We evaluated the expression levels of PD-1, PD-L1 and PD-L2 in a panel of human TC cell lines derived from PTC (BcPAP, TPC-1) or ATC (8505c, CAL62, SW1736, FRO, BHT101, HTH7, OCUT1) compared to a primary human thyroid cell culture (H-6040). Cytofluorimetric analysis demonstrated that all the cell lines expressed PD-1 on the plasma membrane, though to a lesser extent than PD-Ls, and that PD-1 protein levels were higher in cancer compared to normal thyroid cells (Fig. 1A). PD-1, PD-L1 and PD-L2 mRNA levels were comparable between normal and cancerous thyroid cells, suggesting that post-translational mechanisms could be responsible for the protein increase observed in cancer cells (Suppl. Fig. 1).
Immunohistochemical (IHC) staining of whole sections from 34 PTC surgical samples with anti-PD-1 antibodies showed that PD-1 is expressed in TC cells (Fig. 1B), but not in normal thyroid epithelial cells (not shown). Figure 1B shows a representative PTC case with negative PD-1 staining (PTC1), intense PD-1 staining in the tumour immune infiltrate (PTC2), and PD-1 immunoreactivity, cytosolic and/or localized at the plasma membrane, in thyroid cancer epithelial cells (PTC3).
PD-1 expression was detectable in epithelial cancerous cells of 47% of tumour samples (Table 1). By analysing clinic-pathologic features of the PTC samples, we found that tumour stage and lymph-nodal metastasis significantly correlated with PD-1 staining (Table 1) in our casistic.
Table 1
Relation between PD-1 epithelial TC cell expression and clinic-pathological features. * P< 0.05 among groups
|
Epithelial cell PD-1 staining
|
|
|
Negative/low
|
Positive
|
P
|
BRAF (n=34)
not mutated
mutated
|
2
16
|
2
14
|
0.9
|
Thyroiditis (n=33)
No
Yes
|
12
5
|
11
5
|
0.9
|
TNM (n=34)
T1
T2
T3
N0
N1
M0
M1
|
11
4
3
17
1
11
7
|
9
3
4
6
10
15
1
|
0.77
0.04*
0.9
|
Progression free survival (n=28)
No
Yes
|
3
12
|
4
9
|
0.51
|
Stage (n=29)
1
2
3
4
|
12
0
1
2
|
6
0
5
3
|
0.04*
|
Infiltrative phenotype (n=34)
No
Yes
|
15
3
|
13
3
|
0.87
|
PD-1 promotes thyroid carcinoma cell proliferation and motility
We selected 8505c and TPC-1 cells - derived from a human ATC and PTC, respectively - to analyse the biologic effects of PD-1 enforced expression or of PD-1 stimulation by soluble PD-L1 (sPD-L1 - 1 mg/ml). Endogenous PD-1 protein expression levels in these cell lines, together with levels of PD-1 expression upon transient transfection, is shown in Suppl.Fig. 2A. We demonstrated that transient PD-1 overexpression (pFLAG PD-1 compared to pFLAG) or PD-1 activation (sPD-L1 vs untreated - NT) significantly increased DNA synthesis, as assessed by BrdU incorporation (Suppl. Fig. 2A) in both TC cell lines. Accordingly, cell cycle analysis showed an increased percentage of cells in S and G2/M phases in PD-1-transfected compared to empty vector-transfected TC cells (Suppl. Fig. 2B). No effects of PD-1 overexpression/activation were observed on cell viability (Suppl. Fig.2C). In order to confirm these observations, we evaluated the effects of PD-1 inhibition on the same cellular functions. To this aim, PD-1 expression was inhibited by siRNA or Nivolumab (anti-PD-1 moAb) in TPC-1 and 8505c cells. Both siRNAs targeting PD-1 (siPD-1 vs siCTR - 100 nM; Suppl. Fig. 2D) and Nivolumab (10 mg/ml) (Nivo vs IgG4) were able to significantly inhibit BrdU incorporation (Fig. 2B) and cell cycle progression (Suppl. Fig. 2E) of TC cells in comparison to the relative controls, without affecting cell viability (Suppl. Fig. 2F).
