Agonists of the TRAIL death receptor DR5 sensitize intestinal stem cells to chemotherapy-induced cell death and trigger gastrointestinal toxicity

Abstract

The combination of TRAIL death receptor agonists and radiochemotherapy to treat advanced cancers continues to be investigated in clinical trials. We previously showed that normal cells with a functional DNA damage response (DDR) upregulate the expression of death inducing receptor DR5/TRAILR2/TNFRSF10B in a p53-dependent manner that sensitizes them to treatment with DR5 agonists. However, it is unclear if targeting DR5 selectively sensitizes cancer cells to agonist treatment following exposure to DNA-damaging chemotherapy, and to what extent normal tissues are targeted. Here, we show that the combined administration of the DR5 agonistic monoclonal antibody (mAb) and chemotherapy to wildtype mice triggered synergistic gastrointestinal toxicities (GIT) that were associated with the death of Lgr5+ crypt base columnar (CBC) stem cells in a p53- and DR5-dependent manner. Furthermore, we confirmed that normal human epithelial cells treated with the human DR5-agonistic mAb and chemotherapeutic agents were also greatly sensitized to cell death. Interestingly, our data also indicated that genetic or pharmacologic targeting of Chk2 may counteract GIT without negatively impacting the antitumor responses of combined DR5 agonist/chemotherapy treatment, further linking the DDR to TRAIL death receptor signaling in normal cells. In conclusion, the combination of DR5-targeting agonistic mAbs with DNA damaging chemotherapy may pose a risk of developing toxicity-induced conditions, and the effects of mAb-based strategies on the dose-limiting toxicity of chemotherapy must be considered when establishing new combination therapies.

Introduction

Novel targeted therapeutics in oncology hold the promise to selectively kill the tumor cell population with reduced toxicity to normal tissues that hampers current standard radiochemotherapy that often has a narrow therapeutic index. One strategy that has been under development is the targeting of the extrinsic apoptotic signaling pathway for the treatment of advanced systemic malignancies. Tumor necrosis factor-related apoptosis inducing ligand (TRAIL) is a member of the tumor necrosis factor superfamily. Endogenous TRAIL and its death inducing receptors DR4 and DR5 (TNFRSF10B) in humans play a powerful role in tumor immune surveillance, inflammation and tumor suppression in vivo (1-4). Ligand-dependent clustering of the DR4 and DR5 receptors and activation of downstream caspases triggers fairly rapid and fulminant apoptosis selectively in cancer cells. The specific molecular mechanisms that render cancer cells increasingly susceptible to apoptosis triggered through the TRAIL-system remain to be fully understood although a role for TRAIL decoy receptors has been suggested to protect normal cells.

Despite the observations that TRAIL death receptor agonists (TDRAs) are generally non-toxic to normal cells and are overall well tolerated, some precautions have been suggested. Certain preparations of recombinant TRAIL have been found to be toxic to human hepatocytes in vitro (5,6) and some agonistic mAb's targeting DR4 and DR5 can kill normal human hepatocytes in vitro (7). Human hepatocytes isolated from steatotic and hepatitis C-positive livers appear to be sensitive to both untagged and tagged TRAIL (8). In experimental mouse models, high dose treatment with MD5-1, an agonistic mAb targeting mouse DR5, triggered cholangitis with a histological appearance reminiscent of human primary sclerosing cholangitis (9). Some early clinical trials reported DLT's of DR5 mAb's that may involve liver toxicity. High dose (20 mg/kg) treatment with lexatumumab resulted in asymptomatic and reversible transaminase and amylase elevations in a phase 1 trial of patients with advanced malignancies (10). Similarly, transaminitis was noted in 1 out of 37 patients subjected to Apomab, an agonistic DR5 mAb (11). It is important to note that it can be difficult to attribute liver toxicity to any particular therapy in patients with metastatic disease to the liver especially when it is progressing. Activation of p53 has been shown to sensitize spermatocyte-like cells to recombinant TRAIL or DR5-targeting mAb's (12).

More recent clinical trials have focused on the integration of DR5-targeting mAb's, that have significantly longer plasma half-life than recombinant TRAIL, with first-line radiochemotherapy that remains the mainstay in oncology in order to help improve response rates. Some data from early phase clinical trials suggest toxicity when DR5-targeting mAb's are administered in combination with chemotherapy. A clinical phase 1 and 2 study assessing the DR5-targeting antibody Conatumumab in combination with FOLFOX6 plus Bevacizumab for the treatment of metastatic CRC was unable to document an improved a response rate with Conatumumab. In general Conatumumab was well tolerated but five (5) percent of the patients receiving 10 mg/kg bw of Conatumumab experienced grade 4 diarrhea and fifteen (15) percent experienced grade 3 hypokalemia (13). A randomized, placebo-controlled phase 2 study of Conatumumab in combination with FOLFIRI for second-line treatment of mutant KRAS metastatic CRC was able to document a trend towards improved response in the FOLFIRI/Conatumumab arm (14). However, the potential of improved response in the FOLFIRI/Conatumumab arm was associated with a trend towards an increased number of adverse events such as diarrhea, neutropenia, fatigue, anemia and abdominal pain as compared to the FOLFIRI/placebo arm.

We show that targeting of DR5 in mice concomitant with treatment with 5-FU and CPT-11 trigger a moribund state in the animals not present following either treatment alone and this phenotype is associated with p53-dependent erosion of the GI epithelium. In the case of CPT11 plus DR5 targeting, this GIT is also Chk2-dependent. GIT is preceded by apoptosis in the stem cell region of the GI tract associated with loss of Lgr5+ cells. Furthermore, toxicity following this treatment is highly dependent on specific components of the apoptotic machinery downstream of the receptors in a chemotherapy-specific manner. Our results point to an unanticipated molecular complexity of the interaction between the extrinsic and intrinsic apoptosis pathways that has relevance to pro-apoptotic cancer therapy combinations in the clinic. Interestingly, we show that pharmacologic targeting of cell cycle checkpoint kinase 2 (Chk2) protects the GI epithelium and facilities dose-escalation without negative impact when treating p53-deficient syngeneic colon cancer cells. A detailed understanding of these interactions along with insights into their potential reversal through pharmacological intervention may improve the therapeutic index during cancer therapy with pro-apoptotic drugs that target the TRAIL death receptor pathway.

