated with total ATM antibody. (B) Controls using IgG or in which the peptide was omitted were included. Assays were conducted in triplicate. Results shown are mean ± SEM of 3 independent experiments. Each dot represents 1 independent experiment. Data are graphed as fold-change, but statistical testing was performed on raw numbers. Differences were assessed with Tukey-adjusted mixed model regression. *P < 0.0 5 vs. IgG + Etop. (B) For ATM kinase assay, Western blotting was used to detect total ATM or autophosphorylated ATM (phospho-Ser 1981) and to verify equal protein amount in the immunoprecipitated samples for each experiment. Representative blots are shown. Quantification of protein expression is depicted in Supplemental Figure 1C. (C) Comet assay of hNCC harvested 24 hours after etoposide treatment. Single-cell gel electrophoresis was conducted and Olive Tail Moments assessed on at least 200 cells/per slide for each experiment. Results shown are mean ± SEM. Control, untreated cells. **P < 0.01 vs. control. Differences were assessed with Tukey-adjusted mixed model regression.

To elucidate mechanisms for DDR suppression by GH, we tested expression of proteins involved in ATM regulation. TRIM29 suppresses histone acetyltransferase Tip60 (46), which in turn acetylates ATM, inducing activation and autophosphorylation (42). Treatment of hNCC with etoposide or GH for 24 hours markedly enhanced TRIM29 expression, but addition of GH did not further increase high TRIM29 in etoposide-treated cells. By contrast, GH treatment decreased Tip60 expression in both control and etoposide-treated cells (Figure 3A and Supplemental Figure 4A). Comparable results were observed in HCT116 cells (Supplemental Figure 5), in which GH pretreatment increased TRIM29 expression and suppressed Tip60 in both control and etoposide-treated cells.

Figure 3

GH suppresses DDR in hNCC by inducing TRIM29 and suppressing Tip60. (A and B) hNCC were pretreated with 500 ng/ml GH and treated with 5 μM etoposide. Western blots of TRIM29 and Tip60 in hNCC harvested 24 hours (A) or 1 and 3 hours (B) after etoposide treatment. Shown are representative blots from at least 3 independent experiments. Quantification of protein expression is depicted in Supplemental Figure 4. (C and D) Three-dimensional intestinal organoids were pretreated with 500 ng/ml GH overnight, treated with etoposide for 24 hours, and harvested. Western blots of (C) TRIM29 and Tip60 and (D) DDR. Shown are representative blots from 3 independent experiments. Quantification of protein expression is depicted in Supplemental Figure 7. (E) Comet assay of organoids pretreated with 500 ng/ml GH, treated with 3 or 5 μM etoposide for 24 hours, and harvested. Results shown are mean ± SEM of 3 independent experiments. Differences were assessed with Tukey-adjusted mixed model regression. Control, untreated organoids. **P < 0.01 vs. control.

At earlier time points, at 1 and 3 hours after treatment, TRIM29 was markedly induced in cells treated with etoposide or GH only, but etoposide did not further induce TRIM29 in cells pretreated with GH (Figure 3B and Supplemental Figure 4B). Activated TRIM29 downregulated Tip60 in GH-treated cells (Figure 3B and Supplemental Figure 4B), which likely resulted in the observed decrease in ATM, H2AX, p53, and Chk2 phosphorylation in response to etoposide (Figure 1B). Thus, GH-induced TRIM29 and the resultant decreased Tip60 likely lead to decreased DDR activity.

A product of the multidrug resistance 1 (MDR1) gene protects cells from genotoxic effects of chemotherapy (47). We found that MDR1 was not changed in cells treated with GH or in cells overexpressing GH after etoposide treatment (Supplemental Figure 6), indicating that protective GH effects on DNA damaged cells are likely not mediated by GH-induced MDR1.

GH suppresses DDR in human intestinal organoids. We next examined effects of GH on DDR in human intestinal organoids by pretreating with GH overnight and then treating with etoposide for an additional 24 hours. While TRIM29 was induced by both etoposide and GH, Tip60 was suppressed by the addition of GH to etoposide (Figure 3C and Supplemental Figure 7A). Phosphorylation of ATM, H2AX, and p53 were markedly reduced in organoids treated with both GH and etoposide compared with organoids treated with etoposide only (Figure 3D and Supplemental Figure 7B).

