X-MAN™ HCT116 Parental (RRID : CVCL_HD91) and HCT116 PTEN^-/- (RRID : CVCL_HD92) cells were generated by Horizon from homozygous knock-out of PTEN by deleting exon 5 which encodes the active site of the protein in the CRC cell line HCT116 (Horizon Discovery www.horizondiscovery.com) (9). Isogenic cell lines HCT116, HK2–6, and HKE-3 were performed by Shirasawa’s group by KRAS specific targeting technique (10, 11). LS180, HT29, RKO, and SW480 cell lines were kindly provided by Dr. Federica Di Nicolantonio as previously described (11). Human foreskin fibroblasts (HFF), green fluorescent protein (GFP)-labeled fibroblasts, and BJ cell lines were kindly provided by Dr. Maurizio Fanciulli. CRC cell lines were routinely maintained in RPMI 1640 medium (Euroclone, Milan, Italy) supplemented with 10% inactivated fetal bovine serum (FBS) (Gibco, Life Technologies, California, USA), whereas HFF and BJ were routinely maintained in DMEM medium supplemented with 15% inactivated FBS. All the growth media were added with 2 mM L-glutamine (Gibco, Life Technologies, California, USA) and antibiotics (Penicinillin/Streptomycin) (Gibco, Life Technologies, California, USA); cells were grown in a humidified atmosphere with 5% CO2 at 37°C.

Normal fibroblasts (NF) were isolated from normal colon tissue, derived from patients surgically treated at Regina Elena Cancer Institute. The study was reviewed and approved by the ethics committee of the Regina Elena National Cancer Institute. Normal colon tissue was digested by 0.35% Collagenase type I (Gibco, Life Technologies, California, USA); DMEM medium supplemented with 15% inactivated FBS was added with 2X Antibiotic Antimycotic Solution (Sigma Aldrich, St. Louis, USA). Isolated fibroblasts were routinely maintained in DMEM medium supplemented with 15% inactivated FBS, added with 2 mM L-glutamine (Gibco, Life Technologies, California, USA) and antibiotics (Penicinillin/Streptomycin) (Gibco, Life Technologies, California, USA); cells were grown in a humidified atmosphere with 5% CO2 at 37°C.

Based on previous experience of our group (11), HFF-CM was obtained under standardized culture conditions: 1x10⁶ cells were seeded and after 24 h, cells were washed with PBS 1X and medium was replaced by serum-free medium (wo FBS); CM was collected after 72 h. Serum-free conditions were chosen based on the notion that production of soluble factors is tightly dependent on the type of growth media used; moreover, serum-free conditions are also necessary to characterize the pattern of expression of soluble factors released by fibroblasts (see below, paragraph 2.10), using specific ELISA and membrane-based arrays, since FBS contains basal levels of cyto/chemokines that could interfere with the detection of mediator(s) involved in the observed CM effects (12, 13).

For recombinant protein experiments, 1000 pg/mL of interleukin (IL)-8, IL-6 and monocyte chemoattractant protein (MCP)-1 were added in DMEM wo FBS. Recombinant human IL-8, IL-6, and MCP-1 were purchased from R&D Systems (R&DSystems, MN, USA).

For cell counting, Thoma chamber was used. All cell lines tested negative for mycoplasma contamination.

Trametinib (GSK1120212, Tram) was kindly provided by GlaxoSmithKline (Brentford, Middlesex, UK). Alpelisib (BYL719, Alp), dactolisib (BEZ235, Dact), and Geda (PF05212384) were purchased from Selleck Chemicals (Huston, TX, USA). MK-2206 (MK) was kindly provided by Merck and Co. (Kenilworth, NJ, USA). Alp, Geda, MK, and Tram were dissolved in DMSO as a 1 mM (Geda, MK and Tram) and 5mM (Alp) stock solution, dact was dissolved in DMF as a 1mM. All the drugs were stored at -20°C (Alp, Dact, MK, and Tram) or -80°C (Geda). Eve (RAD001) was kindly provided from Novartis Pharma (Basel, Switzerland) and was dissolved in 100% ethanol as a 10 mM stock solution and stored at -20°C.