Material and Methods

Mice and treatment

Mice were purchased from Jackson Laboratory (Jackson Laboratory, ME). Chk2^-/- and DR5^-/- mice have been described previously(3,15) 3p53^F/F; mT/mG; lgr5 and DR5^-/-; mT/mG; lgr5 mice were generated in house. All animal care and treatment procedures employed were approved by an Institutional Animal Care and Use Committee.

In vivo labeling of apoptotic cells

B6.129P2-Lgr5tm1(cre/ERT2)Cle/J mice were injected with SR-FLIVO™ (ImmunoChemistry Technologies LLC) according to the manufacturer's instructions. Small intestinal crypts were isolated as previously described (16). Labeled cells (GFP+) and/or FLIVO (red fluorescent) were quantitated with an inverted fluorescence microscope (Zeiss, Axiovert100) coupled to a CCD camera.

Quantitation of histological and immunohistochemical findings

All slide sections were coded, analyzed and quantified blindly by counting ten randomly selected 60× or 100× image fields of three non-serial sections of the same specimen. The area percentage of the total GI area subjected to histological evidence of injury was estimated using ImageJ (NIH Image 1.62 software). Also, the number of TUNEL-positive cells (see ‘Histology and Immunohistochemistry’) was assessed on digital images of specimens with the use of ImageJ (NIH Image 1.62 software). Appropriate statistics were applied to the generated data (see “Statistical Analysis”).

Non-invasive high-resolution near-infrared (NIR) optical imaging

Non-invasive NIR imaging was performed as described previously (17).

Expression and purification of recombinant human TRAIL

Expression and purification of rhTRAIL was carried out as described previously (18).

Results

DR5-targeting sensitizes to the GI toxicity inflicted by 5-FU and CPT-11

In order to assess the toxicity and dose-response characteristics of two commonly used chemotherapeutics fluorouracil (5-FU) and irinotecan (CPT-11) in the absence and presence of DR5-targeting mAb's we treated mice with an isotype-specific control antibody (IgG) or MD5-1. MD5-1 triggers apoptosis in malignant cells that is dependent on FLIP and completely inhibited by the pan-caspase inhibitor z-VAD-fmk (19). To our surprise the supplementation of MD5-1 to either 5-FU or CPT-11 caused anorexia and increased lethality at lower than expected doses of the chemotherapeutic agents (Fig. 1A, B and Fig. S1A, B). For example, all the mice in the 250 mg/kg bw-group of 5-FU (Fig. 1A, right panel) succumbed when challenged with MD5-1 whereas 60% of the mice where alive when IgG was administered in place of MD5-1. Administering MD5-1 alone at this dose-level, without the presence of either chemotherapeutic, was not toxic to mice for an observational period of up to 90 days (Fig. S1C and data not shown). From the two parameters (weight loss and moribund state) that were being assessed, weight loss was the more sensitive end-point (with an approximately 2-fold lower ED₅₀, data not shown). Subsequently we used weight loss area under the curve (AUC) values to generate dose-response relationships for CPT-11 and 5-FU in the presence or absence of the DR5-targeting mAb MD5-1 (Fig. S1A, B, and Fig. 1C, D). From the dose-response curves it is clear that the presence of the DR5-targeting antibody MD5-1 sensitized the mice to weight loss and anorexia. The ED₅₀-values for CPT-11 was 416.6 mg/kg bw (95% CI: 383.9 – 452.0 mg/kg bw) in the presence of control IgG and 175.7 mg/kg bw (95% CI: 143.5 – 215.2) in the presence of MD5(Fig. 1). -1 The corresponding values for 5-FU were 375.7 mg/kg bw (95% CI: 260.6 – 541. 6) and 106.1 mg/kg bw (95% CI: 97.44 – 115.6) in the absence (IgG) and presence of MD5-1, respectively (Fig. 1). Thus, based on weight loss as a toxicity end-point, mice were able to withstand 2.4- and 3.6-fold less CPT-11 and 5-FU respectively following targeting of DR5 with agonistic mAb's in vivo. Given that CPT-11 and 5-FU are commonly administered to patients in combination for the treatment of metastatic CRC as part of FOLFIRI-regimen we employed allometric dose conversion factors from human to mouse for doses of 5-FU and CPT-11 employed in clinical protocols. Following treatment with 5-FU/CPT-11/MD5-1 we found that wild-type mice developed anorexia and succumbed to treatment when the DR5-targeting antibody MD5-1 was included in the treatment regimen (Fig. 1E and F and ‘data not shown’). Interestingly, mice lacking one or both alleles of DR5 (DR5^+/- and DR5^-/-) were resistant to anorexia (P<0.001, student's t-test) and protected from the lethality (P<0.001, Log-rank Mantel Cox test) triggered by the CPT-11/5-FU/MD5-1 combinatorial treatment (Fig. 1E, F). By contrast to our findings with MD5-1, when recombinant mouse TRAIL (rmTRAIL) or rhTRAIL was combined with CPT-11 no sustained anorexia or moribund state was encountered (Fig. S1D). This could potentially implicate a protective role for DcR's that may be abundantly expressed on normal cells. Collectively, our data suggest that DR5-agonistic antibodies may exacerbate toxicity when combined with chemotherapy commonly used for the treatment of CRC.

Figure 1. The DR5-targeting antibody MD5-1 triggers synergistic toxicity in mice when combined with chemotherapy.

