Proteasome inhibition by bortezomib parallels a reduction in head and neck cancer cells growth, and an increase in tumor-infiltrating immune cells

Head and neck cancer (HNC) has frequently an aggressive course for the development of resistance to standard chemotherapy. Thus, the use of innovative therapeutic drugs is being assessed. Bortezomib is a proteasome inhibitor with anticancer effects. In vitro antitumoral activity of Bortezomib was investigated employing human tongue (SCC-15, CAL-27), pharynx (FaDu), salivary gland (A-253) cancer cell lines and a murine cell line (SALTO-5) originated from a salivary gland adenocarcinoma arising in BALB-neuT male mice transgenic for the oncogene neu. Bortezomib inhibited cell proliferation, triggered apoptosis, modulated the expression and activation of pro-survival signaling transduction pathways proteins activated by ErbB receptors and inhibited proteasome activity in vitro. Intraperitoneal administration of Bortezomib delayed tumor growth of SALTO-5 cells transplanted in BALB-neuT mice, protracted mice survival and adjusted tumor microenvironment by increasing tumor-infiltrating immune cells (CD4+ and CD8+ T cells, B lymphocytes, macrophages, and Natural Killer cells) and by decreasing vessels density. In addition, Bortezomib modified the expression of proteasome structural subunits in transplanted SALTO-5 cells. Our findings further support the use of Bortezomib for the treatment of HNC and reveal its ineffectiveness in counteracting the activation of deregulated specific signaling pathways in HNC cell lines when resistance to proteasome inhibition is developed.

). The mean result of two independent experiments were reported in Fig. 2. Representative flow cytometry plots in which the effect of Bortezomib on DNA content was compared to that obtained with DMSO in the different cell lines were reported in Supplementary Fig. S1.  www.nature.com/scientificreports/ To confirm the induction of apoptosis, cells were simultaneously treated with Bortezomib at the highest dose and with the universal caspase inhibitor, Z-VAD-FMK. Administration of this drug was able to significantly reduce the number of cells in the subG1 phase as compared to the single treatment with Bortezomib at the highest dose, thus suggesting the induction of cell death by apoptosis following treatment with Bortezomib in HNC cell lines ( Fig. 2 and Supplementary Fig. S1).  www.nature.com/scientificreports/ Effects of Bortezomib on the expression and activation of ErbB receptors (EGFR and ErbB2) and pro-survival signaling transduction pathway molecules (ERK, JNK, p38, AKT) in HNC cell lines. The MAP (Mitogen Activated Protein) kinase transduction pathway is triggered by the activation of EGFR and ErbB2/neu tyrosine kinase receptors. It has been demonstrated that HNC cell lines overexpress EGFR and ErbB2 receptors, which play a role in their cellular transformation 35 .
In addition, we evaluated the effect of Bortezomib on the expression and phosphorylation of MAP kinases ERK, JNK and p38. Our results showed that Bortezomib inhibited the phosphorylation of ERK1 and ERK2 in SCC-15 (p = 0.0027, for pERK1; p = 0.0009, for pERK2) and A-253 (p = 0.0043, for pERK1; p = 0.0003, for pERK2) cells as compared to untreated cells. In addition, Bortezomib treatment decreased the level of phosphorylation of ERK2 in FaDu cells (p = 0.016), while it increased it in SALTO-5 cells (p = 0.019). Bortezomib did not affect the levels of phosphorylation of ERK1/2 in CAL-27 cells (Fig. 4).

Validation of proteasome inhibition by Bortezomib in HNC cell lines.
In order to evaluate the effectiveness of proteasome inhibition in HNC cells by Bortezomib and to determine whether resistance to the drug may have occurred, SCC-15, CAL-27, FaDu, A-253 and SALTO-5 cells were treated with Bortezomib by following a scheme and doses based on results from the IC 50 values (see Fig. 2). At the indicated time points, Bortezomib-and DMSO-treated cells were harvested and the cytosolic fraction (i.e., where proteasome is mainly represented) was isolated through a non-denaturing lysis procedure. For every cell line, proteasome inhibition and individual particles content was first assayed by native gel electrophoresis 40 (see Materials and Methods for further details) (Fig. 6).