To assess the role of endogenous PD-1 ligands in TC cell proliferation, we treated 8505c cells with blocking anti-PD-L1 or anti-PD-L2 moAb (10 mg/ml) or transiently transfected them with PD-L1 or PD-L2 expressing vectors. PD-L1 or PD-L2 overexpression increased, while anti-PD-L1 or anti-PD-L2 antibodies inhibited, BrdU incorporation in 8505c cells (Suppl. Fig. 2G). No effects of PD-L1 or PD-L2 were observed on TC cell viability (not shown).
Since PD-1 expression levels in human TC samples correlated with lymph-nodal metastasis, we asked whether PD-1 could also stimulate the motility of TC cells. To this aim, we performed migration assays on 8505c cells stably overexpressing or not PD-1 [pCMV3 PD-1 cl13 and cl16 compared to pCMV3 empty vector-transfected cells (Suppl. Fig. 3A)] or on parental 8505c cells treated or not with sPD-L1 (1 mg/ml) in the presence or absence of Nivolumab (10 mg/ml) (Fig. 2C). PD-1 overexpressing TC cells showed increased migratory potential compared to control cells. Consistently, sPD-L1 induced, and Nivolumab inhibited, both basal and sPD-L1-induced migration (Fig. 2C).
These data indicate that PD-1 intrinsic circuit sustains TC cell proliferation and migration.
PD-1 activates the Ras/MAPK signalling cascade in thyroid carcinoma cells
We then asked which signalling pathway was stimulated upon PD-1 overexpression/activation. To this aim, we used specific phospho-antibodies against various signalling proteins. We found that BRAF, MEK and MAPK (p44/p42) are activated, as demonstrated by increased levels of their phosphorylated forms, upon PD-1 transient transfection (Fig. 3A), PD-1 stable transfection (Suppl. Fig. 3B), and sPD-L1 treatment (Fig. 3B) in both 8505c and TPC-1 cells. No significant activation of other signalling proteins was detected (Suppl. Fig. 3C). To confirm these observations, BRAF, MEK1/2 and MAPK activation levels were evaluated upon PD-1 blockade by siPD-1 or Nivolumab treatment. Consistently, both siPD-1 (100 nM) and Nivolumab (10 mg/ml – 15 and 30 min) reduced the levels of phosphorylated BRAF, MEK1/2 and MAPK compared to the relative controls (Fig. 3C) in TC cells.
Since the BRAF/MEK/MAPK signalling is potentiated by PD-1 in TC cells, and Ras GTPase is the main upstream activator of this cascade [28], we asked whether PD-1 could activate Ras. To this end, we used a pull-down assay with the GST-RAF1-Ras binding domain (RBD), which specifically binds the GTP-loaded active form of Ras. 8505c and TPC-1 cells were transiently transfected with empty vector (pFLAG) or PD-1 (pFLAG PD-1) in combination with pCEFL H-Ras AU5 or the relative empty vector (pCEFL). PD-1 enforced expression increased Ras activation, as assessed by Ras pull-down, in comparison to control (Fig. 3D), suggesting that PD-1 potentiates Ras activation in TC cells.
PD-1 recruits and activates the SHP2 phosphatase in thyroid carcinoma cells
In immune cells, PD-1 signalling requires the tyrosine phosphatase SHP2 (PTPN11) [27]. Upon phosphorylation of tyrosine residues in its cytosolic domain, PD-1 binds to the SH2 domains of SHP2 that, in turn, dephosphorylates signalling components of the immune receptors, thus down-regulating the immune responses [29]. In cancer cells, SHP2 acts as a signalling molecule downstream receptor tyrosine kinases (RTKs), displaying oncogenic activity [30]. In particular, SHP2 can contribute to Ras activation either by recruiting the GRB2/SOS complex to the plasma membrane [31] or through its phosphatase activity on Ras inhibitory tyrosine residues [31, 32].