(A) Survival of mice treated with the combination fluorouracil (5-FU) and isotype control antibodies (IgG) (left diagram) or 5-FU and the DR5-targeting MD5-1 antibody (right panel) (*P<0.05 and ^#P<0.05 log rank statistics; N=5) respectively. (B) Survival curves for mice treated with CPT-11/IgG and CPT-11/MD5-1. (*P<0.05 log rank statistics; N=5). Dose-response analysis of AUC for the weight curves following different doses of 5-FU (C) and CPT-11 (D) in the absence (IgG; green curves) or presence (MD5-1; red curves) of DR5-targeting antibodies. Weight loss (E; P<0.001; Unpaired students t test) and survival (F; P<0.001, Log-rank [Mantel-Cox] Test) curves for wild-type and DR5^-/- mice treated with clinically relevant concentrations of the combination 5-FU, CPT-11 and the DR5-agonistic antibody MD5-1.

DR5-dependent liver toxicity occurs following high dose treatment withMD5-1

Previously reported data have shown that liver toxicity may arise in mice following higher doses of MD5-1 than applied here (9). Mice that became anorexic in our model did not show visible signs of liver choleastatic disease nor where there histological signs of injury in other organs than the GI tract (Figure S2A). We found that MD5-1 alone was well-tolerated without any obvious signs of toxicity up to a dose level of 20 mg/kg bw (Fig. S2B and C). Consistent with the previous report, necropsy indicated that weight loss was well correlated with yellow skin discoloration, enlarged gall bladders and histological signs of periportal loss of cholangiocytes and hepatocyte atrophy (Fig. S2D). Interestingly, the addition of a single dose of CPT-11 (40 mg/kg bw IV) caused sensitization of mice to MD5-1 with respect to choleastatic disease at a dose of 10 mg/kg bw of MD5-1 (Fig. S2C). Furthermore, we assessed the impact of mouse strain on the observed GIT inflicted by combined DR5-targeting and chemotherapy since a strict strain dependency has been described for the manifestation of liver toxicity following MD5-1-exposure (9). By contrast to what has been reported, we found that the combination of 5-FU/MD5-1 triggered anorexia and a moribund state in previously reported “resistant” strains with a slightly delayed response compared to that of C57BL6 (Fig. S2E and F). One exception to this was the CB17 SCID mouse strain that was found to be highly sensitive to the 5-FU/MD5-1 combination. This may indicate that chemotherapy sensitizes to DR5-targeting independently of mouse strain genetics associated with sensitization of the liver to MD5-1 alone. Furthermore, we aimed to test the impact of treatment of chemotherapy and MD5-1 in mice lacking B cells since such mice have been suggested to fail to trigger apoptosis and an anti-tumor response following MD5-1 treatment (20). Curiously, such mice were somewhat protected from GIT as evident by a reduction in weight accumulation in young (4-6 weeks old) mice following treatment with CPT-11/MD5-1 (Fig. S2G-I). This indicated that the presence of B cells is not critical for the manifestation of GIT in our model in contrast to what has been previously reported with respect to MD5-1's activity in tumor models.

Low doses of MD5-1 trigger DR5-dependent GIT when combined with DNA damaging chemotherapy

Necropsy of moribund animals in our MD5-1 ‘low dose’ model revealed occasional hemorrhage and distension of the distal small intestine and proximal colon whereas other organs (liver, spleen, kidney, heart, lung and brain) had a normal gross appearance (Fig. 2A, S2A and data not shown). Histological evidence suggested epithelial erosion, occasional lymphocyte infiltration and mucosal wall thickening in the small intestine and colon was associated with the manifestation of toxicity (Fig. 2A and S2A). Quantitative analysis of H/E stained slides of the GI tract isolated from mice of various treatments showed that mice subjected to either the combination of CPT-11/MD5-1 or 5-FU/MD5-1 had significantly fewer intact crypts per millimeter of alimentary tract as compared to mice treated with a combination of CPT-11/IgG (CPT-11) or 5-FU/IgG (5-FU) respectively (Fig. 2B and C). IHC for the GI epithelial-specific marker claudin-7 or the mesenchymal-specific marker vimentin supported epithelial erosion and expansion of mesenchymal cells in the GI tract of mice subjected to the combination of 5-FU, CPT-11 and MD5-1 (Fig. 2D and E). We employed near-infrared (NIR) imaging using the inflammatory-specific probe Prosense680 (PS680) that has been previously used to study inflammation in the GI tract (17). There was a clear increase in PS680-labeling of the small intestine and colon of mice treated with CPT-11/MD5-1 (Fig. 2F) Taken together our data indicate that the DR5-dependent inflammatory injury to the GI tract is associated with the observed gross toxicity observed in mice following the drug combinations 5-FU/MD5-1 and CPT-11/MD5-1.

Figure 2. The DR5-targeting antibody MD5-1 triggers synergistic GI toxicity characterized by epithelial erosion and inflammation in mice when combined with 5-Fluorouracil and CPT-11.

(A) H/E staining of histological slide sections of the small intestine and colon from wild-type and DR5^-/- mice treated with CPT-11 plus MD5-1. Quantification of the number of crypts/mm in histological sections of the small intestine (B) and the colon (C) from mice treated with chemotherapy and MD5-1. DSS was used as an inducer of epithelial atrophy and as a positive control for mucosal injury. Error bars represent the standard deviation from the mean. (D) IHC of the colon from wild-type and DR5^-/- mice for claudin 7 (green; Cy2), vimentin (red; Cy3) and using DAPI (blue) to visualize cellular nuclei. Colons were stained at 7 days post-treatment. Representative pictures from at least 3 mice are shown. (E) IHC for claudin 7, vimentin and BrdU (blue; Cy5) of the small intestine (ileum) from wild-type mice subjected to 4 days post the indicated treatments. Representative pictures from at least 3 mice are shown. (F) Prosense680 (PS680) cleavage and retention (red signal) in the small intestine and colon of wild-type but not DR5^-/- mice following CPT-11 plus MD5-1. ‘BF’, bright field; ‘p’, proximal; ‘d’, distal; ‘AC’, ascending colon; ‘DC’, descending colon. Quantification of PS680-signal emission in the near-infrared spectrum from the colon and the small intestine of mice treated with CPT-11 plus MD5-1 indicates increased probe retention in wild-type animals.