We then tested the extent of DNA damage in organoids treated with 3 μM and 5 μM etoposide in the presence or absence of GH. Similar to its effects in hNCC, GH exacerbated DNA damage caused by both doses of etoposide treatment (Figure 3E). Levels of phosphorylated ATM, Rad50, p53, and H2AX were also accordingly decreased in organoids treated with GH (Supplemental Figure 8, A and B).

GH suppresses endogenous DDR and increases unrepaired DNA damage. After finding that GH suppressed etoposide-induced ATM kinase phosphorylation and prevented complete DDR activation, we tested whether GH also acts to alter baseline DDR. In hNCC treated with 500 ng/ml GH for 24 hours, GH induced TRIM29 expression while suppressing Tip60. Downregulation of Tip60, in turn, likely resulted in suppressed ATM phosphorylation, as we observed subsequent decreased phosphorylated H2AX, p53, and Rad50 in cells treated with GH (Figure 4A and Supplemental Figure 9A). Comet assay showed an increase of approximately 50% in baseline DNA damage (Figure 4B), indicating that GH also suppresses endogenous DDR and may lead to increased unrepaired DNA damage.

Figure 4

GH suppresses endogenous DDR via TRIM29/Tip60 pathway in vitro. (A) Western blot analysis of DDR and (B) comet assay in hNCC treated with 500 ng/ml GH for 24 hours. Results shown are mean ± SEM of 3 independent experiments. Controls represent untreated cells. Differences were assessed with Tukey-adjusted mixed model regression. (C) Western blot of hNCC stably expressing short hairpin scramble (shScr) or shTRIM29 RNAi and treated with 500 ng/ml GH for 24 hours. Shown are representative blots of at least 3 independent experiments. For A and C, quantification of protein expression is depicted in Supplemental Figure 9.

To test whether GH suppresses ATM kinase phosphorylation at lower doses, we treated hNCC cells with 100 ng/ml GH and harvested cells 24 hours later. ATM phosphorylation was also markedly suppressed even at this modest GH dose (Supplemental Figure 9C).

GH suppresses DDR via TRIM29/Tip60. As high TRIM29 may reduce ATM kinase activity by suppressing Tip60 (42), we generated hNCC stably expressing shTRIM29. In scrambled shRNA transfectants, GH induced TRIM29 and suppressed Tip60, and it also reduced phosphorylation of ATM. When TRIM29 was downregulated, Tip60 was significantly induced, likely as a result of TRIM29 suppression and independent of GH treatment. Induction of Tip60, in turn, was associated with increased pATM as compared with cells stably expressing scrambled RNA and treated with GH (Figure 4C and Supplemental Figure 9B). These results suggest that GH action on Tip60 and pATM requires TRIM29, at least under baseline conditions.

High GH action on TRIM29/Tip60 and DNA damage in vivo. We next considered whether high GH levels would similarly impact the TRIM29/Tip60 pathway in vivo. Athymic nude mice were injected with HCT116 cells stably infected with lentivirus-expressing murine GH (lentiGH) or empty vector (lentiV). lentiGH xenografts recapitulate the systemic GH increase, as observed in acromegaly, and lead to increased circulating GH (Supplemental Figure 10A). We found induced colon TRIM29 and suppressed Tip60 in 6 of 7 lentiGH mice (Figure 5A and Supplemental Figure 10C). As exogenous DNA damage was not triggered, levels of pATM were difficult to detect; however, we still observed decreased phosphorylation of p53 and Rad50, likely resulting from downregulation of endogenous pATM (Figure 5A and Supplemental Figure 10C). In randomly selected animals carrying the xenograft, hepatic TRIM29 was similarly upregulated and Tip60 downregulated (Figure 5B and Supplemental Figure 10D).

Figure 5

GH suppresses endogenous DDR via the TRIM29/Tip60 pathway in vivo. (A–B) Athymic nude mice were injected s.c. with 500,000 HCT116 cells stably infected with lenti mGH (GH) or lenti vector (lentiV) and sacrificed 5 weeks after injection. Western blot analysis of individual colon tissues (A) and liver tissues (B) derived from mice bearing xenograft tumors. Quantification of protein expression is depicted in Supplemental Figure10, C and D. (C) Comet assay in colon tissue of mice bearing HCT116 lenti mGH xenograft tumors. Results shown are mean ± SEM. n = 6 mice/group. Controls represent mice bearing lentiV xenografts.