We first asked whether PD-1 could physically interact with SHP2 in TC cells. Reciprocal co-immunoprecipitation experiments showed that endogenous and exogenously expressed PD-1 bind SHP2 in 8505c and TPC-1 cells (Fig. 4A). Moreover, pull-down assays with N- or C-terminal SH2 domain of SHP2 demonstrated that SHP2 can bind PD-1 mainly through SHP2 C-terminal SH2 domain (Fig. 4B). In support of these observations, we found that both endogenous and exogenous PD-1 are tyrosine phosphorylated in TC cells (Suppl. Fig. 4A), condition necessary to allow the SH2 domains of SHP2 to bind PD-1 [31].
Cell fractionation of 8505c cells transiently or stably transfected with PD-1 was used to demonstrate that PD-1 binding to SHP2 enforced the membrane localization of SHP2. Subcellular fractions of membranes (M) or cytosol (C) were obtained from PD-1 overexpressing and from control cells (pFLAG-PD-1 vs pFLAG or pCMV3 PD-1 cl 16 vs pCMV3). Enrichment of SHP2 levels in the membrane fractions was observed in PD-1 overexpressing cells compared to empty-vector transfected cells. Normalizations of each extract were obtained by using antibodies to transferrin receptor for membrane fraction and a-tubulin for cytosolic extract (Fig. 4C). In agreement with these observations, immunofluorescence (IF) assay of PD-1 overexpressing TC cells showed a significant increase of SHP2 staining at the plasma membrane in cells overexpressing PD-1 compared to controls (Fig. 4D and Suppl. Fig. 4B).
Furthermore, in 8505c cells transfected with PD-1-GFP, we demonstrated by IF that SHP2 and PD-1-GFP co-localize at the plasma membrane (Suppl. Fig. 4C).
SHP2 dephosphorylates and activates Ras in TC cells
We then searched for the molecular mechanism of Ras activation mediated by the PD-1/SHP2 complex. We first asked whether PD-1 could enhance GRB2 recruitment by SHP2. To this aim, we used pull-down assays with GST-SH2-GRB2 fusion proteins and co-immunoprecipitation assays showing no increased GRB2 binding to SHP2 in PD-1 transfected TC cells compared to controls (Suppl. Fig. 4D). In accordance with these observations, PD-1 enforced expression did not significantly increase SHP2 tyrosine phosphorylation levels (Suppl. Fig. 4A), on which GRB2 binding to SHP2 is dependent, nor changed substantially GRB2 compartmentalization as demonstrated in cell fractionation experiments (Fig. 4C).
Since the GRB2/SOS complex is not involved in PD-1-mediated Ras activation, we asked whether Ras could be activated by SHP2 through the dephosphorylation of its inhibitory tyrosine residues [27, 33]. We evaluated the phosphatase activity of SHP2 and, in parallel, the levels of Ras tyrosine phosphorylation in cells overexpressing or not PD-1. We used a specific SHP2 phosphorylated substrate in the presence of the Malachite Green tracer, a colorimetric method for the detection of free inorganic phosphate [27]. We observed that SHP2 phosphatase activity was significantly increased in PD-1- versus empty-vector-transfected TC cells (Fig. 4E). Similar results were obtained in PD-1 stably transfected cells (not shown). Consistently with the increased phosphatase activity of SHP2, Ras total phosphorylation levels, in the presence of PD-1, were significantly reduced in TC cells transfected with pCEFL H-Ras AU5 (Suppl. Fig. 4E). To assess whether Ras dephosphorylation occurs in its inhibitory residues 32 and/or 64 [27], we used (pan)Ras immunoprecipitation followed by immunoblotting with anti-phospho Y32 (Ras) or Y64 (Ras) antibodies. These experiments demonstrated that PD-1 enforced expression in 8505c cells reduced the Ras phosphorylation levels in the inhibitory tyrosine residues 32 in pCEFL Ras AU5-transfected cells compared to controls (Fig. 4F). Similar results were obtained in TPC-1 cells (not shown). No differences in phosphorylation levels of inhibitory residues 64 were observed (not shown).
Taken together, these data indicate that, in TC cells, PD-1 binds SHP2, which in turn dephosphorylates Ras in its inhibitory tyrosine, thus leading to the activation of the MAPK signalling cascade.