DR5-targeting does not exacerbate dose-limiting myelosuppression following 5-FU and CPT-11

We assessed the impact of combining MD5-1 with chemotherapy on bone marrow toxicity, a common DLT with the chemotherapy. Although, the bone marrow showed signs of atrophy following such treatments, this was not well correlated with a moribund state (Fig. S2A). Moreover, mice receiving treatment with either CPT-11 or 5-FU combined with control IgG showed similar levels of apoptotic (sub-G1 positive) cells and atrophy in the bone marrow (Fig. S3A, B and data not shown). In order to further address the possibility that the toxicity phenotype was a result directly of rate-limiting damage to the GI tract, we subjected mice to whole-body γ-irradiation (WBR) using an “abdominal shield” (AS) constructed of lead covering the mouse from the sternum to the pelvic region effectively protecting the GI tract from radiation-induced injury and subsequently injected mice with MD5-1 (Fig. S3C). Wild-type mice subjected to WBR/MD5-1 and lacking AS showed a trend to increased weight loss as compared to wild-type mice that had their mid-section shielded and received WBR/MD5-1 (Fig. S3D and E). Furthermore, we subjected mice to lethal doses of WBR and performed bone marrow transplantation (BMT) with bone marrow lacking DR5 (DR5^-/- BM) (Fig. S3F). Female recipient mice received male DR5^-/- BM and after 40 weeks post BMT we were able to confirm complete engraftment of the DR5 “null” BM in the mice (Fig. S3G). However, mice receiving DR5^-/- BM showed equal sensitivity towards 5-FU/MD5-1 when compared to wild-type mice suggesting that loss of DR5 specifically in the BM did not rescue mice from the toxicity of DR5-targeting in combination with chemotherapy (Fig. S3H and S3I). Taken together our data indicates that the GI tract and not the BM is the bona fide site for our observed acute organ toxicity following targeting of DR5 in combination with 5-FU and CPT-11.

DR5-targeting sensitizes to 5-FU- and CPT-11-induced cell death in the GI tract

Given the propensity of DR5-targeting to trigger apoptosis we assessed the GI tract for apoptotic cells following 5-FU, CPT-11 and MD5-1. H&E staining of colon isolated at twenty-four (24) hr following treatment with 5-FU and MD5-1 suggested that presence of apoptotic cells by morphology and TUNEL staining following both treatments (Fig. 3A-D). Loss of DR5 completely protected the crypts from cell death induced by MD5-1 (Fig. 3D and S4A, B). Flow cytometric sub-G1 analysis and in vivo FLIVO-labeling of isolated intestinal epithelial cell confirmed the increased presence of apoptotic cells when DR5-targeting MD5-1 antibodies was combined with CPT-11 (Fig. S4C and D). We used the untransformed human fetal GI epithelial cell line FHs74Int to assess the impact of DR5-targeting following CPT-11 and 5-FU. Consistently with our mouse in vivo findings the FHs74Int cells were sensitized to apoptosis when the human DR5-agonistic monoclonal antibody mAb631 was combined with either CPT-11 or 5-FU (S4E-F). Furthermore, caspase-3/7 was increasingly activated as compared to control following mAb631 and CPT-11 (Fig. 3E). Indeed, a substantial increase in caspase-3/7 activation over that of either compound alone was observed following the combination of CPT-11/mAB631. The combination treatment significantly inhibited growth of the FHs74Int cells over 72 hrs in comparison to either compound alone (Fig. 3F). Interestingly, caspase-3/7 activation following mAb631 alone was poorly correlated with long-term cell survival of mAb631-treated FHs74Int cells. Furthermore, CPT-11 dose-dependent activation of caspase-3/7 was only observed in the presence of mAb631 and subsequently correlated with growth inhibition (Fig. 3G, H and S4G). Thus untransformed human GI epithelium might be increasingly sensitized to undergo apoptosis following DR5-targeting in the presence of high doses of CPT-11. Our findings are consistent with and appear to validate our earlier mouse model observations on apoptosis induction.

Figure 3. Apoptosis in the GI tract following targeting of the gastrointestinal epithelium with chemotherapy and DR5-targeting antibodies.

(A) H/E staining of the colon from mice treated with isotype control antibodies (IgG), DR5-agonist antibodies (MD5-1), 5-Fluorouracil (5-FU) or 5-Fluorouracil combined with DR5-agonist antibodies (F/M). Red arrows indicate apoptotic cells in colonic crypts. Representative images are shown. (B) Image analysis of TUNEL-stained histological sections of colon isolated from mice treated with 5-Fluorouracil and DR5-agonist MD5-1 antibodies. Error bars represent the standard error from the mean. (C) TUNEL-staining (brown stain) of the colon from treated mice and (D) the graphical representation of image analysis of TUNEL-stained slide sections from treated mice. Error bars represent the standard error from the mean. (E) and change in cellularity (F) of the normal GI epithelial cell line FHs 74Int following treatment with the DR5-receptor agonist antibody mAb1613 (M; 1.0 μg/mL), CPT-11 (C; 280 μg/mL) and the combination thereof (C/M). Error bars represent the standard error from the mean. Statistical analysis was performed by 2-way ANOVA test using Bonferonni correction. P<0.05 was considered significant.