The final concentration of drugs was obtained by dilution with culture medium.

Effects on cell growth after 72 h of different treatments were monitored by Crystal Violet assay, as previously described (10).

X-MAN™ HCT116 Parental and HCT116 PTEN^-/- cells and HFF cells were seeded alone or with a 1:1 ratio in 35 mm plates (Falcon BD, Oxford, UK) in RPMI 1640 (Euroclone, Milan, Italy) supplemented with 10% inactivated FBS (Gibco, Life Technologies, California, USA), as previously described. After 24 h, culture media were replaced with serum-free DMEM (Euroclone, Milan, Italy) containing different drug concentrations alone or in combination. After 72 h of treatment, cells were suspended in ice-cold RPMI 1640 (Euroclone, Milan, Italy) supplemented with 10% inactivated FBS (Gibco, Life Technologies, California, USA). The total amount of cells was counted using the Thoma chamber and the percentage of CRC and HFF cells were analyzed by flow cytometry analysis that allows to differentiate between HFF cellular populations, fluorescently labeled in comparison to unlabeled X-MAN™ isogenic HCT116 cells.

X-MAN™ HCT116 Parental cell line was transfected with either GFP-PTEN C124S or GFP-Q399STOP plasmids, which encode for mutant PTEN C124S and PTEN Q399STOP proteins, respectively (14). Stable transfection of plasmids was performed using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions.

Total RNA was prepared from cells using the RNA extraction kit, RNeasy Mini Kit (Qiagen, Hilden, Germany) as per the manufacturer’s instructions. Of total RNA, 1 μg was converted into single-strand cDNA using Superscript II (Invitrogen, Carlsbad, CA, USA) as per the manufacturer’s instructions. Quantitative real-time PCR (RT-qPCR) was performed with Fast SYBR^®Green quantitative PCR kit (Applied Biosystems, Foster City, CA, USA) for RPL19 (Forward primer sequence: 5′-CGGAAGGGCAGGCACAT-3′ and Reverse primer sequence 5′-GGCGCAAAATCCTCATTCTC-3′) and PIK3CD (Forward primer sequence: 5′-CCCACATGAAGAGGAACTGAGAT-3′ and Reverse primer sequence 3′- GGTTGGCAGGCTCAGTGACT-5′). Expression of PIK3CD was then normalized with RPL19.

Whole cell extracts were obtained by SDS lysis buffer, containing 20 mM TrisHCl (pH 7.4) and 2% SDS, with 1X phosphatase and protease inhibitors (Thermo Fisher Scientific, Rochester, NY, USA). Protein quantification, separation, and detection were assessed as previously described (11). Membranes were probed with the primary and secondary antibodies reported in Supplementary Materials .

Chip-Grade Protein A/G Magnetic Beads (Thermo Fisher Scientific, Rochester, NY, USA) were incubated with 2 μg of FRAP (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. Precleared beads were then incubated overnight at 4°C with 1 μg of protein. The immunoprecipitates were collected and after two washes in CHAPS buffer (containing 5 mM EDTA, 0.3% CHAPS, 50 mM Tris HCl (pH 7.4), 150 mM NaCl, 10 mM MgCl₂, 1mM KCl, and 1X phosphatase and protease inhibitors (ThermoFisher Scientific, Rochester, NY, USA)) re-suspended in 35 µl of the same buffer and 1X Ladder buffer. The immune complexes and 20 μg of protein total cell extract were analyzed by Western blot analysis as described above.

5x10⁴ cells were seeded on 22x22 mm coverslips pretreated with 2% type A from porcine skin gelatin (Sigma Aldrich, St. Louis, USA); after 24 h, culture media were replaced by wo FBS DMEM (Euroclone, Milan, Italy) or HFF CM containing the indicated concentration of drug. After 24 h from treatments, cells were fixed in 4% formaldehyde in 1X PBS for 10 min, permeabilized in 1X PBS containing 0.5% Triton X-100 for 5 min at room temperature. Coverslips were incubated with primary and secondary antibodies as previously described (11). The used antibodies are reported in Supplementary Materials . Images were taken using Leica DMi8 microscopy (Leica, Solms, Germany) and Zeiss LSM 880 Confocal Microscope (Carl Zeiss, Oberkochen, Germany) equipped with ×63 lenses and operated by Zen 09 black software (Carl Zeiss).