For all cell lines, a marked inhibition of the activity of all three main assemblies existing in the cell cytosol (i.e., the uncapped 20S, the doubly 19S:20S:19S and single-capped 19S:20S, respectively) was observed in the presence of Bortezomib (Fig. 6a). With respect to control cells, SCC-15, CAL-27, and SALTO-5 cells displayed the greatest extent of proteasome inhibition (> 50%) in the presence of the lowest Bortezomib concentration (25 nM in SCC-15 and SALTO-5, 12.5 nM in CAL-27) at the first time-point (7 h in SCC-15 and 12 h in SALTO-5 and CAL-27) of stimulation (Fig. 6a, upper panel; Supplementary Fig. S2). Conversely, in the case of A-253 cells, a > 50% extent of proteasome inhibition at 12 h of stimulation was achieved only at 50 nM Bortezomib. Interestingly, FaDu cells showed a less marked inhibition of proteasome activity as compared to the other cell lines. The extent of inhibition was roughly 50% in the presence of either 25 or 50 nM Bortezomib at 12 h and slightly lower at 24 h (Fig. 6a, upper panel; Supplementary Fig. S2). A modest rescue of proteasome activity was further observed in SALTO-5 cells stimulated with 25 and 50 nM Bortezomib for 24 h, as compared to the effect induced by the same Bortezomib concentrations after 12 h of stimulation (Fig. 6a, upper panel; Supplementary Fig. S2).
In the case of SCC-15 cells, an unexpected finding was observed. The overall proteolytic activity was found to decrease in untreated cells over time (14 h vs 7 h) ( Fig. 6a upper panel; Supplementary Fig. S2).
In accordance with a general inhibition of proteasome activity, the poly-ubiquitinated proteins, (i.e., the natural substrates of capped proteasome assemblies), which were assayed by denaturing and reducing Western blotting, turned out to be significantly increased in the presence of Bortezomib under all experimental conditions ( Supplementary Fig. S3).
To further validate their identity, proteasome particles were then transferred to a nitrocellulose filter and probed with an antibody which recognizes all catalytic assemblies of proteasome (i.e., 30S, 26S, and 20S), since www.nature.com/scientificreports/ it is raised against a peptide covering residues shared by the α1-7 subunits of 20S proteasome, with the exception of α4 (hereafter referred to as pan-α-subunits). By probing the filters some unexpected findings were observed   www.nature.com/scientificreports/ cells treated with Bortezomib showed the appearance of non-canonical proteasome assemblies with an apparent mass/charge ratio very similar to that reported by other authors in a similar experimental condition 41 (Fig. 6a, bottom panel, red arrow). It is worth pointing out that SCC-15 cells displayed a decrease in overall proteasome content also in control cells harvested after 14 h (with respect to control cells harvested after 7 h), thus confirming the observed loss of proteasome activity (Fig. 6). Conversely, SALTO-5 cells appeared to be somewhat resistant to this phenomenon, since this cell line did not display any significant decrease of proteasome particles content (Fig. 6a, bottom panel, and Fig. 6b) with the exception of cells treated with 50 nM for 12 h. In addition, A-253 and CAL-27 cells displayed a very high 20S/capped assemblies (30S + 26S) ratio, even though the pattern of activity (Fig. 6a, bottom panel, and Fig. 6b) was comparable to that of other cell lines analyzed which showed a more balanced 20S/capped assemblies ratio.
To better address these findings, the same crude cell extracts, run by native gel electrophoresis, were analyzed by denaturing, and reducing Western blotting. Interestingly, the content of free α7 and Rpt5 subunits, which are representative of the 20S and 19S particles, respectively, ranged from unchanged or even increased under the investigated experimental conditions ( Supplementary Fig. S3).
To further verify the observed phenomenon, FaDu cells, which displayed the apparent highest extent of proteasome loss, were further stimulated with 25 and 50 nM Bortezomib for 30 min and 2 h (Fig. 7). Proteasome activity was almost null as early as after 30 min of stimulation in the presence of 50 nM Bortezomib, whilst a residual activity was observed in the presence of 25 nM Bortezomib (Fig. 7, left panel). After 2 h of stimulation, the proteolytic activity was undetectable in the presence of all Bortezomib concentrations. Interestingly, by probing the filter with the anti-pan-α subunits antibody, it was observed that the proteasome assemblies were unaffected by Bortezomib after 30 min of stimulation (Fig. 7, left panel). Conversely, after 2 h of Bortezomib treatment, a very robust loss of proteasome assemblies was observed in the presence of 50 nM Bortezomib, whereas residual proteasome assemblies were detected in the presence of 25 nM Bortezomib (Fig. 7, right panel).