Combined selective DR5-targeting with CPT-11 and 5-FU sensitizes Lgr5+ stem cells to cell death

It was somewhat surprising that MD5-1 triggered cell death in the GI tract of mice without co-treatment with chemotherapy (Fig. 3A-D and Fig. S5A-C). However, the magnitude of the cell death and its crypt spatial involvement was altered following the addition of either 5-FU or CPT-11. Scoring the crypt positions most frequently affected following CPT-11/MD5-1 (C/M) and 5-FU/MD5-1 (F/M)-treatments revealed that apoptosis (as determined by TUNEL-staining) was more frequently affecting the crypt bottom and CBC cells (Fig. 4A-C and S4A-C). In order to verify that CBC cells were increasingly targeted to death following the inclusion of DR5-targeting mAb's with chemotherapy we performed IHC for Lysozyme C (green fluorescence), a marker for paneth cells in the intestine, in combination with TUNEL-staining (red fluorescence). Following C/M we found that cells were increasingly dying in the stem cell niche between paneth cells (Fig. 4C) as compared to either treatment alone. However, the combinatorial treatment had no such effect in DR5^-/- mice subjected to C/M (Fig. S5A-C). Considering the deleterious impact the combined DR5-targeting had on the GI cell homeostasis we sought to address if the inclusion of MD5-1 with chemotherapy increasingly included Lgr5+ CBC cells in the cell death response by generating DR5^-/-; lgr5-EGFP-Cre-ERT2 (DR5^-/-; lgr5+) mice (21). These mice express EGFP under the Lgr5-promoter and EGFP-expressing cells can readily be detected by performing IHC for EGFP. Indeed, mice lacking DR5 never displayed loss EGFP-positive cells following treatment with either CPT-11 or 5-FU combinations as compared to treatment with single modalities of 5-FU, CPT-11 or MD5-1 (Fig. 4D and E). EGFP+ cells with apoptotic morphology were a feature observed almost exclusively in wild-type lgr5-EGFP-Cre-ERT2 (lgr5+) mice treated with the combinations of either C/M or F/M (Fig. 4D). We also used in vivo SR-FLIVO labeling to detect apoptotic lgr5+ cells following crypt isolation (Fig. S5D, E and Fig. 4F). Mice subjected to treatment with C/M showed increased SR-FLIVO uptake in GFP+ crypt cells ex vivo compared to crypts isolated from lgr5+ mice treated with vehicle, MD5-1 and CPT-11 (Fig. 4F). Moreover fewer Lgr5+ (EGFP+) cells were obtained following treatment with the combination treatment of C/M compared to treatment with either compound alone (Fig. S5E). Taken together our data indicate that lgr5+ cells are a de novo target for selective DR5-targeting when combined with chemotherapy.

Figure 4. GIT following combined DR5-targeting and chemotherapy is associated with depletion of Lgr5+ GI stem cells.

Combinatorial treatment with 5-FU (F)/MD5-1 (M) (A) and CPT-11 (C)/MD5-1 (M) (B) triggers more frequent apoptosis in the crypt bottom (Cell position 0-3) as compared to either modality alone. Error bars represent the standard error from the mean. (C) IHC for paneth cells (Lysozyme C; Cy2; green fluorescence) and apoptotic cells (TUNEL; Cy3; red fluorescence). Representative pictures from at least 3 mice are shown. (D) IHC of transverse SI (ileum) sections showing crypt bottoms positive for EGFP (Lgr5+ cells) in Lgr5-EGFP-Cre-ERT2+ and DR5^-/-; Lgr5-EGFP-Cre-ERT2+ transgenic mice following treatment with C/M. Representative pictures of three mice are shown. (E) IHC for GFP on small intestines isolated from Lgr5-EGFP-cre-ERT2 mice subjected to No treatment (V), MD5-1 (M), 5-FU (F) or the combination of F/M. (F) In vivo labeling of apoptotic cells in Lgr5-EGFP-Cre-ERT2 mice using SR-FLIVO and subsequent quantification of crypts containing cells with combined green and red fluorescence following treatment with vehicle, MD5-1, CPT-11 or the combination of C/M. Error bars represent the standard deviation from the mean.

Both 5-FU and CPT-11 triggered the stabilization of p53 in the colonic epithelium of mice (Fig. 5A). Following 5-FU, induction of p53, p21 and at later time-points c-Myc was observed (Fig. 5B). Expression profiling of colonic mRNA at twenty-four (24) hr following either CPT-11 or 5-FU revealed the induction of several ‘apoptosis-inducers’ as well as ‘p53-dependent genes’ (Fig. S6A and B). Indeed, mice lacking p53 (p53 ‘null’) in all tissues were completely protected from the moribundicity triggered by F/M (Fig. 5C). We also subjected transgenic mice lacking critical molecules for cell cycle arrest and apoptosis signaling following DNA damage to treatment with F/M and C/M (Fig. S7). Predictably, we found that the toxic responses to both treatments in the mice required the presence of DR5 (Fig. S7B and H), however surprisingly the reliance's on caspase-3, bid, puma and p21 to trigger GIT were remarkably different for the two drug combinations. Although puma was upregulated in the colonic epithelium following treatment with either 5-FU or CPT-11, puma was only required to trigger toxicity following F/M (Fig. S7C and I). By contrast, bid appears to be required for toxicity following C/M although this trend did not reach statistical significance (P=0.1712, Log-rank test). Loss of bid appeared to protect the colonic epithelium from apoptosis (as evaluated by TUNEL-staining) following C/M but not F/M (data not shown). Interestingly, loss of the cell cycle regulator p21 also sensitized to C/M (P=0.0002, Log-rank test) (Fig. S7D, E, J and K). Taken together our data support the requirement of intact p53-signaling to trigger GIT. However, downstream p53 targets show a differential requirement for the triggering of GIT following DR5-targeting in combination with chemotherapy.

Figure 5. Targeting p53 in Lgr5+ cells protects from eradication of Lgr5+ cells and their progeny following treatment with chemotherapy and DR5-targeting mAb's.