5x10⁶ HCT116 Parental cells were seeded into 100 mm dishes (Greiner bio-one, Kremsmünster, Austria) in RPMI 1640 (Euroclone, Milan, Italy) and after 24 h from plating cell culture media were replaced by wo FBS DMEM (Euroclone, Milan, Italy) or HFF CM containing the indicated concentration of drug. After 24 h from treatment, cells were processed according to the manufacturer’s instructions of the Subcellular Protein Fractionation Kit (Thermo Fisher Scientific, Rochester, NY, USA). Membrane and cytoplasmic extract were analyzed by using Western blot assay.

3x10⁵ HFF, BJ, and NF cells were plated into 60 mm dishes (Falcon BD, Oxford, UK). After 24 h from the plating, the culture medium was replaced by wo FBS DMEM, and after 48 h from the plating media were collected and cells were counted. Cells culture media were analyzed by Human Angiogenesis Antibody Array (R&DSystems, MN, USA) according to the manufacturer’s protocol. Cell culture media were also analyzed by IL-8 and IL-6 (Enzo Life Sciences, Farmingdale, NY, USA), and MCP-1 (R&DSystems, MN, USA) specific ELISA assay, according to the manufacturer’s protocol. Absorbance was read at 450 nm. IL-8, IL-6, and MCP1 expression were represented as pg/mL and then related to the control assuming as to 100%.

6x10⁶ HCT116 Parental cells were seeded into 100 mm dishes (Greiner bio-one, Kremsmünster, Austria) in RPMI 1640 (Euroclone, Milan, Italy) and after 24 h from plating cell culture media were replaced by wo FBS DMEM (Euroclone, Milan, Italy) or HFF CM containing the indicated concentration of drug. After 24 h from treatment, cells were processed according to the manufacturer’s instructions of the PTEN activity ELISA assay (Echelon Biosciences Inc., Salt Lake City, UT, USA). Absorbance was read at 450 nm. PTEN phosphatase activity was represented as PIP2 concentration (pmol/μL) and then related to the control assuming as to 100%.

Results are expressed as average of three independent experiments. Results with two-tailed p values <0.05 were judged to be statistically significant. The dose of drug that causes 50% of cell growth inhibition (IC₅₀) was calculated according to the Chou-Talalay method using the Calcusyn software (Biosoft, Cambridge, United Kingdom). The same software was used for deriving combination indexes (CI): the average CI at the ED50, ED75, and ED90 <1 indicates synergism, = 1 indicates additivity, and >1 indicates antagonism, respectively. Statistical analysis for co-localization was conducted using GraphPad Prism (RRID : SCR_002798) (www.graphpad.com).

We evaluated the role of PTEN status and microenvironmental interactions in the modulation of CRC response to single or combined MAPK/PI3K inhibition. The X-MAN™ isogenic HCT116 cell lines [HCT116 Parental (PTEN-competent) and HCT116 PTEN^-/- (PTEN-loss)] were treated with increasing concentrations of the MEK inhibitor Tram, the PI3K/mTOR double inhibitor Geda, or their combination (Tram+Geda, 1:1 fixed dose-ratio) for 72 h under different culture conditions: serum-free medium (wo FBS), fibroblast-CM (CM HFF), or direct CRC/fibroblast co-culture (co-co), using GFP-tagged immortalized skin fibroblast (HFF). Pathway inhibitors (alone or in combination) displayed little, if any, growth inhibitory activity on isolated HFF fibroblasts (data not shown).