Toxicological evaluation of Bortezomib treatment by histological analysis. Hematoxylin and eosin staining was performed on tissue sections from multiple organs (liver, lung, kidney, spleen, heart) collected from BALB-neuT mice treated i.p. with Bortezomib for 4 weeks. Our results showed no cytological and architectural alterations in any of the organ examined. In particular, hematoxylin and eosin staining showed a well-preserved liver parenchyma with a normal portal triad; no alterations in both renal glomeruli and tubules; a well-preserved lung parenchyma; no signal of hypertrophy and inflammation in the heart; a normal architecture of the spleen ( Supplementary Fig. S4).  (Fig. 8a,b). This significant difference was maintained after 4 weeks (97.4 vs 513.5 mm 3 ; p = 0.0002) and until the 5th week, when two control-treated mice were sacrificed due to the excessive size of the tumor (251.6 vs 1035 mm 3 ; p = 0.0001) (Fig. 8d). The remaining controltreated mice were sacrificed at 6 (5 mice) and 7 (1 mouse) weeks ( Fig. 8a,b). At this stage (7 weeks after tumor challenge) only one Bortezomib-treated mouse was sacrificed due to excessive size of the tumor. In contrast, the remaining Bortezomib-treated mice were sacrificed at week 9 (3 mice), 10 (2 mice), 11 (1 mouse) and 12 (1 mouse) (Fig. 8a,b). The mean survival significantly increased for the mice treated with Bortezomib, as compared to the control-treated mice (9.5 vs 6 weeks, Bortezomib-treated mice vs control-treated mice; p = 0.0001) (Fig. 8c)   www.nature.com/scientificreports/ The presence of apoptotic cells was evaluated by the expression of cleaved caspase 3 in cancer cells employing immunohistochemical analysis (IHC) (Fig. 8e). The number of apoptotic cells within the tumors from Bortezomib-treated mice (62.8 ± 5.1) was higher than that from control-treated mice (8.4 ± 1.1) (p = 0.0001). www.nature.com/scientificreports/ The expression of ErbB2 on tumor cells in vivo was evaluated by IHC as well (Fig. 8e). Tumors from Bortezomib-treated mice showed a significantly lower expression of ErbB2 than that from control-treated mice (1.0 ± 0.4 vs 2.3 ± 0.8 staining intensity, p = 0.04). In addition, AKT phosphorylation was significantly decreased in tumors from mice treated with Bortezomib, as compared to those treated with DMSO (0.3 ± 0.2 vs 0.8 ± 0.2 staining intensity, p = 0.04). On the other hand, the same treatment did not affect AKT expression in tumors (Fig. 8e).

Histological analysis of tumors from mice treated with Bortezomib by Optical
Moreover, we investigated the effects of Bortezomib treatment on the number of tumor-infiltrating immune cells. The presence and/or modifications of the immune infiltrate was analyzed by IHC analysis by evaluating the number of CD4 + , CD8 + , CD57 + , CD25 + , F480 + and CD20 + cells. In addition, tumor vessels density and expression of some proteasome structural subunits (PSMA4, PSMD4, PSME1) and ubiquitin were evaluated. Our results demonstrated that Bortezomib treatment deeply affected the number of tumor-infiltrating immune cells. Indeed, a significant increase in the number of helper T lymphocytes (CD4 + ), cytotoxic T lymphocytes (CD8 + ), B lymphocytes (CD20 + ) and macrophages (F480 + ) were observed in the Bortezomib-treated group as compared to the control-treated mice ( Fig. 9 and Table 2). In addition, Bortezomib-treated tumors showed a significant increase in NK cells (CD57 + ) and CD25 + immunoregulatory T cells ( Fig. 9 and Table 2). These results suggest that Bortezomib can potentiate the recruitment of tumor-infiltrating immune cells. Moreover, a significant reduction in tumor vessels density from Bortezomib-treated mice, as compared to those from control-treated mice was observed ( Supplementary Fig. S5 and Table 2).
To investigate whether Bortezomib affected T cells activation and IFN-γ production, intratumoral and spleens (SPL) lymphocytes from Bortezomib-or PBS + DMSO-treated mice carrying transplanted (SALTO-5) cells were collected and analyzed by flow cytometry. T cells CD69 expression was employed to measure T cells activation. Tumor infiltrating CD4 + and CD8 + lymphocytes (TIL) were highly activated and CD8 + lymphocytes were particularly prone to IFN-γ production both in Bortezomib-and in DMSO-treated mice ( Fig. S6).
Proteasome structural subunits, both PSMA4 (e.g., the 20S α3 subunit) and PSMD4 (e.g., the 19S Rpn10 subunit) were faintly detectable in DMSO-treated tumors, with the exception of some isolated foci that displayed stronger immunoreactivity. However, the immunostaining of these subunits turned out to be significantly stronger (p < 0.001 for PSMA4, and p < 0.0001 for PSMD4) in Bortezomib-than in DMSO-treated tumors ( Fig. 10 and Table 2). Strikingly, immunostaining of PSME1 was markedly reduced in Bortezomib-vs DMSO-treated tumors (p < 0.0002) ( Fig. 10 and Table 2). PSME1 encodes for the α subunit of the PA28 complex, a regulatory particle alternative to the canonical 19S, whose expression is often transcriptionally upregulated upon delivery of stressors, at least in vitro. Nevertheless, with respect to DMSO-treated tumors, ubiquitin immunostaining was unaltered between the two experimental groups ( Fig. 10 and Table 2).