PD-1-induced biologic activities in thyroid cancer cells require the SHP2/BRAF/MEK signalling proteins
To investigate the role of SHP2 in PD-1 functional activity, we treated TC cells, overexpressing or not PD-1, with siRNA targeting SHP2 (siSHP2 – 100 nM) or with a SHP2 allosteric inhibitor that blocks its phosphatase activity (SHP099 – 350 nM) [34]. As shown in Figure 5A, siSHP2 was able to significantly reduce SHP2 protein levels compared to scrambled siRNAs (siCTR). By BrdU incorporation assays, we demonstrated that siSHP2 significantly decreased DNA synthesis (Fig. 5B) in PD-1-, and to a lesser extent in empty vector- transfected, 8505c cells. Consistently, SHP099 inhibitor significantly reduced PD-1-induced DNA synthesis in 8505c cells (Fig. 5C).
To investigate the role of the downstream signalling cascade in PD-1 dependent biologic TC responses, we conducted BrdU incorporation assays in TC cells overexpressing or not PD-1, in the presence or in the absence of Vemurafenib (Vemu – 10 mM) [35], a BRAF inhibitor, or Selumetinib (Selu – 10 mM) [36], a MEK inhibitor. As shown in Figure 5D, both drugs were able to significantly revert PD-1-induced DNA synthesis in 8505c cells.
Similar experiments were performed to assess the role of the signalling cascade in PD-1-mediated TC cell migration. Figure 5E shows that SHP099 and Vemurafenib, and to a lesser extent Selumetinib, were able to inhibit the migration of 8505c cells induced by sPD-L1. Similar results were obtained in TC cells transfected with PD-1 (not shown).
These data demonstrate that PD-1-induced cell proliferation and motility of TC cells are dependent on the SHP2/BRAF/MEK pathway.
Intrinsic PD-1 signalling enhances xenograft growth of TC cells in immunocompromised mice
To verify whether PD-1 intrinsic signalling and biologic activity could affect tumorigenicity of TC cells, we xenotransplanted 8505c pCMV3 PD-1 (two clones) and control 8505c pCMV3 (a mass population) cells in athymic mice. 8505c pCMV3 PD-1 xenografts displayed increased tumour growth rate that was statistically significant at 4 weeks after injection, in comparison to empty vector transfected cells (Fig. 6A). End-stage tumours were excised and analysed for cell proliferation (Ki-67), apoptotic rate (cleaved-caspase 3) and vessel density (CD31) by immunohistochemistry. 8505c pCMV3 PD-1 and 8505c pCMV3 xenografts exhibited statistically significant differences in cell proliferation rate, but not in apoptotic rate or vessel density (Fig. 6B and Suppl. Fig. 5A).
To verify whether the inhibition of PD-1 by Nivolumab could affect xenograft growth of parental 8505c cells, mice were xenotransplanted, randomized in two homogeneous groups, and administered intraperitoneally (i.p.) with Nivolumab or control IgG4 (30 mg/kg) twice a week. 5 weeks after xenotransplantation, Nivolumab-treated tumours showed a significant decrease in growth rate in comparison with the IgG4-treated group (Fig. 6C). Consistently, Nivolumab significantly reduced TC xenografts’ proliferation without affecting apoptotic rate or vessel density (Fig. 6D and Suppl. Fig. 5B).
Despite these experiments were carried out in immunocompromised mice, we could not exclude that Nivolumab anti-tumour activity could be ascribed to its ability to affect innate immunity that is present and functional in athymic mice. Thus, we analysed the density and activation of immune cells infiltrating 8505c xenografts treated with Nivolumab or with IgG4 by cytofluorimetric analysis. We found that Nivolumab treatment did not change the percentage of CD45+ leucocytes infiltrating xenografts in comparison to IgG4 controls, at least at 5 weeks of treatment. Moreover, the density and the expression of polarization/activation markers of tumour-associated macrophages (TAM), of Ly6C+ and Ly6G+ immature myeloid cells, of mature and immature dendritic cells and of regulatory or activated NK, and NKT cells, were comparable between Nivolumab- and IgG4-treated 8505c xenografts (Suppl. Table 1).
These data indicate that, in our model system, PD-1 blockade by Nivolumab inhibits TC cell xenograft growth by affecting tumour cell rather than immune cell compartment.