In CRC cells, sensitivity to the growth-inhibitory activity of Tram significantly decreased in the absence of a functional PTEN (HCT116 PTEN^-/- cells), regardless of cell culture conditions ( Supplementary Figure S1 and Table 1 ). Conversely, exposure to fibroblast-CM specifically sensitized PTEN-competent (HCT116 Parental), but not PTEN-loss (HCT116 PTEN^-/-) CRC cells to Geda [48% reduction in the IC₅₀ in HCT116 Parental cells under CM HFF, as compared with wo FBS, conditions ( Table 1 ), p-value <0.05 at all the tested doses ( Figure 1A left panels)]; direct co-co had no effect on sensitivity to Geda, regardless of PTEN status ( Figure 1A right panels and Table 1 ). As we previously reported (9), pharmacological interactions between Tram and Geda were highly synergistic in PTEN-loss and antagonistic in PTEN-competent cells, under isolated cell culture (wo FBS) conditions ( Figure 1B left panels, Supplementary Figure S2A and Table 1 ). Exposure to HFF CM and direct co-co turned the interaction between Tram and Geda into highly synergistic in the HCT116 Parental cell line (CI: 0.01 and 0.0007, respectively), whereas only direct cell-to-cell contact influenced pharmacological interactions in HCT116 PTEN^-/- cells, reverting synergism (CI: 0.99) and reducing (albeit not significantly) the growth inhibitory response to combined treatment ( Figure 1B right panels, Supplementary Figure S2B and Table 1 ).

We focused on the observed sensitization to the growth-inhibitory effects of PI3K/mTOR inhibition under exposure to fibroblast-CM. First, the same functional sensitization effects observed with HFF CM were obtained using CM from another immortalized normal foreskin fibroblast cell line (BJ) and from primary fibroblasts isolated by normal colon tissue ( Supplementary Figures S3A, B ). Moreover, another double PI3K/mTOR inhibitor (Dact) exerted functional effects similar to those observed with Geda ( Supplementary Figure S4A and Supplementary Table S1 ). Single-step inhibition along the PI3K cascade was not sufficient to obtain such functional effects: indeed, response to specific inhibitors targeting PI3K, AKT, or mTOR individually (Alp, MK, and Eve, respectively) was not modulated by exposure to fibroblast-CM in either isogenic cell line ( Supplementary Figures S4B–D and Supplementary Table S1 ). These results confirm that exposure to normal fibroblast-CM specifically sensitizes PTEN-competent CRC cells to double PI3K/mTOR inhibition.

We next explored the effect of fibroblast-CM in modulating functional response to Geda in four PTEN-competent CRC cell lines (LS180, HT29, RKO, and SW480), characterized by different genetic background (KRAS-mut/PIK3CA-mut, KRAS-wt/PIK3CA-wt, KRAS-wt/PIK3CA-mut, and KRAS-mut/PIK3CA-wt, respectively) ( Figures 2A, B ). Each cell line was exposed to increasing concentrations of Geda (0.1-1000 nM) under wo FBS or HFF CM conditions for 72 h. Similar to HCT116 Parental cells, only LS180 (PIK3CA-mut/KRAS-mut/PTEN-competent) cells displayed a statistically significant difference in their response to Geda according to culture conditions (p-value < 0.05 for the lowest concentrations of Geda; Figures 2A, B ). Conversely, in the KRAS-wt/PIK3CA-mut RKO cell line exposure to fibroblast-CM induced significant resistance to Geda at the highest concentrations. No significant differences were observed in the HT29 and SW480 cell lines ( Figure 2A ).

Since HCT116 cells are known to express high levels of the p110δ PI3K isoform (15), we also evaluated whether the pattern of p110δ PI3K expression correlated with functional response to Geda in the presence or absence of HFF-CM. As shown in Supplementary Figure S4E , p110δ PI3K (PIK3CD) expression by RT-qPCR was highest in HCT116 Parental and SW480 cells when compared in a panel of five cell lines, and did not correlate with HFF-CM induced functional sensitization to Geda-mediated growth inhibition.