Thus, high levels of circulating GH results in downregulation of DDR. To assess the consequence of GH-induced DDR suppression, we examine DNA damage in the colon tissue of experimental mice. Detected by comet assay, animals with high-circulating GH demonstrate 61% ± 14.5% higher levels of unrepaired DNA damage as compared with controls (Figure 5C).

GH affects DDR through the GHR. To confirm that GH signaling activates TRIM29 and subsequently suppresses DDR, we pretreated hNCC with 20 mg/ml pegvisomant, a GHR antagonist (48), for 1 hour and then treated cells with 500 ng/ml GH and harvested them 24 hours later. In control cells, as expected, GH treatment induced TRIM29 with subsequent downregulated Tip60 and decreased phosphorylation of ATM and p53. By contrast, in cells treated with pegvisomant, blocking GH action led to downregulated TRIM29 and induction of Tip60, likely leading to the observed increased phosphorylated ATM and p53 (Figure 5A, Figure 6A, and Supplemental Figure 11A). These results indicate that GH, acting through the GHR, activates TRIM29 and suppresses DDR, leading to increased unrepaired DNA damage.

Figure 6

GH suppresses DDR via GHR. (A) Western blot of hNCC treated with 20 mg/ml pegvisomant (P) for 1 hour and then with 500 ng/ml GH for 24 hours; C, untreated control. This experiment was repeated 3 times with similar results; representative blots are shown. (B) Comet assay of hNCC stably expressing shGHR or scramble shRNAi (control). Results shown are mean ± SEM of 3 independent experiments. (C) Endogenous DNA damage in GHR^–/– and WT colon tissue. WT(control) and GHR^–/– mice were paired for analysis according to age and sex. Results shown are mean ± SEM (each dot represents 1 pair). For B and C, differences were assessed with Tukey-adjusted mixed model regression. (D) Western blot of hNCC stably expressing shGHR or scramble shRNAi and treated with 5 μM etoposide for 24 hours. Representative blots from 3 independent experiments are shown. For A and D, quantification of protein expression is depicted in Supplemental Figure 11, A and C.

In support of this observation, we generated hNCC stably expressing shGHR and found a 30% decrease in DNA damage as assessed by comet assay compared with controls (Figure 6B). Similarly, abrogated GH signaling in GHR^–/– mice resulted in decreased levels of unrepaired male colon DNA. In female GHR^–/– mice, the difference did not reach significance (Figure 6C). In hNCC where GHR was suppressed, etoposide treatment resulted in marked pATM upregulation (Figure 6D and Supplemental Figure 11, B and C).

GH attenuates DNA repair. As the DDR pathway is not fully activated in the presence of GH, we sought to examine whether GH affects NHEJ or homologous recombination (HR). Rapid DSB repair is mediated principally via NHEJ (49); HR occur with DNA damage at DNA replication forks (33, 50, 51). We measured rejoining of DSB repair reporters integrated into genomic DNA of hNCC (52) by cotransfecting these stable transfectants with plasmid encoding I-SceI endonuclease to induce DSBs, as well as with plasmid encoding DsRed to control for transfection efficiency; then we treated these transfectants with 500 ng/ml GH and assessed the results after 4 days. NHEJ efficiency was 35% ± 3.3% lower in cells treated with GH (P < 0.05) (Figure 7, A and B). We also observed a small but consistent decrease in HR efficiency (11% ± 2.04%, n = 6 independent experiments) 5 days after GH treatment.

Figure 7

GH attenuates endogenous NHEJ DNA repair by inhibiting DNA-PKcs phosphorylation. (A) hNCC containing a chromosomally integrated NHEJ reporter cassette were cotransfected with I-SceI and DsRed expression vectors and treated with 500 ng/ml GH overnight. The intact reporters are negative for GFP. Upon induction of a DSB by I-SceI digestion, the functional GFP gene is reconstituted. Cells were also transfected with pDsRed2-N1 as transfection control, and the percent of DsRed⁺ cells indicates transfection efficiency. Cells were analyzed by FACS on day 5 after GH treatment, and the relative efficiency of DNA DSB repair was calculated as the ratio of GFP⁺ cells/DsRed⁺ cells. Representative analysis is shown. (B) Graph shows fold change in GFP positivity ± SEM in 5 independent assays. Controls represent cells not treated with GH. Differences were assessed with Tukey-adjusted mixed model regression. (C) Western blot analysis of DNA-PKcs phosphorylation in hNCC treated with 500 ng/ml GH for 24 hours. Representative blots from 5 independent experiments are shown. Quantification of protein expression is depicted in Supplemental Figure 12.

