Signaling interactions of rapamycin combined with erlotinib in cervical carcinoma xenografts

Aberrant signaling pathways are responsible for the features of cancer: unregulated growth, suppression of apoptosis, tumor angiogenesis, invasion, and metastasis. Consequently, a large effort is being made to develop novel agents that selectively target signaling elements for the treatment of cancer. Despite some clinical successes, response rates to single agents usually occur only in a minority of cancer patients and are typically incomplete. Given that cancers progress through multiple genetic abnormalities and the complexity of the interdependent signaling pathways, a combination of molecular directed drugs seems legitimate. The challenge would be then to rationally develop a mix of signaling inhibitors with enhanced anticancer activity and acceptable toxicity.

Excessive signaling through epidermal growth factor receptor (EGFR) due to receptor overexpression, autocrine stimulation, or activating mutations occurs in an extensive range of tumors and correlates with worse prognosis, thus justifying the efforts directed to inhibit its activity (1–3). Clinical trials using EGFR inhibitors, such as erlotinib, have shown partial responses or stabilization in various types of EGFR-positive tumors, with relatively mild side effects (1, 2).

The mammalian target of rapamycin (mTOR) plays critical roles in growth, differentiation, proliferation, migration, and survival (4, 5). It is an atypical serine-threonine kinase member of phosphatidylinositol kinase–related kinases family that acts as a sensor for the level of growth factors, nutrients, ATP, and oxygen (4–9). In response to these stimuli, mTOR regulates the activity of p70S6K and 4E-BP/PHAS, which have essential roles in ribosome biogenesis and cap-dependent translation via S6 ribosomal protein and eIF-4E, respectively (5, 6). mTOR also regulates other cellular processes, including cell-cycle proteins and RNA polymerases, protein phosphatases, hypoxia-inducible factor-1α, and vascular endothelial growth factor (VEGF) signaling (4–6, 8–10). Progress has been made recently understanding the interactions of mTOR, particularly within the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (PKB) pathway, with the identification of several key modulators, including TSC1/2 and Rheb (4, 5, 11, 12). Rapamycin, a natural macrolide antibiotic, is an exquisitely specific inhibitor of mTOR downstream signaling (4, 5, 8). It has antiproliferative effects against a broad range of tumor cell lines and antiangiogenic properties that translated into tumor growth inhibition in animal models (4, 12–16). This prompted the development of several analogues for testing as anticancer drugs (4).

Growth factors induce activation of tyrosine kinase receptors and consequently can signal through both PI3K/PKB/mTOR and mitogen-activated protein kinase (MAPK) pathways (2, 3, 5, 17). These pathways have multiple cross-connections at several levels, such as tuberin, p70S6K, and 4E-BP1 as illustrated in Fig. 1 (5, 10, 13, 17–29). p70S6K is believed to be the principal kinase for S6 ribosomal protein, and its activity correlates well with S6 phosphorylation (5, 6, 25). Besides experimental studies, immunohistochemical analysis of clinical samples is consistent with the dual activation model for p70S6K (30). The activation of p70S6K is triggered by the hierarchical phosphorylation of at least eight sites (5, 6, 31). Phosphorylated extracellular signal-regulated kinase (ERK) seems to be involved especially in the initial steps of activation, whereas mTOR-sensitive phosphorylation sites count for full activation of p70 (23–25, 31). S6 ribosomal protein represents the main substrate of the two isoforms of p70S6K (S6K1 and S6K2), and its activation is therefore sensitive to rapamycin. Based on these considerations, we hypothesized that simultaneous inhibition of mTOR and MAPK pathways would result in enhanced inhibition of S6 activation, in addition to other synergistic effects expected when targeting two major pathways that display extensive cross-talk. Furthermore, there is an extensive literature reporting the antiangiogenic effects of rapamycin and EGFR inhibitors (2, 10, 12–14, 32–34). VEGF, the key molecule for angiogenesis, is mainly regulated through hypoxia-inducible factors. Although hypoxia-inducible factor is controlled differently during normoxia versus hypoxia, dual regulation from PI3K/PKB/mTOR and MAPK is required at all times (7, 9, 10, 13, 17, 27–29, 31, 35). Up-regulation and modulation of hypoxia-inducible factor protein levels is considered to be achieved through PKB/mTOR, whereas its transactivation relies essentially on MAPK, which recruits the coactivators (9, 28, 29). These interactions were examined in a series of human cervical cancer xenografts.