We next investigated the response of isogenic KRAS HCT116 cell lines (parental HCT116 - heterozygous KRAS-mut, HK2-6 - homozygous KRAS-mut, and HKE-3 cell lines - homozygous KRAS-wt, respectively) to Geda under different culture conditions. As shown in Figure 2C , the most pronounced fibroblast-CM dependent functional sensitization to Geda was observed in the homozygous KRAS-mut isogenic cell line HK2-6. Overall, these results suggest that, in addition to the expression of a functional PTEN protein, the mutational status of both KRAS and PIK3CA plays a role in the observed functional sensitization to double PI3K/mTOR inhibition induced by fibroblast-CM in CRC cells.

We next investigated whether fibroblast-CM might influence PTEN expression and function and PI3K pathway activation. Exposure of HCT116 Parental cells (PTEN-competent) to fibroblast-CM for 24 h induced PTEN phosphorylation at Ser 380/Thr 382/383, along with relative protein accumulation, and p70^S6K1 phosphorylation at Thr 389, with relative protein accumulation; under these conditions, double PI3K/mTOR inhibition by Geda paradoxically induced PTEN and p70^S6K phosphorylation under unstimulated conditions, while substantially decreasing their phosphorylation induced by fibroblast-CM ( Figure 3A left panel). A similar modulation in p-PTEN Ser 380/Thr 382/383 and p70^S6K1 Thr 389 in response to both fibroblast-CM exposure and Geda treatment was also observed in the PTEN-competent LS180 cell line ( Supplementary Figure S5A ) and comparable results were observed using a different double PI3K/mTOR inhibitor (Dact) in HCT116 Parental cells ( Supplementary Figure S5B ). Conversely, in HCT116 PTEN^-/- cells p70^S6K1 Thr 389 phosphorylation levels were similar under both unstimulated and fibroblast-CM stimulated conditions and were consistently decreased by Geda under either culture condition ( Figure 3A right panel). Modulation in other pathway elements (such as reduction in p-mTOR Ser 2448, Ser 2481, and Thr 2446, and total mTOR protein; increase in p-4E-BP1 Thr 37/46 and total 4E-BP1 protein levels) upon exposure to fibroblast-CM was consistent between PTEN-competent (HCT116 Parental, left panel) and PTEN-loss (HCT116 PTEN^-/-, right panel) isogenic CRC cells ( Supplementary Figure S5C ).

Among soluble factors potentially involved in the fibroblast-CM induced PI3K pathway stimulation in PTEN-competent CRC cells, IL-8, IL-6, and MCP-1 were the most prominently produced by all three normal fibroblast cell lines tested (HFF, BJ, and NF; Supplementary Figure S5D ); however, exogenous addition of recombinant IL-8, IL-6, or MCP-1 (alone or in combination) was not able to induce p70^S6K1 phosphorylation at Thr 389, despite upregulation of p-PTEN Ser 380/382/383 levels ( Supplementary Figure S5E ).

To further investigate the mechanisms underlying paradox PI3K pathway activation in response to fibroblast-CM stimulation in PTEN-competent cells, we examined mTOR complex formation under different culture conditions. Co-immunoprecipitation experiments shown in Figure 3B indicated specific binding of RAPTOR to, and dissociation of, PRAS40 from mTOR under fibroblast-CM stimulated conditions.

Overall, such evidence indicates that yet unidentified soluble factors released by normal fibroblasts paradoxically activate the PI3K axis in PTEN-competent cells, by specifically activating the mTORC1 complex.