As the catalytic subunit of DNA-dependent protein kinase (DNA-PKcs) is required for NHEJ repair, we examined effects of GH on this protein. Similar to ATM, DNA-PKcs is phosphorylated in response to DNA damage (33, 53). We found that GH suppressed phosphorylation of this protein in hNCC (Figure 7C and Supplemental Figure 12), implying that lower NHEJ efficiency in the presence of GH may be attributed to decreased DNA-PKcs activation.

GH action on proliferation, cell cycle, survival, and anchorage-independent growth. We tested GH action on hNCC proliferation when pretreated with GH for 6 hours and then treated with 5 μM etoposide. Seventy-two hours after the beginning of etoposide treatment, no difference was observed in cells incorporating BrdU among cells treated with etoposide only vs. both GH and etoposide (35.6% ± 7.2% vs. 33% ± 5.5%, respectively).

We examined GH effects on cell cycle progression. hNCC and HCT116 were plated in full medium, and when cells reached approximately 60% confluency, medium was changed to serum free with 0.1% BSA, and 500 ng/ml of GH was added. After a 6-hour incubation, cells were treated with 5 μM etoposide for 24 hours. Cells were then harvested, fixed in 70% ethanol, stained with propidium iodine, and analyzed by FACS in triplicate. In hNCC, GH treatment resulted in a modest 8% ± 2.7% increase in the number of cells in S-phase compared with untreated cells. In cells treated with GH and etoposide, the number of cells in S-phase was 20.7% ± 4.3% higher than in cells treated with etoposide only. These results indicate that GH reduced replication delay caused etoposide-induced DNA damage. No significant cell cycle changes were observed in HCT116 cells treated with GH.

We also tested whether GH modifies colon cell survival following etoposide treatment using clonogenic survival assays. Cells were pretreated with GH for 6 hours and treated with etoposide for 1, 3, or 24 hours; fresh GH was added every third day, and colonies were assessed 8 days after the end of etoposide treatment. Fewer colonies were detected in GH-pretreated cells compared with etoposide-only–treated cells (Figure 8A), indicating that, despite acute antiapoptotic effects of GH (54, 55), accumulation of DNA damage in GH-treated cells eventually may result in further cell death.

Figure 8

GH decreases survival but increases anchorage-independent growth and metastasis. (A) Number of colonies per well in hNCC pretreated with 500 ng/ml for 6 hours and treated with etoposide for 1, 3, and 24 hours and normalized to number of colonies in etoposide-only treated cells (control). The number of colonies was assessed 8 days later, and assays were performed in triplicate. Results shown are mean ± SEM of 3 independent experiments. Controls represent untreated cells. (B) Number of colonies formed by hNCC in soft agar treated as in A with 100 and 500 ng/ml GH. Results shown are mean ± SEM of duplicates in 4 independent experiments. In A and B, differences were assessed with Tukey-adjusted mixed model regression. Controls represent cells treated with etoposide only. (C–D) Athymic nude mice were injected with HCT116 lenti mGH (GH) or HCT116 lenti vector (vector) cells. Four weeks later, when mice develop ~0.5³ cm xenograft tumors, they received intrasplenic injections of 500,000 HCT116 cells and were sacrificed 4 weeks later. (C) Percent of mice that developed distant metastasis. (D) Number of metastases per mouse adjusted to the circulating GH levels. One-way ANCOVA, F = 4.66, degrees of freedom = 1.16, P = 0.045.

The ability to survive and grow in the absence of extracellular matrix anchorage correlates closely with tumorigenicity in animal models (56). We examined GH action on soft agar colony formation by pretreating hNCC with 2 different doses of GH (100 and 500 ng/ml) for 6 hours and then treating with etoposide for 1, 3, and 24 hours. Fresh GH was added every third day. In cells treated with etoposide for 3 and 24 hours, both concentrations of GH significantly increased the number of colonies compared with cells treated with etoposide only (Figure 8B), suggesting that GH enhances neoplastic transformation of normal colon cells that survive DNA damage.