Erlotinib was obtained from OSI Pharmaceuticals, Inc. (Uniondale, NY) and dissolved in DMSO at 5 mg/mL freshly for each treatment. Rapamycin was purchased from Calbiochem (San Diego, CA). EGF was from Sigma (St. Louis, MO), and UO126 was from Cell Signaling Technology (Beverly, MA). The following antibodies were purchased from Cell Signaling Technology: phosphorylated ERK Thr²⁰²/Tyr²⁰⁴ monoclonal E10 for immunofluorescence staining; phosphorylated ERK polyclonal for Western blots; phosphorylated S6 ribosomal protein polyclonals to Ser²³⁵/Ser²³⁶, Ser²⁴⁰/Ser²⁴⁴; polyclonal phosphorylated PKB Ser⁴⁷³; and polyclonal cleaved caspase-3. Rabbit monoclonal anti-phosphorylated p70S6K Thr³⁸⁹ was from Epitomics (Burlingame, CA). Anti-α-tubulin was from Oncogene (Cambridge, MA), CD31 was from BD Biosciences (San Diego, CA), and Ki-67 (clone MIB-1) was from DAKO (Glostrup, Denmark).

For testing the antiproliferative effects of the drugs, SiHa cells were cultured in 96-well plates at 3,000 per well in 100 μL complete medium. The drugs were added in various concentrations, and after 48 hours, according to the manufacturer's instructions, the viable cell number was determined using a colorimetric method based on the cellular bioreduction of the tetrazolium compound 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) into a soluble colored formazan product (CellTiter 96 Aqueous One Solution Cell Proliferation Assay, Promega Corp., Madison, WI).

The human cervical squamous cell carcinoma cell lines SiHa, Me180, and CaSki were obtained from the American Type Culture Collection (Manassas, VA) and cultured in α-MEM, McCoy's medium, and RPMI 1640, respectively, all supplemented with 10% fetal bovine serum. Cells were plated in 100-mm dishes at appropriate concentration and treated for the indicated periods. The cells were washed twice in ice-cold PBS, and 650 μL of lysis buffer [50 mmol/L HEPES (pH 8.0), 10% glycerol, 1% Triton X-100, 150 mmol/L NaCl, 1 mmol/L EDTA, 1.5 mmol/L MgCl₂, 100 mmol/L NaF, 10 mmol/L NaP₂O₇, 1 mmol/L Na₃VO₄, supplemented with protease inhibitor cocktail from Roche Diagnostics, Mannheim, Germany] were added to each dish; lysis was conducted for 1 hour on ice. Equivalent amounts of protein (assayed with bicinchoninic acid protein assay from Pierce PerBio, Rockford, IL) were separated by SDS-PAGE gels. Proteins were transferred to polyvinylidene difluoride membranes (Millipore, Bedford, MA) and probed with the appropriate antibodies according to the manufacturer's instructions. Detection was conducted using enhanced chemiluminescence kit (Amersham Biosciences, Buckinghamshire, United Kingdom).