Immunohistochemical staining (A) and western blot (B) for p53 and downstream target genes following 5-Fluorouracil (5-FU) and irinotecan (CPT-11). (C) Mice with somatic deletion of p53 are resistant to toxicity following combinatorial treatment with 5-FU plus MD5-1 (*P<0.05; log-rank test) (D) Immunohistochemistry (IHC) for p53 seven (7) days post tamoxifen administration to p53^f/f;lgr5 and p53^f/+;lgr5 mice. Mice were challenged with 5-FU to induce p53 expression in the crypts. Crypts lacking expression of p53 (middle panel) and having ‘chimeric’ expression of p53 was found in p53^Δ/Δ;lgr5 mice. (E) Deletion of p53 in Lgr5+ cells through the treatment of p53^flox/flox; lgr5 mice with tamoxifen for twenty-four (24) hrs and subsequent treatment with CPT-11/MD5-1. Mice were sacrificed at seven (7) days post treatment with CPT-11/MD5-1. (F) Lineage tracking of Lgr5-dependent progeny using p53^flox/+; mT/mG; lgr5 and p53^flox/flox; mT/mG; lgr5 mice. Mice were sacrificed at twelve (12) days post the initiation of treatment with CPT-11 plus MD5-1. Lgr5+ cells lacking p53 were increasingly viable and able to generate progeny following treatment with CPT-11/MD5-1.

In order to more specifically assess the impact of lost p53 expression in Lgr5+ cells we generated p53^f/f; lgr5; mT/mG mice that would allow deletion of p53 in Lgr5+ cells and subsequent lineage tracking of cell progeny using the Cre-reporter mT/mG transgene (22). Successful deletion of p53 was also assessed by IHC for p53 on serial section of GI tracts isolated from mice subjected to treatment with 5-FU (Fig. 5D). Deletion of the p53 gene in Lgr5+ cells through the use p53^f/f; lgr5 mice concomitantly with treatment with C/M partially rescued villous blunting and reduced toxicity (Fig. 5E). Furthermore, lineage tracking of the progeny of Lgr5+ cells was improved when p53 was deleted in Lgr5+ cells through administration of tamoxifen to p53^f/f; lgr5; mT/mG mice immediately prior to treatment with C/M (Fig. 5F). These data indicates that Lgr5+ cells in the GI tract are increasingly targeted when DR5 agonists are combined with 5-FU and CPT-11.

Targeting of Chk2 prevents GIT following the combination CPT-11/MD5-1

Considering that some key molecules involved in the pro-apoptotic DNA damage response (DDR) were found to be critical determinants of the toxic response to chemotherapy combined with DR5-targeting mAb's we hypothesized that targeting of the cell cycle checkpoint kinase 2 (Chk2) could potentially modulate such toxicities. Interestingly, mice lacking the Chk2-gene (Chk2^-/-) were susceptible to toxicity following treatment with F/M but not C/M (Fig. 6A). In fact, Chk2^-/- mice appeared to be completely resistant to developing a moribund state triggered by the C/M drug combination. These findings translated into less epithelial atrophy and a reduction in crypt apoptosis in the colons of Chk2^-/- mice following C/M (Fig. S8A and 6C). Using a pharmacologic Chk2 inhibitor (Chk2 inhibitor II ‘CI-II’) (23) resulted in a decreased moribund state and reduced the PS680 signal from the GI tract of wild-type mice treated with C/M (Fig. 6D and S8B-D). Thus, our data indicate that targeting of Chk2 at the genetic level or the kinase function of Chk2 may prevent DLT's when the DR5-receptor is being targeted in combination with CPT-11.

Figure 6. Chk2 controls GI toxicity following DR5-targeting and CPT-11 treatment.

(A) Body weight loss and moribund state following treatment with CPT-11 plus MD5-1 is controlled by Chk2. Error bars represent the standard error. (B) TUNEL-staining of colonic crypts at 24 hr following treatment with combinations as indicated and histological evaluation (right panels). Representative images are shown. (C) Weight loss and survival curves generated following co-treatment of mice with C/M with Chk2 inhibitor II (CI-II). The data were analyzed for statistically significant differences (P<0.05) using the 2-way ANOVA and the log-rank test. Error bars represent the standard error. (D) Western blot of colonic scrapings from wild-type and Chk2^-/-mice subjected to treatment with 5-FU and CPT-11 for 24 hr. (E) Western blot assessment of siRNA-mediated survivin/birc5 knock-down efficacy in FHs 74Int cells. (F) Activation of caspase-3/7 in FHs74Int cells subjected to scrambled (CTRL) siRNA, siRNA targeting survivin and following treatment with CPT-11 (280 μg/mL) and mAb631 (1.0 μg/mL) for eighteen hours. An unpaired student's t test was used for the statistical analysis. P<0.05 was considered statistically significant and an N=3/treatment group was used. The mean with standard error bars are shown.

Gene expressions analysis indicated that CPT-11 triggered repression of inhibitor-of-apoptosis protein (IAP) survivin (birc5) (Fig. 6E and data not shown). We targeted expression of survivin using siRNA in human FHs74Int cells and subjected such cells to treatment with CPT-11/mAb631 (Fig. 6F). Indeed, targeting of anti-apoptotic survivin resulted in increased activation of caspase-3/7 following CPT-11/mAb631 suggesting that survivin has a functional role in repressing apoptosis following DNA damage and DR5-targeting. Given the in vivo role of Chk2 to protect mice from toxicity following C/M we tested the capacity of three (3) different Chk2-inhibitors in PV1019 (24), CI-II and CI-III (25) to prevent caspase-3/7 activation in FHs74Int cells following CPT-11/mAb631 (Fig. S8E). Indeed, a dose-dependent repression of caspase-3/7 activation was observed following co-treatment with the Chk2 inhibitors suggesting a functional role of Chk2 in mediating cell death in human cells following DNA damage and DR5-targeting.