GH effects on metastasis in vivo. To confirm these results in vivo, athymic nude mice were injected with HCT116 cells stably infected with lentiGH or lentiV. Four weeks after injection, when xenografted tumors reached ~0.5³ cm volume, we implanted intrasplenic HCT116 cells to recapitulate the metastatic process, as described (57, 58). Four weeks after intrasplenic injection, mice were sacrificed, and circulating GH was measured to confirm increased levels (Supplemental Figure 10B). Distant pleural, ovarian, and peritoneal metastases were detected in 7 of 9 mice with high-circulating GH, compared with metastases observed in 3 of 10 control animals bearing lentiV xenografts (Figure 8C). There was a strong overall correlation between metastasis count and GH levels (Pearson’s correlation = 0.4857, P = 0.0350). After adjusting for GH levels, the lentiGH group developed significantly more metastases than the lentiV group (Figure 8D).

In this study, we examined effects of GH in nontumorous colon epithelial cells and tissues, using etoposide as a tool to activate DDR. We show that GH acts to attenuate DDR by inhibiting ATM kinase activity; decreasing phosphorylation of key DDR effector proteins including Chk2, p53, and H2AX; and resulting in increased unrepaired DNA. We found that GH reduced NHEJ by more than 30%, thus disrupting normal DNA repair mechanisms.

Loss of genomic integrity due to DDR inactivation may enhance the risk of accumulating oncogenic mutations. Early-stage human colon tumors show elevated levels of DNA damage, which contributes to neoplastic cell growth and proliferation (59, 60). Our results suggest that normal colon cells with GH-induced unrepaired DNA damage are more likely to undergo neoplastic transformation, increasing anchorage-independent growth in vitro and metastasis development in vivo. This hypothesis is reinforced by our findings that the number of metastases correlates strongly with the levels of circulating GH.

Previous studies have linked the GH/IGF-1 axis with the DNA damage pathway. In prostate cancer cells lines, IGF-1 receptor (IGF-1R) influences DNA repair via DSB repair pathways, while IGF-1R depletion impairs ATM kinase activity in murine melanoma cells and enhances radiosensitivity in human prostate cancer cells (61, 62). GH treatment protects Chinese hamster ovary–4 cells from γ-irradiation (63), and administration of human GH shields normal rat intestinal tissue but not adenocarcinoma xenografts from radiotherapy-induced damage (64, 65). We show here that colon cells treated with GH exhibit suppressed DDR pathways and increased unrepaired DNA damage both at baseline and after stimulation by etoposide. Mice subjected to prolonged high-circulating GH due to GH-secreting xenografts showed ~60% more unrepaired DNA damage in colon tissue, potentially enabling a mucosal microenvironment permissive for transformation. Moreover, we also show that endogenous DNA damage is decreased in colon cells when GHR signaling is blocked. These results were confirmed in vivo in the colon of male GHR^–/– mice, consistent with previous observations showing that GH-deficient Snell dwarf and GHR-KO mice have increased expression of hepatic proteins associated with DNA repair (66). Thus, the effects of GH on DDR are mediated by GHR. Although the exact downstream pathway is yet unknown, it was shown that KRAS is required for GHR-mediated activation of p44/p42 MAPK (67) and that the KRAS/MAPK/ERK pathway is important for radiation resistance (68).

We found that high GH leads to increased unrepaired DNA damage. Some cells that accumulate DNA damage eventually die due to late triggering of apoptosis, as observed in colon cells overexpressing GH (32) and evidenced by reduced colony formation in cells treated with both etoposide and GH compared with etoposide alone. P53 phosphorylation and stabilization results in temporal proliferation block to allow cells to repair DNA damage (69, 70). We previously showed that GH suppresses total p53 (24). Here, we show that GH also reduces ATM activity and destabilizes p53, thus enabling some cells to evade p53-dependent senescence or apoptosis (71, 72) and to continue proliferating despite accumulation of unrepaired DNA damage. Our finding that GH increases the number of hNCC cells entering S phase after etoposide treatment suggests that, by destabilizing p53, GH unblocks replication delay induced by etoposide-activated DNA damage and supports our hypothesis that high GH, by suppressing DDR and DNA repair, allows damaged cells to replicate.