Animal experiments were done according to institutional guidelines for animal welfare. Cell suspension (0.1 mL, containing ∼5 × 10⁵ cells) was injected into the left gastrocnemius muscle of severe combined immunodeficient mice. After ∼2 weeks, the tumors became visible and the tumor plus leg diameter was measured between 7.5 and 8.5 mm. At this point, the mice were randomly assigned to four groups of five mice. The treatments included rapamycin given twice weekly at 2.5 mg/kg and erlotinib given daily at 10 mg/kg. Both drugs were dissolved in DMSO and given by i.p. injection. For the combined treatment, rapamycin was given first and erlotinib 1 hour later. The control group was injected with the same amount of DMSO as was used as a vehicle for the double-treated animals. Animal weight and tumor measurements were made every 2nd day. At the end of treatment, the mice were killed and the tumors were excised and processed as follows: each tumor was dissected from the surrounding tissues and cut into several pieces (approximately 3 × 3 × 3 mm), which were randomized and split in three groups. Pieces from first group were placed in OCT embedding medium (Miles, Inc., Elkhart, IN) and snap frozen in liquid nitrogen; pieces from the second group were fixed in formalin for 24 hours and paraffin embedded. The remaining tissues were lysed 1 hour on ice in 1.5 mL lysis buffer as for the cell lysates.

To examine the short-term effects of drugs, 15 SiHa tumors were allowed to grow 13 to 14 mm in tumor plus leg diameter. Mice were randomized in four groups: three for controls and four in each rapamycin, erlotinib, and rapamycin plus erlotinib groups. DMSO and drugs were given as above; tumors were excised 60 minutes later and processed as above.

Plasma VEGF was measured using an ELISA assay (R&D Systems, Minneapolis, MN) that recognizes both human VEGF₁₆₅ and VEGF₁₂₁ and, at 50 ng/mL, does not cross-react with mouse VEGF. To allow for effects of the individual tumor sizes on the plasma levels of VEGF, we expressed these as (pg/mL) / mm³ tumor volume × 100 in addition to the actual plasma concentrations.

Serial sections from the SiHa tumors were cut, and one was stained with H&E for selecting the tumoral areas on which the further image analysis was to be done. The other sections were labeled with primary antibodies against (a) phosphorylated S6 (Ser²³⁵/Ser²³⁶) and phosphorylated ERK, (b) CD31 and cleaved caspase-3, and (c) Ki-67. Secondary antibodies used alone were control for nonspecific background. All sections were counterstained with 1 μg/mL 4′,6-diamidino-2-phenylindole (DAPI) to outline the nuclear area.

Tissue sections were examined using a fluorescence image analysis system (Microcomputer Imaging Device, Imaging Research, Inc., St. Catharines, Ontario, Canada) equipped with a cooled CCD camera (Quantix, Photometrics, Tucson, AZ). The analysis included the entire viable tumoral area identified on the slide using a scanning autostage to create composite images of individual fields at 20× (2.24 mm²) as described previously (32). The tumoral map in which the actual measurements were made was initially determined using the H&E sections and then carefully adjusted on the composite DAPI images. For each marker, we measured at least two variables: the intensity of the signal (integrated absorbance) and the percentage of positive stained area, based on the binary images.

To determine activated ERK, we used both phosphorylated ERK and DAPI (binarized) images. By overlaying these, we were able to identify and count phosphorylated ERK-positive nuclei (a value that was normalized to the total number of nuclei in the area given by DAPI), whereas cytoplasmic phosphorylated ERK was obtained by subtracting nuclear phosphorylated ERK area from the total area that stained positive for phosphorylated ERK (total phosphorylated ERK) and normalizing it to the total area of the map. On the phosphorylated S6 binarized images, a closing operation was done (dilation followed by erosion) to fill in the unstained nuclear area (S6 being a cytoplasmic marker), so an estimate of the area of the cells that stained positive for phosphorylated S6 was obtained and that was normalized to the tumoral map area.

The cleaved caspase-3 and Ki-67 images were also binarized. To count the number of positive objects for the cellular markers, we used size criteria to eliminate the background and a mean value of a tumoral cell/nucleus area to resolve the clusters. Microvascular density was measured using CD31 staining. On the binarized image, a closing operation was done to fill the vascular lumen. For microvessel density (MVD), we introduced additional size and shape criteria to obtain the number and size distribution of the vessels.

