The combination of AZD0156 and SN38 affects proliferation and cellular regrowth in CRC cell lines
The IC50 across 4 CRC cell lines evaluated was close to 5 uM, though at these high doses there is concern for off-target effects of closely related kinases [4] (data not show). As anticipated, treatment with single-agent AZD0156 at a dose of 50 nM or 100 nM was not associated with a significant decrease in proliferation as compared to vehicle control across the 12 cell lines evaluated by IncuCyte assay (Fig. 1 and Supplemental Figs. 1 and 2, Additional File 1).
The four CRC cell lines with the best combination effect across the 12 tested (HCT8, RKO, LOVO, and HT29) were selected for further in vitro assessments. The combination of AZD0156 (100 nM) and SN38 (10 nM) resulted in decreased proliferation as compared to single-agent AZD0156 across all cell lines (HCT8 P < 0.0001, RKO P = 0.0005, LOVO P < 0.0001, HT29 P = 0.0002). When compared to single-agent SN38, the decrease in proliferation of the combination was not statistically significant, though a trend was noted in the HCT8 and LOVO cell lines (Fig. 1). This trend was also observed when the lower dose of AZD0156 (50 nM) was combined with SN38 (10 nM) (Supplemental Fig. 1, Additional File 1).
In the CRC cell lines, HCT8, RKO, LOVO and HT29, a decrease in percent confluence was observed by clonogenic assay following treatment with the combination of AZD0156 and SN38 followed by regrowth for 72 h. This decrease was statistically significant at the 50 nM dose of AZD0156 in combination with SN38 when compared to single-agent AZD0156 50 nM in the HCT8 (P = 0.0001), RKO (P < 0.0001) and HT29 (P < 0.0001) cell lines. There was a statistically significant decrease in percent confluence following treatment with this same combination when compared to SN38 in the HCT8 (P < 0.0001) and HT29 (P < 0.0001) cell lines. Similar patterns were seen at the 100 nM dose of AZD0156 in combination with SN38 when compared to single-agent AZD0156 (HCT8 P = 0.0002, RKO P < 0.0001, HT29 P < 0.0001), and single-agent SN38 (HCT8 P = 0.0002, RKO P < 0.0001, LOVO P = 0.0204, HT29 P < 0.0001) (Fig. 2).
The combination of AZD0156 100 nM and SN38 demonstrated synergy according to the Bliss Additivity model in the HCT8, LOVO, a HT29 cell lines, with values of 1.20, 1.38, and 1.68, respectively. A value of 1.04 was documented in the RKO model consistent with an additive interaction. When statistical analysis was performed as per Demidenko E, et al. [12], none of the 4 cell lines demonstrated statistically significant synergy (HCT8 P = 0.1213, RKO P = 0.9849, LOVO P = 0.0715, HT29 P = 0.6363), with findings more consistent with an additive effect of the AZD0156 and SN38 combination (Supplemental Fig. 3, Additional File 1).
Increased G2/M arrest is observed in CRC cells upon exposure to the combination of AZD0156 and SN38 in CRC cell lines
In the CRC cell lines, significant effects on the cell cycle were not observed upon treatment with AZD0156 as a single-agent. Significant decreases in G1 phase cells were noted in three of four cell lines exposed to SN38, as compared to untreated HCT8, RKO, and HT29 cells (P < 0.0001 for each). This corresponded to increased G2/M arrest, with statistically significant differences observed in the RKO (P = 0.0005) and HT29 (P = 0.0192) cell lines. When AZD0156 was added to SN38, further decreases in G1 phase cells were observed in cell lines with AZD0156 at 50 nM (HCT8 P < 0.0001, RKO P < 0.0001, LOVO P = 0.0089, HT29 P < 0.0001) and 100 nM (HCT8 P < 0.0001, RKO P < 0.0001, LOVO P = 0.0039, HT29 P < 0.0001) when compared to AZD0156 at respective doses as a single agent. The proportion of cells in G2/M phase increased accordingly with the combination of SN38 and AZD0156 at 50 nM (HCT8 P = 0.0384, RKO P < 0.0001, LOVO P = 0.0020, HT29 P = 0.0035) and 100 nM (HCT8 P = 0.0129, RKO P < 0.0001, LOVO P = 0.0025, HT29 P = 0.0012) (Fig. 3).