To address if the newly found molecular mechanism could be used to reduce GIT and improve treatment efficacy of the combination of CPT-11 and DR5-targeting mAb's we used a reduced dose of CPT-11 (116 mg/kg bw) to treat syngeneic mouse colon cancer cells (p53dmc/Ras/Myc) overexpressing mutated K-Ras and c-Myc in syngeneic mice in vivo. The combination of CPT-11/MD5-1/CI-II (C/M/CI-II) appeared to improve the response-rate of subcutaneous p53dmc/Ras/Myc grafts as compared to that of mice treated with CPT-11/MD5-1/vehicle (C/M/V). Although the observed difference in tumor load was not statistically significantly different between the treatments at the end of the experiment, (P=0.26, Mann-Whitney test) the tumor doubling times in days for control (vehicle alone), C/M/V and C/M/CI-II were 2.4 (95%-CI: 2.24-2.54), 4.6 (95%-CI: 3.56-6.58) and 6.2 (95%-CI: 4.25-11.67) indicated significant differences between the treatment groups (Fig. S9A and B). Interestingly, mice subjected to the combinatorial treatment of C/M/CI-II lost less body weight throughout the experiment as compared to mice subjected to treatment with C/M/V (Fig. S9C and D). No sign of increased toxicity was observed in the mice when CI-II was combined with CPT-11 and MD5-1 suggesting the Chk2 inhibitor was ‘nontoxic’ in this context.

In summary, our data indicates that the GIT observed in our models differs with respect to the molecular underpinnings in a chemotherapy-specific manner despite sharing a similar pathological manifestation in the GI tract of mice (Fig. 7). Both treatments are dependent on intact DR5 and p53-signaling however, the relative dependency on canonical downstream molecules of the p53-repsonse is altered dependent on the specific chemotherapeutic when DR5-targeting mAb's are included. The combination of CPT-11/MD5-1 relies increasingly on a cell cycle (p21) dependent mechanism, bid and Chk2 to trigger GIT (Fig. 7A) whereas the combination of 5-FU/MD5-1 relies mainly on puma to inflict injury to the GI tract of mice (Fig. 7B). Although we did not address this directly, previous data of others and ours suggest that Myc-upregulation may sensitize to DR5-targeting (26). Thus, Myc-upregulation and the inherently rapid GI epithelial regeneration following injury might cause Lgr5+ GI stem cells to become increasingly sensitized to death receptor agonists.

Figure 7. Model schematics.

(A) Molecular determinants of gastrointestinal toxicity (GIT) involved following treatment with CPT-11 in combination with DR5-agonistic monoclonal antibodies (mAb). (B) Genetic determinants of GIT following the combination of 5-FU plus MD5-1. Dark red solid boxes indicate mediators of GIT following treatment as indicated by genetic evidence. Solid green boxes indicate genes that limit GIT following chemotherapy and DR5-targeting. Dashed dark red boxes indicate molecules with putative involvement based on their known biology and correlative expression.

Discussion

Our work addresses the dose limiting toxicity in the GI tract due to combinations of chemo or radiotherapy plus specific targeting of the DR5 death receptor pathway, and provides insights that in the future may be pursued to reduce toxicities while maintaining or improving the therapeutic index. In particular, targeting Chk2 may be an attractive strategy to pursue further to reduce dose-limiting toxicity in the GI tract. In order to address the possibility of toxicity when DR5 is targeted in combination with chemotherapy we treated mice with a DR5 agonistic mAb (MD5-1) in combination with 5-FU and CPT-11. Our data indicate that low dose (1.0 mg/kg bw) supplementation of DR5 targeting mAb's in combination with either 5-FU and CPT-11 is sufficient to trigger a lethal GIT syndrome in mice (Fig. 1, 2 and Fig.S1 and 2A-F). The GIT observed with combined DR5 targeting plus chemotherapy is completely dependent on the presence of DR5 and the tumor suppressor p53 in mice (Fig.1E, F, 5C). Furthermore, we found that this toxicity was dose-limiting for the treatment of syngeneic colorectal tumor grafts (Fig. S9A-D) and correlated with increased apoptosis in the TA zone of the gut as the disappearance of Lgr5+ CBC cells (Fig. 4A-C and Fig. S5D-F).

Lgr5+ crypt basal columnar cells have been reported to be indispensible for epithelial restitution and as a predictor of lethality in mice from the radiation-induced gastrointestinal syndrome (27-29). These prior findings are in agreement with findings in our model where chemotherapy (or DR5-targeting alone) is sufficient to trigger cell death in the TA zone of (ileal) small intestinal crypts. However, cell death in the TA zone per se does not appear to be sufficient to permanently disrupt crypt homeostasis without the addition of a DR5-targeting mAb. Thus it seems plausible that the result of the perturbed GI epithelial cell homeostasis could result from TA zone injury that sensitize Lgr5+ cells to DR5-targeting. Further model development is required to address those observations in depth and elucidate differences in death receptor signaling between normal and stem cells in the GI tract. Our reported data clearly stand in contrast to the majority of preclinical findings that suggest targeting of DR5 is essentially non-toxic to normal cells and tissues in combination with DNA damaging chemotherapy. One potential reason for this is that assessment of toxicity following treatment with TDRA in mice is difficult due to inherent differences between the genome of mice and humans with respect to genes that influence TRAIL-signaling. For example the mouse genome harbors only a single death-inducing TRAIL receptor gene in DR5 in contrast to humans who have two in DR4 and DR5. Subsequently mouse models would not be useful to address potential toxicity of DR4 targeting that may potentially manifest in a different manner from that of DR5-targeting. Thus it is important to stress that the mouse model presented here limits itself to assessing toxicity following selective targeting of DR5 due to the mentioned differences. It is unclear to what extent any of the clinically relevant DR5 mAb's cross-react with mouse DR5 and toxicity assessment and the establishment of a LOAEL in rodent models may not be possible for some TDRA.