Either upregulated or downregulated/mutated DDR proteins are associated with metastasis (73). Thus, p53 downregulation or mutation induces epithelial mesenchymal transition, migration, and invasion and also inhibits p63, resulting in increased trafficking of β1-integrin, which is intimately involved in metastasis in human breast and prostate carcinomas (74, 75). Downregulation of Tip60 is associated with distant metastases in colon cancer (76) and melanoma (77), and Tip60 overexpression in melanoma cells reduces in vitro cell migration (77). We show here that GH acts to decrease both p53 phosphorylation, as well as Tip60 expression. GH-treated cells demonstrate increased capacity for anchorage-independent growth, consistent with our in vivo results showing increased metastatic potential of colon cells in the presence of high-circulating GH. Others have shown that high GH increased murine pulmonary melanoma metastases (78). These prometastatic effects may be attributed to both GH-mediated DDR suppression and EMT activation (24).

We previously found that in vitro GH treatment and overexpression did not affect IGF-1 or IGF-1R in colon cells (24). However, we cannot exclude that our observed in vivo effects of GH on DDR in colon may also be mediated by IGF-1 signaling, especially as IGF-1R also interacts with the GHR and may augment GH signaling (79). p53 activates IGF-1 binding protein 3 (IGFBP3), which suppresses IGF-1 function (80), and transcriptionally activates IGF-1R (81, 82). GH-induced p53 destabilization may suppress both IGFBP3 and IGF-1R, and decreased IGF-1R may attenuate DNA repair, as has been reported in prostate cancer cell lines (62).

GH action on DDR elucidated here may be cell type specific or specific to normal nontumorous cells with intact DNA damage pathways, as opposed to neoplastic cells harboring multiple mutations. In our studies, GH-induced DNA damage was observed in nontumorous colon and breast cells with intact p53 and DNA damage pathway signaling and in normal human organoids, while less profound effects were found in colon adenocarcinoma HCT116 cells that exhibit DNA repair alterations (83). Others have demonstrated ATM and H2AX phosphorylation in GH-treated malignant basal prostate cells, but only moderate effects on luminal cells (84), as well as decreased DNA damage and increased cell survival with GH overexpression in human endometrial and mammary carcinoma cells (85). Thus, GH effects may differ in normal and malignant tumor cells. In addition, protective effects of GH in DNA damaged cells may be also associated with clearing of chemotherapeutic drug. Although we show that GH does not affect MDR1 expression in hNCC, GH treatment of human melanoma cells induced expression of ATP-binding xenobiotic efflux pumps associated with melanoma drug resistance (86).

We elucidated mechanisms underlying GH effects on ATM. ATM can be regulated via different pathways. After DNA damage, the Mre11/Rad50/Nbs1 (MRN) DNA-binding complex is required for ATM activation (87), and spectrometry-based phosphoproteomics showed that GH rapidly dephosphorylates Mre11 and Rad50 in 3T3-F442A preadipocytes, potentially affecting activity of the MRN complex (88). TRIM29, induced by DNA damage, binds and degrades the histone acetyltransferase Tip60 (46), which also induces ATM kinase activity in response to DNA damage. Suppression of Tip60 blocks ATM kinase activity and prevents ATM-dependent phosphorylation of p53 and Chk2 (42), and Tip60 downregulation is linked to development of advanced colorectal carcinomas (76). We show here that GH markedly induces normal colon cell TRIM29, leading to decreased Tip60 expression, diminished ATM phosphorylation, and prevention of DDR from being fully activated in response to DNA damaging treatment. By contrast, suppression of baseline TRIM29 in hNCC resulted in Tip60 induction with subsequent ATM activation. Further confirming the importance of GH signaling for this pathway, pharmacological blockade of the GHR abolished TRIM29 and induced Tip60 expression, with induced phosphorylation of ATM and p53. Cellular responses to DNA damaging agents is complex. TRIM29 has several functions in tumorigenesis, depending on cell and tissue type (89), including acting as a scaffold protein for DNA repair proteins (41). The mutual relationship between TRIM29 and Tip60 is not clear-cut in cells exposed to etoposide. However, the results shown here suggest that GH induces TRIM29 and suppresses Tip60, and may be at least partially responsible for decreased ATM phosphorylation in cells treated with both etoposide and GH. Mechanisms underlying GH effects on TRIM29 expression are yet unknown.

ATM orchestrates cellular responses to DNA damage by activating checkpoint molecules, apoptosis, senescence, chromatin structure alterations, and DNA repair, while the primary role of DNA-PK is to promote NHEJ (33, 90). MRN-dependent ATM stimulation triggers phosphorylation of DNA-PKcs to induce NHEJ, while also directly promoting HR (91). Although we cannot exclude other pathways for reduction of DNA-PKcs phosphorylation, we detected reduced DNA-PKcs phosphorylation with decreased NHEJ in cells treated with GH, suggesting that GH action on DNA-PKcs may be mediated by decreased ATM activity.