Ultrastructural analysis of SALTO-5 cells in vitro treated with Bortezomib. Ultrastructural anal-
ysis was performed by transmission electron microscopy on SALTO-5 cells treated with Bortezomib or DMSO at the concentration of 25 nM for 24 h. DMSO-treated cells showed heterogeneous forms with a predominance of round over stretched cells. The nuclei appeared large, mainly formed by euchromatin with low dense heterochromatin in the periphery. In the cytoplasm several mitochondria and cisterns of rough endoplasmic reticulum were detected, with few vacuoles (Supplementary Fig. S7a,b). Conversely, Bortezomib-treated cells showed mainly necrotic features ( Supplementary Fig. S7c) with cytoplasm swelling and the presence of numerous cytoplasmic vacuoles surrounded by a single or double membrane, the latter being probably of autophagic origin (Supplementary Fig. S7d). Apoptotic cells were also visible ( Supplementary Fig. S7e).

Discussion
The use of new therapeutic agents is being evaluated in HNC [5][6][7][8] . Among the approaches currently employed in clinical trials, the delivery of UPS inhibitors, mainly those targeting proteasome catalytic activity, is gaining considerable interest. The purpose of proteasome inhibitors is to block the turnover of the intracellular proteome, bringing about the formation of a toxic environment and the subsequent need to activate apoptotic processes 42,43 . In this regard, Bortezomib is a reversible inhibitor of the chymotrypsin-like activity of the 20S proteasome 14,20,21 and it has been reported to possess potent anticancer activity, both in vitro and in vivo [23][24][25][26]44,45 .
Preclinical studies have shown that Bortezomib, when used alone, inhibited cell growth, induced apoptosis, autophagy, and modulation of individual signaling transduction pathways in HNSCC (Head and neck squamous cell carcinoma) cells in vitro [46][47][48][49][50][51][52][53][54][55][56] . To our knowledge, this is the first report that analyzes in a comprehensive and detailed approach, the broaden cellular effects of Bortezomib, thus providing molecular clues to join together the proteasome inhibition with the activation of apoptosis, autophagy, as well as the modulation of cell survival signaling transduction pathways involved in cellular transformation of HNC cell lines, that is a concern which no previous single study has been dealing with. Bortezomib was previously shown to activate apoptosis by modulating the expression of apoptotic proteins, activation of caspases, cleavage of PARP-1 protein and hypodiploidia and www.nature.com/scientificreports/ www.nature.com/scientificreports/ phosphatidylserine externalization in HNSCC cell lines 46,48-52,56,57. Bortezomib was shown to stimulate autophagy in HNSCC cells in two studies showing autophagosomes formation, upregulation of LC3-I, -II, Beclin-1 and the JNK-dependent phosphorylation of Bcl-2 after Bortezomib treatment 58,59 . It was also suggested that autophagy attenuated the Bortezomib cytotoxicity 59 . Bortezomib was able to modulate expression and activation of signaling pathways in HNSCC, by inhibiting AKT activation and mTOR 46,60 , and by upregulating STAT3 55 in HNSCC cell lines. Only two studies showed that Bortezomib affects proteasome activity by inducing the accumulation of ubiquitylated proteins in larynx (UM-SCC-11A, -11B) 56 and mouth floor (SCC1) cell lines 59 . It is worth pointing out that the majority of the studies investigated the effect of Bortezomib on the modulation of individual biological pathways being restricted to SCC cell lines arising in head and neck. In our study the in vitro and in vivo effects of Bortezomib were for the first time analyzed in a salivary gland adenocarcinoma experimental model. Salivary gland carcinomas represent 6-8% of HNC, with heterogeneous morphologies and clinical outcomes 3 . The mainstream therapy for these types of cancer, when feasible, is the surgery followed by radiation therapy. Chemotherapy regimens have demonstrated controversial clinical outcomes with low responses for advanced or metastatic malignant tumors, so that new targeted therapies are under evaluation [61][62][63] . Bortezomib has been evaluated for the treatment of only adenoid cystic carcinoma in combination with doxorubicin in a phase II trial, leading to no complete or partial responses, but only stabilization of disease in patients 32 .
Our findings about apoptosis activation by Bortezomib corroborated previous results employing HNSCC cell lines and showed for the first time Bortezomib-mediated apoptosis in a salivary gland adenocarcinoma cell line (A-253). It is worth remarking that only the A-253 cell line showed a G2/M arrest simultaneous to apoptosis, thus indicating a different response to the drug. This cell cycle pattern after Bortezomib treatment was previously reported in tumor cell lines from larynx cancer 56 , colorectal cancer 64 , non-small cell lung carcinoma 65 , prostate cancer 43 , Ewing's sarcoma 66 , malignant mesothelioma and breast cancer 67 .