Effectors of DNA Damage response are activated upon treatment with SN38 and decreased upon combination with AZD0156 in CRC cell lines
P-RAD50 expression increased upon exposure to SN38 as compared to vehicle and single-agent AZD0156 treated samples across cell lines and at both 24 and 72 h, indicating induction of DNA damage repair pathways. A modest decrease in P-RAD50 expression was observed with the addition of AZD0156 to SN38 as compared to SN38 alone, indicating suppression of ATM signaling. This finding was most pronounced at the 24 h time point. Similar findings were observed with P-CHK2, a down-stream effector of ATM, with increase upon SN38 exposure and decrease in combination with AZD0156, in the majority of models, though in the HT29 model, P-CHK2 increased with the AZD0156 dose of 100 nM at the 24 h time point. In addition, the CHK2 protein was not detected in HCT-8 cells, as previously described [25]. γH2AX, a marker of DNA double-strand breaks, increased in cells exposed to SN38, and this was preserved with the addition of AZD0156. An increase in P-HH3 as a marker of G2/M arrest was also observed upon exposure to SN38 alone and in combination with AZD0156 in the LOVO and HT29 cell lines at 72 h (Fig. 4).
The addition of AZD0156 to irinotecan-based chemotherapy leads to tumor growth inhibition in CRC PDX models
In the CRC026 PDX model (NRAS Q61K MT), doses of irinotecan and 5FU were decreased from 15 mg/kg to 7.5 mg/kg and 60 mg/kg to 30 mg/kg, respectively, on day 11 of treatment due to the sensitivity of the model to these single-agents. Following these dose adjustments, a clear effect of the combination of AZD0156 and irinotecan as compared to single-agent AZD0156 (P < 0.0001) was observed, with evidence of tumor regression. Both single-agent irinotecan and the triplet of AZD0156, irinotecan, and 5FU were associated with significant tumor growth inhibition as well (P < 0.0001 for both). Interestingly, the addition of 5FU to the combination did not further enhance this effect, though tumor growth inhibition was not statistically significantly greater in the AZD0156/irinotecan arm as compared to the triplet arm. A trend toward tumor growth reduction was noted when comparing the combination of AZD0156 and irinotecan to single-agent irinotecan, though the difference was not statistically significant. Also of note in this model, a non-statistically significant increase in tumor volume was noted in the single-agent AZD0156 arm as compared to vehicle control (P = 0.1021). This pattern was not observed across other PDX models (Fig. 5A). No significiant findings were noted in additional treatment arms evaluated (Supplemental Fig. 4, Additional File 1). Tumor growth inhibition in the AZD0156/irinotecan combination arm was 94.7% as compared to 54.0% in the AZD0156/irinotecan/5FU triple combination arm.
In the CRC102 model (KRAS G12V MT), the AZD0156/irinotecan doublet (P = 0.0203) and AZD0156/irinotecan/5FU triple combination (P = 0.0090) were again associated with a statistically significant reduction in tumor volume when compared to single-agent AZD0156. A trend toward tumor growth reduction was noted in the AZD0156/irinotecan doublet (P = 0.0358) and AZD0156/irinotecan/5FU triplet as compared to single-agent irinotecan, but this was not statistically significant (Fig. 5A). No significiant findings were noted in additional treatment arms evaluated (Supplemental Fig. 4, Additional File 1). Tumor growth inhibition in the AZD0156/irinotecan/5FU triple combination arm was 51.1%, as compared to 43.5% in the AZD0156/irinotecan combination arm in this model.