The markers of interest were phosphorylated S6, phosphorylated ERK, cleaved caspase-3, Ki-67, and MVD. Phosphorylated S6 and phosphorylated ERK were evaluated using both integrated absorbance values and percentage of positive area. The other markers were expressed as percentage of positive area. Phosphorylated ERK was also evaluated in cytoplasm as well as in the nucleus.

Because for each marker there were several sections analyzed within each tumor, the comparison between the treatment arms was made using the random effects model (a simple ANOVA test assumes that the records are independent; therefore, it is not appropriate for this data set). For each marker, the normality assumption of the residuals was visually inspected. The percentages of positive area for phosphorylated S6 and phosphorylated ERK were log transformed to fulfill the normality assumption of the residuals. The mean value for each marker and for each treatment was expressed as the least square means calculated using the model. When the marker was log transformed (phosphorylated S6 and phosphorylated ERK percentage of positive area), the means obtained from the model were back transformed to reflect the usual scale.

The analysis of the tumor growth was done using mixed models, considering a spatial correlation between the measurements for the same mouse over time.

Continuous exposure of SiHa cells for 48 hours resulted in nonsignificant growth inhibition using a range of rapamycin and erlotinib concentrations, whereas the combination produced statistically significant inhibition determined by MTS assay (P = 0.01; Fig. 2). This was associated with a decrease in the percentage of cells in S phase. There was no increase in apoptosis, determined by flow cytometric analysis of cleaved caspase-3, under any of the in vitro treatments tested (data not shown).

Whereas MAPK is responsible mainly for relieving an inhibitory conformation of p70S6K, PI3K/PKB/mTOR ensures its complete activation (22, 26). Furthermore, phosphorylation of S6 at Ser²³⁵/Ser²³⁶ sites is also the result of combined mTOR and MAPK signaling (29). As shown in Fig. 2, phosphorylation of ERK, Akt, p70S6K, and S6 was detectable in all three cell lines when these were grown in the presence of 10% fetal bovine serum. As expected, S6 phosphorylation (Ser²³⁵/Ser²³⁶) was suppressed by rapamycin at all concentrations tested, whereas treatment with the MAPK/ERK kinase inhibitor U0126 resulted in decreased phosphorylated ERK (Fig. 2). Although erlotinib also decreased ERK activation in all three cell lines, the effect on phosphorylated S6 was much less. Following EGF stimulation, both rapamycin and erlotinib partially suppressed phosphorylated S6 (Ser²³⁵/Ser²³⁶), whereas the two drugs combined produced almost complete inhibition in all three cell lines. Similar results were seen combining the MAPK/ERK kinase inhibitor UO126 with rapamycin. However, unlike erlotinib, UO126 alone did not significantly reduce phosphorylated S6 in EGF-stimulated cells nor had any effect on PKB/Akt phosphorylation. The activation of the other pair of serine sites on S6 (Ser²⁴⁰/Ser²⁴⁴) exhibited a pattern similar to Ser²³⁵/Ser²³⁶ but showed greater sensitivity to rapamycin and erlotinib, consistent with these sites showing greater dependence on PI3K/PKB/mTOR relative to MAPK signaling (29).

Effects of the Combined Therapy on Tumor Growth. As shown in Fig. 3, increased growth delay was seen in SiHa tumors treated with the combination of drugs when compared with either the control group (P < 0.0001) or single drug-treated animals (P = 0.0002 versus erlotinib; P = 0.01 versus rapamycin). Rapamycin alone had a statistically significant effect when compared with the controls (P = 0.002), whereas for erlotinib a trend was present (P = 0.08). There was no significant difference between rapamycin and erlotinib when considering the tumor progression. In terms of drug toxicity, although erlotinib-treated mice presented weight loss (between 7% and 13%), in the group that also received rapamycin the animals displayed an apparent overall healthier appearance.