In addition, our results indicated that Bortezomib mediated biological effect in a cell line specific-dependent modality. We found a decreased expression of the EGFR and ErbB2 receptors only in tongue, salivary gland but not pharynx cancer cell lines. EGFR and ErbB2 are often over-expressed in HNC cells 35,68 and are frequently prone to heterodimerization that confers tumor growth advantage 69,70 . Bortezomib inhibited the phosphorylation of ERK1 and/or ERK2 in FaDu, SCC-15 and A-253 cells, while the modulation of the p38 activation induced by the drug was cell lines dependent. Bortezomib induced an increase in the phosphorylated form of JNK p54 and/or p46 in CAL-27, SCC-15, and A-253 cells but not in FaDu cells. This finding corroborated other studies in which it has been shown that Bortezomib induced apoptosis by activating p38 and/or JNK kinase in several types of cancer [71][72][73][74][75][76] . Our finding extends this observation to the salivary gland adenocarcinoma cell line. In addition, other studies have shown that the activation of JNK kinase is necessary for the activation of death by autophagy in HNSCC cell lines 47,58 . Accordingly, our results showed that Bortezomib induced autophagy in human HNSCC cells, but the process was then blocked, as showed by the increase of p62 39 . The same effect was observed for the first time in the salivary gland adenocarcinoma cell line. The block of the autophagic flux by Bortezomib was reported in ovarian cancer cells, hepatocellular carcinoma cells and endometrial cancer cells 77 , breast cancer cells 78 , and B-Raf-mutated melanoma cells 79 . Table 2. Immunohistochemical evaluation of markers for tumor infiltrating cells and molecules involved in the proteasome activity. *Two-tailed Student's T test. a Positive cell count/HPF averaging 5 representative microscopic fields. b Positive vessel count/HPF averaging 5 representative microscopic fields. c A combined scoring system was used for the IHC evaluation. Specifically, the total score (0-6) was obtained by adding the score associated to the number of positive cells and the score related to the signal intensity. The score associated to the number of positive cells was defined as follow: 0 (1 ≤ Positive cells/HPF), 1 (2 ≤ x ≤ 10 Positive cells/HPF), 2 (11 ≤ x ≤ 20 Positive cells/HPF), 3 (≥ 21 Positive cells/HPF). The score related to the signal intensity was defined as follow: 0 (absent/very low Intensity/HPF), 1 (low Intensity/HPF), 2 (moderate Intensity/HPF), 3 (high Intensity/HPF).

PBS + DMSO (mean ± SD)
Bortezomib (mean ± SD) p Value* www.nature.com/scientificreports/ Furthermore, the treatment with Bortezomib inhibited AKT phosphorylation and activation, which triggers a cell survival signal 80 , in both the tongue squamous carcinoma cell lines (SCC-15, CAL-27) and in the salivary gland adenocarcinoma cell line (A-253). The inhibition of AKT phosphorylation by Bortezomib is a key molecular event for Bortezomib-mediated apoptosis in HNC 46,60 and non-small cell lung cancer cells 76 . However, our findings showed that Bortezomib had no effect on AKT expression and phosphorylation on the FaDu pharynx cell line.
Since we observed different responses on the modulation of the signaling pathway molecules in HNC cell lines, we evaluated whether these differences were dependent on a different sensibility of cells to Bortezomibinduced inhibition of proteasome activity. The analysis of the structural and functional effects of Bortezomib on the proteasome assemblies in HNC cell lines is a further novelty of our study. Indeed, only two studies showed that Bortezomib affects proteasome activity by inducing the accumulation of ubiquitylated proteins in larynx 56 www.nature.com/scientificreports/ and mouth floor cells 59 . We found that the different responses observed upon Bortezomib treatment in the HNC cell lines could be due to a different extent of proteasome inhibition. Indeed, the pharyngeal carcinoma cell line (FaDu), which is found to be the most resistant to the action of Bortezomib, was that displaying the lowest extent of proteasome inhibition after 12 h and 24 h of stimulation, regardless the Bortezomib concentration administered. It is important to recall that, as highlighted above, Bortezomib was not able to induce modulation of EGFR, ErbB2, JNK, p38 as well AKT proteins in FaDu cells. The ineffectiveness of Bortezomib in modulating these signal transduction pathways may thus parallel the low efficacy of Bortezomib in inhibiting the proteasome activity. However, the Bortezomib inhibitory effect on overall proteolytic activity was several-fold greater when the proteasome assemblies were harvested and analyzed at earlier time-points. Without ruling out the possibility that FaDu cells may have evolved canonical mechanisms of drug resistance (e.g., drug secretion and/or detoxification, or selective downregulation of 19S subunits 81 ), the resistance of these cells to Bortezomib-induced apoptosis, which is in sharp contrast with the complete early proteasome inhibition after 2 h, can be likely explained through two different and not mutually exclusive hypotheses, namely: i) Bortezomib, being a reversible inhibitor, is displaced from the β5 catalytic site at a higher rate than in other cells; ii) among all cells employed in this study, FaDu are those which more readily synthesize de novo proteasome assemblies. Hypothesis i) indeed reflects a chemical property of Bortezomib which contributes to the resistance through which the cells can bypass the drug-induced death. However, although speculative at this stage, we envisage that hypothesis ii) may be of greater relevance to explain the observed behaviour for two main reasons: • FaDu cells were the only cells clearly inducing the formation of alternative proteasome assemblies at 12 h and 24 h (but not at 20 min or 2 h), which clearly resembles non-canonical complexes, such as PA28-20S, as previously proposed to occur as an adaptative response to proteasome inhibition in experimental models other than those herein discussed 41 . • FaDu cells showed the greatest extent of proteasome loss during treatment, a phenomenon which, to the best of our knowledge, has never been reported in the presence of a proteasome inhibitor. Remarkably, this Bortezomib-induced effect was observed, though to a variable extent, in all human cell lines tested so far, underscoring that it may be a general issue of pharmacological relevance, if confirmed in vivo.