Acromegaly patients with high ambient circulating GH levels develop colon polyps and other soft tissue tumors (21, 22) and have increased risk for developing colon adenocarcinoma (22). As colon polyp and adenoma frequency also significantly increase with age (92), our results point to the risk of GH replacement therapy in pituitary-replete healthy adults. As functional DDR pathways and adequate DNA repair are protective against neoplastic insults (93), GH administration to pituitary-replete adults may confer an unacceptable risk for epithelial proliferation and could contribute to field cancerization due to colon epithelial cell reprograming (94). In line with this hypothesis, short-stature adults with Laron syndrome (28), as well as animals with GHR deficiency (20, 29, 95), are protected from developing neoplasms.

Our results identify potentially novel mechanisms underlying protumorigenic properties of high GH levels. These results raise awareness that long-term inappropriate GH treatment may present a risk of developing colon epithelial cell transformation, as well as activation of preexisting low-grade tumors.

Supplemental methods are available online with this article.

Mice

Animal experiments were approved by the Cedars-Sinai Institutional Animal Care and Use Committee (IACUC 5587). GHR^–/– mice (B6N[Cg]-Ghr^{tm1b[KOMP]wtsi/3J}) were purchased from The Jackson Laboratory. Breeding was performed with heterozygous males and females so WT and GHR^–/– mice were obtained from the same breeding. In the course of breeding, heterozygous mice were backcrossed with WT mice at least 5 times.

Xenograft model. HCT116 cells stably infected with lenti-murine GH (5 × 10⁵ cells in 0.05 ml PBS) were mixed (1:1) with Matrigel (Corning) and injected s.c. into the right flank of athymic nude female mice (The Jackson Laboratory) to establish a model of excess systemic GH. Control mice were injected with HCT116 cells infected with empty vector. All mice developed xenograft tumors. Increased circulating GH and IFG-1 levels were confirmed with ELISA (MilliporeSigma).

Experimental metastases. A colorectal cancer mouse model exhibiting spontaneous metastases originating from intrasplenic primary tumor was employed (57, 96). When xenograft tumors reached 0.5³ cm, HCT116 cells (500,000 cells/100 μl PBS per animal) were injected into the spleen of xenograft bearing mice. Each group included 14 animals. Mice were weighed once a week, and metastasis development was monitored. Four weeks later, mice were sacrificed and abdominal organs examined for metastases. One mouse from the lentiGH group developed abdominal abscess, had to be euthanized 1 weeks after intraspleen injection, and was excluded from the study.

Cells and treatments

Human colon carcinoma HCT116 cells and human nontumorous breast cells (MCF12A) were obtained from American Type Culture Collection (ATCC). HCT116 were cultured in McCoy’s 5A medium (Invitrogen) and 10% FBS. MCF12A were cultured in DMEM/F12 medium (Invitrogen) with 0.5 μg/ml hydrocortisone (MilliporeSigma), 10 μg/ml insulin (MilliporeSigma), 20 ng/ml EGF (Invitrogen), and 5% horse serum (Omega Scientific). hNCC were purchased (Applied Biological Materials) and cultured in PriGrow III Media (Applied Biological Materials) supplemented with 5% FBS. All media were supplemented with antibiotic/antimycotic solution from Gemini Bio-Products. Cells were infected or treated before passage 4, per manufacturer instruction.

Pegvisomant was donated by Pfizer (to SM). Etoposide (MilliporeSigma) was prepared as a 10 mM DMSO stock solution. Cells and 3-dimensional intestinal organoids were treated at the indicated doses for the indicated times. Recombinant human GH1 (Biovision) was reconstituted in PBS pH8 containing 0.1% BSA. Cells were placed in culture medium free of serum containing 0.1% BSA, GH was added at a concentration of 500 ng/ml, and cells were harvested at the indicated times.