In contrast to SiHa, treatment with erlotinib produced no significant effect on Me180 or CaSki tumor growth, but the combination of erlotinib plus rapamycin produced significantly greater growth delay in Me180 xenografts compared with rapamycin alone (P = 0.02). As shown in Fig. 3, CaSki tumors were highly responsive to rapamycin, and there was no additional benefit with the addition of erlotinib.

Western Blots. In preliminary experiments, the acute effects on signaling pathways were tested in SiHa xenografts by treating quadruplicate groups of animals with 2.5 mg/kg rapamycin, 10 mg/kg erlotinib, or the combination followed by sacrifice after 1 hour. Treatment with erlotinib resulted in a large decrease in phosphorylated ERK. Although there was a decrease in phosphorylated S6 following rapamycin, which was more pronounced with the combination, this effect was much less than that seen following 1 hour of drug exposure in vitro (data not shown). In contrast, in the 19-day continuous dosing experiments, phosphorylated S6 (Ser²³⁵/Ser²³⁶) was almost undetectable in the rapamycin-treated SiHa tumors, and any additional effect of erlotinib was not detectable by Western blot. Phosphorylated ERK suppression seemed similar in erlotinib alone treated mice and in the combined treatment group. Although ineffective in tumor growth delay, erlotinib produced decreased phosphorylated ERK in Me180 xenografts, associated with an enhanced effect of rapamycin on phosphorylated S6. Whereas CaSki cells grown in tissue culture showed decreased phosphorylated ERK following treatment with erlotinib (Fig. 2), this effect was not seen in the CaSki xenografts, and there was no enhancement of rapamycin on S6 phosphorylation (Fig. 3).

Immunofluorescence Staining. Tissue sections from the SiHa 19-day continuous treatment experiment were double immunostained for fluorescence image analysis of phosphorylated S6 (Ser²³⁵/Ser²³⁶) and phosphorylated ERK. Because these measurements were made only in the viable tumor areas, unlike the Western blot analyses, they exclude necrotic or nontumoral tissue. Decreased integrated absorbance values and positive pixel areas for phosphorylated S6 were found following treatment with either rapamycin or erlotinib, and this effect was greater with the combination of drugs (Fig. 4A). All of the differences were statistically significant (Fig. 4B). Total phosphorylated ERK, expressed as percentage of positive stained area, was less when tumors were treated with erlotinib compared with the controls and further decreased with the combination. By using a DAPI nuclear mask, we were also able to measure nuclear phosphorylated ERK and found that this followed the same distribution pattern as the total activated protein (Fig. 4B), whereas erlotinib treatment did not decrease cytoplasmic phosphorylated ERK (data not shown).

The Ki-67 labeling index obtained for the SiHa viable tumoral areas showed a significant decrease following rapamycin treatment (P = 0.03) but not with erlotinib (P = 0.16). No significant differences in cleaved caspase-3 staining in tumor tissue were seen between the groups (data not shown).

The image analysis technique used to measure MVD is illustrated in Fig. 5A, and the results obtained from the SiHa continuous dosing experiment are shown in Fig. 5B. There was no significant effect using either of the single agents, but the combination of erlotinib and rapamycin resulted in a significant decrease in MVD (P = 0.012). To investigate the antiangiogenic effects of the drug combination further, we measured plasma VEGF levels following continuous dosing. As shown in Fig. 5C, there was a nonsignificant trend for lower VEGF in the erlotinib-treated group and a significant decrease in plasma VEGF following combined treatment (P = 0.017), consistent with the effects seen on MVD. We also tested plasma VEGF from the Me180 and CaSki groups of mice but found that this was not detectable above background.