Specific growth rate (SGR) analysis was performed for both models [26] (Fig. 5B). In the CRC026 model, a significant difference in specific growth rate was observed between the single-agent AZD0156 and the AZD0156/irinotecan doublet arm (P < 0.0001), as well as the triple combination arm (P = 0.0005) and single-agent irinotecan arm (P = 0.0007). When the doublet and triplet combination were compared to single-agent irinotecan, the differences were not significant. In the CRC102 model, no statistically significant differences in specific growth rate were noted.
In two additional PDX models, CRC001 (KRAS G12D MT) and CRC042 (KRAS G13D MT), less pronounced effects on tumor growth inhibiton were observed with the addition of AZD0156 to chemotherapy (Supplemental Fig. 5, Additional File 1). However, in the CRC042 model, the specific growth rate of the AZD0156/irinotecan doublet arm was lower than both the AZD0156 and single-agent irinotecan arms, and the SGR of triple combination arm was similarly lower than that of both the AZD0156 arm and irinotecan arm (Supplemental Fig. 5, Additional File 1).
DNA damage response is affected by chemotherapy alone and in combination with AZD0156 in CRC PDX Models
The pharmacodynamic effects of AZD0156 and irinotecan in CRC026 and CRC102 PDX models were evaluated by immunofluorescence assessment of γH2AX and P-RAD50 in tumor tissues at the end of treatment [27]. In the CRC026 model, this corresponds to 4 days following the last dose of irinotecan and 5FU, and 1 h following the last dose of AZD0156. In the CRC102 model, this corresponds to 6 days following the last dose of irinotecan and 5FU, and 1 h following the last dose of AZD0156. In the CRC026 model, a significant increase in γH2AX was observed in tumors from mice treated with the combination of irinotecan and AZD0156 as compared to single-agent AZD0156 (P = 0.0009) or irinotecan (P = 0.0005), consistent with increased accumulation of double-strand breaks in the setting of ATM inhibition (Fig. 6A). A similar trend was seen in CRC102 model, though the difference was only statistically significant when the combination arm was compared to control (Fig. 6B). In both models, P-RAD50 expression was decreased in tumors from mice treated with AZD0156 in combination with irinotecan, consistent with a decrease in DDR pathway activation. This difference was not significant in the CRC026 model, but was when compared to both AZD0156 (P = 0.0010) or irinotecan (P = 0.0265) as single-agents in the CRC102 model (Fig. 6).
Alterations in prototypical DNA damage repair genes are variable across in vivo and in vitro models evaluated
Given the potential for improved response to ATM-targeted therapy in the setting of an underlying DDR alteration, prototypical DDR mutations were evaluated across all models utilized in this work (Supplemental Table 1, Additional File 1). All models evaluated had at least one mutation in a DDR gene, but there was no consistent mutation pattern across models. The HCT116 cell line was the only model evaluated that harbors an ATM mutation and was the only cell line evaluated in initial proliferation studies in which an antiproliferative effect was observed with AZD0156 as a single-agent. This finding was limited to the 100 nM dose. Interestingly, the cell lines in which the antiproliferative effect of SN38 was most enhanced by AZD0156, HCT8 and LOVO, both harbor CHEK2 mutations. The HT29 cell line, which had less pronounced, though still statistically significant enhancement of antiproliferative chemotherapy effects with AZD0156, has an TP53 mutation, with no other DDR alterations.
Though the DDR mutation profile of the PDX models evaluated was overall similar, it is possible that key variations impacted treatment effect. In the CRC026 model, in which the most pronounced tumor growth inhibition was observed with AZD0156 and irinotecan, the only unique alteration identified was an NRAS Q61K mutation. The CRC102 model harbors a TP53 mutation, which may explain the altered anti-tumor activity associated with single-agent chemotherapy, and improved anti-tumor activity with the AZD0156/irinotecan/5FU triplet. Interestingly, the CRC042 model, in which the tumor specific growth rate was significantly decreased with the addition of AZD0156 to irinotecan, does harbor an ATR mutation.