At this stage, we cannot rule out a priori that it is a technical artifact, such as, for instance, epitope masking in the presence of undigested poly-ubiquitinylated substrates and further studies are demanded to clarify the biological relevance of this point.
Overall, our in vitro results indicated that the anti-cancer activities of Bortezomib was dependent on the type of HNC cell line, being positively related to the extent of proteasome inhibition exerted by Bortezomib in the different HNC cell lines. In fact, we have shown that the functional impairment of proteasome modulated in turn the signaling pathways involved in HNC progression. For example, Bortezomib had no effect on EGFR, ErbB2, JNK, p38 or AKT in FaDu cells. These effects paralleled the low efficacy of Bortezomib in inhibiting the proteasome activity in those cancer cells. Thus, our in vitro findings suggest that in HNC, showing limited proteasome resistance to Bortezomib and simultaneous upregulation of ErbB receptors-mediated signaling, anti-ErbB receptors antibodies or inhibitors of the ErbB receptors intrinsic tyrosine kinase activity should be used in combination with proteasome inhibitors. Furthermore, since other studies highlighted the development of different mechanisms of Bortezomib resistance in HNC, the investigation of the efficacy of combined treatment with Bortezomib and other inhibitors of different signaling pathways should be envisaged 13,27,55,56,59,82 .
In addition, regarding the analysis of the effects of Bortezomib on experimental models in HNC, there are several studies which have shown that Bortezomib has promising anticancer activities in mouse tumor models 45,46,56,[83][84][85] . However, only two studies analyzed the ability of Bortezomib to counteract the in vivo HNC tumor growth, but they were restricted to xenografts implanted human larynx or tongue squamous cell carcinoma cell lines 46,56 . Accordingly, none of them have investigated the in vivo effect of Bortezomib in salivary gland carcinoma cell line. Thus, this is the first study showing the in vitro and in vivo growth inhibitory properties of Bortezomib in a salivary gland carcinoma cell line (SALTO-5) and the in vivo effects of Bortezomib on tumorinfiltrating immune cells, tumor vessels density, expression of ErbB2, AKT and cleaved caspase 3 and molecules involved in proteasome structural composition in SALTO-5 cell line transplanted in BALB-neuT mice. The Bortezomib effects on SALTO-5 cells were first analyzed in vitro, and we observed apoptosis and inhibition of ErbB2, p38 and AKT, and activation of JNK in Bortezomib-treated SALTO-5 cell line. In contrast to the human cells analyzed, Bortezomib induced an increase of ERK2 phosphorylation in SALTO-5 cells, which was still associated with the activation of the apoptotic process. On the other hand, regarding the effect of Bortezomib on proteasome, SALTO-5 cells displayed a significant recovery of proteasome particles after 24 h and of proteasome subunits content associated to a decrease of poly-ubiquitinated proteins. Although the study of apoptosis pathway did not put in evidence any resistance of these cells greater than that of SCC-15, A-253 or CAL-27 cells, it is likely that this mechanisms of resistance to the drug may emerge over a prolonged time of treatment. Cytotoxic and apoptotic effects of Bortezomib in SALTO-5 cell line were also observed by ultrastructural analysis.