In recent years, EGFR inhibitors and rapamycin analogues have been extensively tested against human solid tumors. Each class of agent has been found to be well tolerated and to produce clinical benefit in a wide range of tumor types. However, responses are typically seen only in a minority of patients and are often of short duration. Given the known interactions between signaling pathways downstream from EGFR and mTOR, we hypothesized that the combination of rapamycin plus erlotinib would show additive or synergistic effects in vivo and tested this in a series of human cervical squamous cell carcinoma cell lines. All three cell lines responded to rapamycin and erlotinib treatment in vitro with reduced phosphorylated S6 and phosphorylated ERK, respectively. SiHa and Me180 xenografts both showed enhanced tumor growth delay with the drug combination, associated with effects on signaling pathways that are consistent with our hypothesis.

It is still unclear what dictates the response to EGFR inhibitors because the status of the receptor itself is not predictive (1, 2). On the other hand, rapamycin and analogues were repeatedly found to do better whenever PKB is constitutively activated (4, 8, 12–15, 36). The cervical carcinoma xenograft models were used because the tumors express EGFR and activated PKB. As well as Western blotting, wide-field multicolor fluorescence image analysis was used to study effects in SiHa xenografts because this has several advantages in studying biological processes at the tumor tissue level (37). Relative to immunohistochemistry, it is more quantitative, and it allows colocalization of multiple markers in the same tissue section. Moreover, in contrast to obtaining data from “hotspots” or randomly chosen fields, the use of composite, tiled field images allows the analysis of large tissue areas, which accounts for the heterogeneity of the tissue itself and of the markers of interest. In the present study, phosphorylated S6 (Ser²³⁵/Ser²³⁶) and phosphorylated ERK were chosen as the appropriate biomarkers to evaluate erlotinib and rapamycin effects on intracellular signaling. Although their levels correlated well in Western blots versus computerized image analysis, the latter measurements were evidently more accurate as they quantify activated proteins exclusively within the viable tumoral tissue.

In SiHa tumors, rapamycin given together with erlotinib had a significantly greater effect on tumor growth when compared with single drug alone, associated with enhanced suppression of phosphorylated S6, lower VEGF plasma levels, and decreased MVD, consistent with pathway interactions shown in Fig. 1. For some signaling inhibitors, it is difficult to discern in vivo between cytostatic/cytotoxic and antiangiogenic effects (14, 34). The mTOR and MAPK pathways regulate protein translation through the common targets p70/S6 ribosomal protein and 4E-BP/eIF-4E (5, 6, 24, 25). However, despite the significant decrease in phosphorylated S6, it is unlikely that the tumor growth delay can be explained entirely based on impaired translation. In the SiHa xenografts, there was a significant decrease in proliferation measured by Ki-67 labeling after rapamycin administration but not with erlotinib. We did not see major changes in apoptosis assessed by caspase-3 cleavage following 19 days of continuous dosing but cannot exclude the possibilities that the drug combination produces increased apoptosis at an earlier stage of treatment or induces a caspase-independent mode of tumor cell death (14, 34, 38).

In contrast to SiHa, erlotinib treatment did not produce significant tumor growth delay in Me180 xenografts despite a reduction in phosphorylated ERK. However, erlotinib enhanced the effects of rapamycin on tumor growth as well as on S6 phosphorylation. This finding supports the idea that the combination of rapamycin and an EGFR inhibitor might have clinical activity against cancers that are refractory to EGFR inhibitors. In contrast, CaSki cells, which were responsive to erlotinib in vitro, showed no effects on tumor growth or phosphorylated ERK when treated with this agent in vivo. These tumors were very responsive to rapamycin in vivo, but there was no additional effect when this was combined with erlotinib, consistent with its lack of effect on ERK and S6 phosphorylation in vivo. Although highlighting the complexity of combination treatment protocols that incorporate multiple signal transduction inhibitors, overall, the results reported here support the further clinical investigation of rapamycin and EGFR inhibitor combinations in the treatment of cancer patients.

We thank Melania Pintilie for the statistical analysis, Trudey Nicklee, May Cheung, Pinjiang Cao, and James Ho for excellent technical assistance, and Ken Iwata (OSI Pharmaceuticals) for his support and helpful comments.