Three-dimensional intestinal organoids

Fibroblasts were obtained from healthy human volunteer donors at Cedars-Sinai (IRB, 00027264). Three-dimensional intestinal organoids were generated from a control fibroblast 83i induced pluripotent stem cell (iPSC) line using an episomal plasmid reprogramming system. To induce definitive endoderm formation, all iPSCs were cultured with a high dose of activin A (100 ng/ml; R&D Systems) with increasing concentrations of FBS over time (0%, 0.2%, and 2% [vol/vol] on days 1, 2, and 3, respectively). Wnt3A (25 ng/ml; R&D Systems) was also added on the first day of endoderm differentiation. To induce hindgut formation, cells were cultured in advanced DMEM/F12 with 2% (vol/vol) FBS, along with CHIR 99021 (2 μM; Tocris) and FGF4 (500 ng/ml; R&D Systems). After 3–4 days, free-floating epithelial spheres and loosely attached epithelial tubes became visible and were harvested. Epithelial structures were subsequently suspended in Matrigel and then overlaid in intestinal medium containing CHIR99021 (2 μM; Tocris), noggin, EGF (both 100 ng/ml; both R&D Systems), and B27 (1×; Invitrogen). Organoids were passaged every 7–10 days thereafter (97).

Image analysis of γH2AX with CellProfiler

To analyze γH2AX intensity, we pretreated hNCC with 500 ng/ml GH for 6 hours and then treated with 5 μM etoposide. Twenty-four hours after etoposide treatment, cells were fixed and confocal images analyzed using the MIT CellProfiler software (Carpenter Genome Biology 2006 7:R100) using the Speckle Counting analysis pipeline (http://cellprofiler.org/examples/#Speckles), which identifies intranuclear foci and computes per-object aggregate measurements, including number of foci per nucleus and intensities. Confocal images were analyzed without speckle-enhancer processing. Cells with intense nonfocal γH2AX signals that may reflect apoptosis were excluded from analysis. Twenty to 30 nuclei per image, and 5 images per group, were analyzed. Speckle intensity measurements were then further analyzed (98).

DSB repair assays

DSB repair was assessed as described (52, 99). Briefly, NHEJ reporter cassette and HR reporter cassette were chromosomally integrated in hNCC. hNCC were nucleofected with DNA reporter cassettes using Amaxa Basic Nucleofector Kit for Primary Mammalian Epithelial Cells (Lonza, catalog VPI-1005) in Amaxa Nucleofector-I (Lonza, program no. W-01). These cassettes contain the GFP gene with recognition sequences for I-SceI endonucleases for induction of DSBs. Stably transfected cells were selected for 10 days with 1 mg/ml of G418 and then cotransfected (by nucleofection) with 5 μg plasmid encoding I-SceI endonuclease (pCBASceI; Addgene, plasmid no. 26477) to induce DSBs, and 0.5 μg plasmid encoding DsRed (pDsRed2-N1; Clontech) to control for transfection efficiency. Intact reporters are negative for GFP. Upon induction of a DSB by I-SceI digestion, the functional GFP gene was reconstituted. The percent of GFP⁺ cells corresponds to the efficiency of DNA DSB repair, and the percent of DsRed⁺ cells indicates the efficiency of transfection. Five days after transfection, the number of GFP⁺ and DsRed⁺ cells was determined by flow cytometry. The ratio between GFP⁺ and DsRed⁺ cells was used as a measure of DSB repair efficiency. In hNCC, NHEJ repair efficiency was 0.77–0.83, and HR efficiency was approximately 0.05 in control cells, which is consistent with other reports (39, 43, 44). FACS analysis was performed using FACSCanto (BD Biosciences). For each treatment, a minimum of 50,000 cells was analyzed by FACS. Final data analysis was done using FlowJo software.

Statistics

Aggregate measurements of γH2AX foci were analyzed by nonparametric Wilcoxon rank sum test. In all other experiments, differences between groups were tested with mixed model regression to allow for random effects of intra-assay variation. Post hoc testing was performed with Tukey’s test to control for multiple comparisons. Residuals were inspected to confirm that data met assumptions necessary for parametric testing. Image quantification in colon tissue was assessed using 2-tailed Student’s t test. Differences were considered significant where P < 0.05. Data are graphed as fold-change or percent of control, but statistical testing was performed on raw numbers. For metastases, analysis data were log-transformed prior to analysis to meet assumptions necessary for parametric testing. Differences in number of tumors was tested between groups in one-way ANCOVA to adjust for circulating levels of GH.

Study approval

Animal experiments were approved by the Cedars-Sinai Institutional Animal Care and Use Committee (IACUC, 5587). Generation of iPSC from fibroblasts obtained from healthy human volunteer donors at Cedars-Sinai was approved by the Cedars-Sinai Medical Center IRB (no. 00027264).