In light of the in vitro results, we also evaluated the in vivo anti-tumor effects of i.p. administration of Bortezomib (0.5 mg/kg, twice a week) on tumor growth in BALB-neuT mice subcutaneously inoculated with syngeneic murine SALTO-5 cells. It has been reported that i.p. administration, at least twice a week, resulted in greater Bortezomib activity with less toxicity 86 . We confirmed the absence of Bortezomib toxicity in organs collected from Bortezomib-treated mice. The preclinical investigations have collectively demonstrated the anticancer activity of Bortezomib when used as monotherapy or in combination with chemotherapy, radiotherapy, or other anti-neoplastic agents [86][87][88][89][90][91] . Our results showed for the first time that Bortezomib reduced the growth of SALTO-5 murine cells in mice and increased the survival of the mice. In addition, the histological examination of tumors www.nature.com/scientificreports/ from Bortezomib-treated mice showed extensive necrosis and presence of apoptotic cells, as compared to the control mice. One previous study showed that tumor specimens, from mice transplanted with a human larynx cell line and treated with Bortezomib, displayed cell nuclear condensation and tissue degradation, as well as apoptotic areas 56 . According to the in vitro results, we provided evidence that Bortezomib inhibited in vivo the expression of ErbB2 simultaneously to that of AKT in tumors. It has to be highlighted that the effect on ErbB2 expression is very likely to have a strong impact on cancer cell growth. In fact, ErbB2 can dimerize with other ErbB receptors expressed by tumor cells, thus conferring a proliferative advantage of cells over homodimerization-induced growth. AKT inhibition by Bortezomib in vivo was previously observed in homogenates from tumors of a tongue squamous cell carcinoma cell line transplanted in mice 46 . Furthermore, we showed that Bortezomib induced an increased number of tumor-infiltrating immune cells, as indicated by the increase in both T and B lymphocytes, macrophages, and NK cells in transplanted (SALTO-5) cells from Bortezomib-treated mice. On the other hand, the increase of tumor-infiltrating immune cells in mice could be triggered by the large areas of cells necrosis induced by Bortezomib and therefore could represent an indirect effect of the drug. Nevertheless, the immune response elicited in the Bortezomib-treated mice could thus cooperate with the drug to inhibit tumor growth in vivo. In addition, we also investigated for the first time T cells activation and IFN-γ production of intratumoral and spleen lymphocytes from Bortezomib-or PBS + DMSO-treated mice transplanted with SALTO-5 cells. Our results showed that there was not a significant difference in the percentage of activated infiltrating lymphocytes and their IFN-γ production between Bortezomib-and DMSO-treated mice. This finding indicates that Bortezomib induces an increased number of tumor infiltrating T cells but not an increase of their functional activity. We also reported, in agreement with a previous study 45 , that the inhibition of the tumor growth by Bortezomib was associated with a decrease in vessel density. For what concerns the in vivo tumor expression of proteasome subunits and ubiquitin, it is first worth underscoring that the immunostaining of two constitutive structural subunits, such as PSMA4 and PSMD4, was very faint in cancer cells of DMSO-treated animals. Interestingly, Bortezomib treatment induced a robust increase of immunostaining strengthening the hypothesis that the apparently low basal content of constitutive proteasome is critical for cancer cells. Thus, it could be due to a compensatory mechanism that the cells may have adopted to survive. Nevertheless, ubiquitin immunostaining was similar between the two groups of mice. Although ubiquitin staining often is a read-out of proteasome bulk proteolytic activity, it is worth underscoring that it is hard to discriminate between free ubiquitin (present at µmol/L concentration inside the cell) and ubiquitin conjugated to substrates. However, it can be likely stated that no accumulation of ubiquitin occurs in the presence of Bortezomib suggesting that compensatory mechanisms for the turnover of the natural substrates processed through the UPS take place. Unexpectedly, PSME1 expression, which encodes for the α subunit of PA28, was high in DMSO-treated cancer cells and very low in Bortezomib-treated cells 15 . The expression of PA28, an alternative regulatory particle which triggers 20S gate opening but has not ATP-dependent activity, is often induced in the presence of metabolic and pharmacological insults and it has been further reported as an adaptative mechanism of cancer cells treated with proteasome inhibitors in vitro 41 . Hence, this evidence raises some questions, that cannot be addressed at this stage, about the basal expression of proteasome sub-populations in this cancer and the metabolic and pharmacological implications this feature entails.
Overall, our results showed that anti-cancer activities of Bortezomib in tongue, pharynx and salivary gland cancer cells were dependent on cell line histotype and associated with the different extent of proteasome inhibition. The inhibition of proteasome was in turn associated with the modulation of the main signaling transduction pathways involved in HNC cellular transformation. Furthermore, for the first time we showed that Bortezomib displayed in vitro and in vivo antitumor activities in an adenocarcinoma of the salivary gland. The inhibition of tumor growth by Bortezomib was associated with tumor necrosis and apoptosis, with the simultaneous inhibition of ErbB2, AKT, and with the induction of a strong intratumoral immune response.
Our in vitro and in vivo findings further support the use of the proteasome inhibitor Bortezomib for the treatment of HNSCC and adenocarcinomas of the salivary gland and reveal its ineffectiveness in counteracting the activation of deregulated specific signaling pathways in HNC cell lines when resistance to proteasome inhibition is developed, thus suggesting the combined use of Bortezomib and specific drugs targeting signaling transduction pathways unaffected by Bortezomib treatment.

Materials and methods
Reagents. DMSO  www.nature.com/scientificreports/ lini (University of Bologna) and kept in DMEM containing 20% fetal bovine serum (FBS). SALTO-5 cells were established from salivary carcinoma arising in BALB-neuT transgenic male mice hemizygous for the p53 172R-H transgene driven by the whey acidic protein promoter 92 . Bortezomib was dissolved in DMSO. For treatments, cells were incubated for the indicated times in the presence of Bortezomib (dose range 6.25-100 nM) or vehicle control (DMSO ≤ 0.1).