As shown in Figure 3A , fibroblast-CM induced prominent phosphorylation of PTEN’s C-tail (Ser 380/Thr 382/383) in PTEN-competent CRC cells. Consistent with the well-known role of PTEN phosphorylation in modulating protein localization and activity, immunofluorescence experiments conducted in HCT116 Parental cells showed predominantly membrane-bound PTEN under basal conditions (wo FBS), while its distribution became spread and predominantly cytoplasmic upon exposure to either Geda or fibroblast-CM ( Figure 4A ); under these conditions, PIP3 distribution essentially paralleled that of PTEN protein, even though the spread, cytoplasmic distribution and relative abundance of PIP3 were less evident in Geda-treated, as compared with fibroblast-CM treated, cells. Regardless, under fibroblast-CM stimulated conditions Geda treatment restored membrane-bound PTEN/PIP3 localization ( Figure 4A ). These findings were further confirmed by cell fractionation experiments ( Figure 4B ): indeed, PTEN protein decreased in the membrane-bound fraction and increased in the cytoplasmic fraction upon exposure to either Geda or fibroblast-CM and accumulated back in the membrane fraction upon combined fibroblast-CM/Geda treatment; interestingly, the mTOR protein changed its subcellular distribution in parallel to that of PTEN ( Figure 4B ). However, when PTEN enzymatic activity was measured in whole cell lysates in the presence of exogenous PIP3 by ELISA (thus not reflecting the potential role of PTEN subcellular localization), Geda treatment did not significantly affect PTEN lipid phosphatase activity under basal (wo FBS) conditions, while exposure to fibroblast-CM reduced enzymatic activity by approximately 50%; combined fibroblast CM/Geda treatment restored PTEN activity to basal levels ( Supplementary Figure 6 ).

Overall, these data suggest that fibroblast-CM induced PI3K pathway activation is due, at least in part, to PTEN C-tail phosphorylation, subcellular redistribution, and impaired enzymatic activity, all of which can be restored by simultaneous PI3K/mTOR inhibition.

Taken together, the above-described results indicate that the presence of a functional PTEN protein is necessary for fibroblast-CM mediated paradox PI3K pathway activation and, hence, to sensitize CRC cells to the growth-inhibitory effects of double PI3K/mTOR inhibition. To dissect whether these molecular and functional effects are mediated by PTEN protein-protein interaction or lipid phosphatase domains, HCT116 Parental cells were stably transfected with two chimeric GFP-mutant PTEN genes exerting a dominant negative function: the Q399STOP construct produces a PTEN protein lacking the PDZ-domain and protein-protein interaction ability; the C124S construct produces a PTEN protein lacking enzymatic activity and the ability to dephosphorylate both proteins and lipids (14, 16, 17).

C124S transfection did not affect response to Geda under either basal (wo FBS) or fibroblast-CM stimulated conditions ( Figure 5A , left panel). Conversely, Q399STOP transfection completely abrogated the Geda-sensitizing effect of fibroblast-CM: indeed, response to Geda was not significantly different between fibroblast-CM stimulated and unstimulated Q399STOP-transfected cells; however, Q399STOP-transfected cells were significantly more sensitive to Geda than their Parental counterpart under unstimulated conditions and less sensitive under fibroblast-CM stimulated conditions ( Figure 5A , right panel). Molecular endpoints were modulated consistently with the observed functional effects: HFF CM-dependent stimulation of p70^S6K1 Thr 389 and 4E-BP1 Thr 37/46 phosphorylation was less pronounced in Q399STOP-transfected cells, as compared to HCT116 Parental and C124S-transfected cells ( Figure 5B ), suggesting a more prominent role for PTEN protein-protein interaction, rather than phosphatase, activity in fibroblast-CM induced paradox PI3K pathway activation. Together with evidence of fibroblast-CM mediated mTORC1 formation only in PTEN-competent cells and concomitant redistribution of both PTEN and mTOR upon treatment, this observation led us to investigate whether PTEN could physically interact with mTOR. To this purpose, we evaluated PTEN and mTOR co-localization under different culture conditions using confocal microscopy. As shown in Figure 5C , a substantial pattern of superimposition between the two proteins was observed in unstimulated HCT116 Parental cells (wo FBS); exposure of HCT116 Parental cells to either Geda or fibroblast-CM significantly reduced the pattern of superimposition (p<0.0003 for both), which returned to basal when fibroblast-CM stimulated cells were exposed to Geda (p-value for the comparison between HCT116 Parental cells wo FBS and HCT116 Parental cells HFF-CM+Geda not significant), closely paralleling data on subcellular PTEN localization.