To the best of our knowledge MD5-1 is the only reported antibody that targets mouse DR5 in an agonistic manner. A recent study showed that the combination of Conatumumab or MD5-1 with recombinant APO2L/TRAIL greatly improved anti-tumor responses (30). Both Conatumumab and MD5-1 require cross-linking to exhibit anti-tumor activity alone (19,31) but this requirement is largely abrogated by the addition of APO2L/TRAIL, potentially as a result of higher order of DR5 clustering on the cell surface by the ligand-antibody combination (30). This study also showed that the co-administration of MD5-1 (at a two hundred times higher dose than the one used for the majority of the experiments in this study) together with APO2L/TRAIL triggered a gross and histologically similar GIT to what was observed here when MD5-1 was combined with CPT-11 and 5-FU. A proposed reason for the GIT was the abundant expression of DR5 in the GI tract of mice, a characteristic of the mouse alimentary tract that is shared with that of primates and humans as well, indicating a translational nature of this particular finding.

It is unlikely that the severity of toxicity observed in our animal model will translate directly to clinical observation for a number of reasons. For example, the mice in this study were given a single (high-dose) bolus of chemotherapy whereas 5-FU is administered as a cycled 46 hr intravenous infusion in patients, a schedule that is challenging to mimic in laboratory animal models. Subsequently, no dose-adjustment would be possible in our model nor did the mice receive supportive care to the same extent that is typical for cancer patients undergoing chemotherapy treatment. Unfortunately, chemotherapy-induced GIT is a common dose-limiting, costly and quality-of-life prohibitive toxicity and in contrast to pre-clinical animal models, accurate quantitative assessment of GIT remains difficult in patients. Estimation of GIT to such treatment is largely based on the presentation of symptoms that may not correlate well with the extent of GI injury or its localization in the GI-tract (32). Although MD5-1 can readily generate significant anti-tumor responses in several pre-clinical models (19,33) (Fig.5F and N), such responses have not been observed in clinical trials where DR5 targeting mAb's are combined with chemotherapy. Therefore the potential for toxicity would be considered in relation to the limited improvement of the response that may be expected when TDRAs are integrated into current clinical chemotherapy protocols. Furthermore our data suggests that chemotherapeutics that inflict DNA damage may non-selectively lower the threshold dose to DR5-targeting mAb's in both normal and malignant tissues or perhaps even preferentially do so in normal cells that harbor an intact P53-response.

Based on our gained knowledge of the DDR pathways involved in mediating toxicity following targeting of TRAIL-death receptors in combination of chemotherapy, we aimed to assess if targeting molecules within the DDR-pathway could modulate GIT. Molecular targeting of Chk2 has previously been proposed as a method to improve therapeutic indices for DNA-damaging modalities such as radiotherapy (34). To our surprise we found that mice lacking the Chk2 gene were highly resistant to toxicity inflicted by CPT-11/MD5-1 but not 5-FU/MD5-1 (Fig.6A and B). Furthermore, pharmacologic targeting of Chk2 kinase function using a 2-arylbenzimidazole Chk2-inhibitor (CI-II) (23) reduced toxicity following CPT-11/MD5-1(Fig. 6D). Although mice devoid of Chk2 were resistant to lethal GIT inflicted by CPT-11/MD5-1 they were not protected from lethal GIT inflicted by CPT-11 alone following dose-escalation (data to be published elsewhere). This finding indicates that the combined agonist targeting of the death receptor DR5 in the face of DNA damage inflicted by the chemotherapeutic CPT-11, triggers GIT through a molecularly distinct mechanism from that observed following DNA damage alone.

Expression data indicates that in contrast to 5-FU, CPT-11 downregulates the expression of Survivin in a chk2-dependent manner (Fig. 6E). Survivin is an established negative regulator of death receptor stimuli (35) suggesting that selective upregulation of Survivin following Chk2-targeting can prevent GIT following CPT-11/MD5-1. Subsequently we addressed the possibility that targeting of Chk2 could help improve therapeutic indices of CPT-11/MD5-1 when treating colorectal cancer. Indeed, in syngeneic grafts of the mouse colorectal cancer cell line p53dmc/Ras/Myc we were able to show that combining CI-II with CPT-11/MD5-1 resulted in reduced GIT and at the same time improved the tumor response rate (Fig. S9A-B). Our data indicate that this type of strategy may not only improve the tumor response rate to treatment but also indirectly translate to increased efficacy as a DLT is prevented that otherwise may result in the discontinuation of treatment.

Chemotherapy currently remains the mainstay in oncology and additional knowledge of how such modalities influence the response to DR5-targeting could help shape strategies that minimize toxicity to normal tissues. The potential of toxicity should also be considered when new combination therapies including DR5 agonists are designed since our data indicate that DR5-targeting can augment dose-limiting toxicities of conventional chemotherapy.

Abbreviations

5-FU

5-Fluorouracil

ADCC

antibody-dependent cell mediated cytotoxicity

AE

adverse events

AUC

area under the curve

CBC

crypt basal columnar

Chk2

cell cycle checkpoint kinase 2

C/M

CPT-11/MD5-1

CPT-11

irinotecan

CRC

colorectal cancer

DcR

decoy TRAIL receptor

DR5

TRAIL death receptor 5

DSB

DNA double-strand breaks

DLT

dose-limiting toxicity

F/M

5-FU/MD5-1

GI

gastrointestinal

GIT

gastrointestinal toxicity

HDACi

histone deacetylase inhibitors

IAP

inhibitor of apoptosis protein

IV

intravenous

mAb

monoclonal antibodies

PFS

progression-free survival

PI

propidium iodide

SSB

DNA single strand breaks

SRY

sex-related protein Y

TA

transit amplifying

TI

therapeutic index

TRAIL

Tumor necrosis factor-related apoptosis inducing ligand

TDRAs

TRAIL death receptor agonists