Sulforhodamine B (SRB) assay. SRB assay was used to investigate cell proliferation, by measuring the cellular protein content of adherent cultures. SRB is a dye which stoichiometric binds to basic aminoacids under mild acidic conditions and dissociates using basic conditions 93 . Cells were seeded at 5 × 10 3 cells/well in 96-well plates and incubated at 37° C to allow cell attachment. After 24 h, the medium was changed and cells were incubated for 24, 48 and 72 h with Bortezomib (6.25-100 nM) or with DMSO (amount equivalent to that administered at the highest concentration of Bortezomib). Cells were then fixed with cold trichloroacetic acid (final concentration 10%) for 1 h at 4° C. The assay was then performed as previously described 94 . Briefly, after four washes with distilled water, the plates were air-dried and stained for 30 min with 0.4% (wt/vol) SRB in 1% acetic acid. After four washes with 1% acetic acid to remove the unbound dye, the plates were air-dried and cell- Equal loading and transfer of proteins was verified by Ponceau red staining of the membranes and by analyzing actin expression. The assay was then performed as previously described 101 . A densitometric analysis of autoradiographic bands was performed with the ImageJ 1.53e software (NIH, MD, USA) after blot scanning and expressed as bar graphs in the figures.
Native gel electrophoresis. Crude cell extracts (e.g., soluble fraction of the cell cytosol) were extracted under non-denaturing conditions through freeze-thawing cycles in 250 mM sucrose, 20% glycerol, 25 mM Tris-HCl, 5 mM MgCl 2 , 1 mM EDTA, 1 mM DTT, 2 mM ATP, pH 7.4. Thereafter, lysates were cleared by centrifugation at 13,000 rpm, 20 min, 4 °C and the protein concentration was determined by Bradford assay. For each experimental condition, 75 µg of proteins were separated under native conditions employing 3.5% acrylamide gel. Gels were then harvested in a clean dish and soaked in the reaction buffer (50 mM Tris, 5 mM MgCl 2 , 1 mM ATP, pH 7.5), which had been supplemented with 75 µM 7-amino-4-methylcoumarin (AMC) labeled Suc-Leu-Leu-Val-Tyr -AMC peptide (referred to as LLVY-AMC) (Boston Biochem, Boston, USA), a highly specific fluorogenic substrate of the proteasome chymotrypsin-like activity. This enzymatic proteolytic activity, which has been proven to be linearly correlated with the light intensity, was then recorded through a gel-documentation system (excitation λ = 365 nm; emission λ = visible) 102 . Proteins were then transferred to a HyBond-ECL nitrocellulose filters and probed with an antibody which recognizes an epitope shared by α1-7 subunits, but not by α4 (hereafter referred to as pan-α-subunits) (Protein-tech Group, Manchester, UK), diluted 1:3000 in 0.02% Tween-PBS fat-free milk, and then incubated with a Horseradish Peroxidase-conjugated antirabbit or anti-mouse IgG antibody (Biorad, Hercules, CA, USA), diluted 1:50,000 in 0.2% Tween-PBS fat-free milk.
Treatment of BALB-neuT mice with Bortezomib. Transgenic  www.nature.com/scientificreports/ 400 µl PBS + DMSO, twice a week) or with vehicle only (400 µl PBS + DMSO, twice a week) one week after the SALTO-5 tumor challenge. Mice were sacrificed at the first signs of distress 105 . In addition, another group of mice (n = 3) were treated i.p. with Bortezomib (0.5 mg/kg in 400 µl PBS + DMSO, twice a week) for 4 weeks to assess the toxicity of the drug. After 4 weeks, liver, lung, kidney, heart, and spleen were collected from these mice for histological examination after hematoxylin/eosin staining using 3 μm thick paraffin sections.

Analysis of antitumor activity in vivo.
Tumor growth was monitored weekly until tumor-bearing mice were sacrificed when the tumor exceeded a 20 mm width by cervical dislocation. Tumors were measured by a caliper in two perpendicular dimensions, and the volumes were calculated using the formula: width 2 x length/2 97,106 .
Histological analysis of tumors from mice treated with Bortezomib by optical microscopy. At sacrifice, tumors from three animals from each group of mice were used for histological examination after hematoxylin/eosin staining using 3 μm thick paraffin sections. Necrotic areas were measured using ImageJ 1.53e software on 10 microscopic fields. IHC was used to analyze the presence of caspase 3-positive cells (apoptotic cells) and the expression of ErbB2, AKT and phospho-AKT in tumors from PBS + DMSO-and Bortezomib-treated mice 107 . For IHC, antigen retrieval was performed on 3 μm thick paraffin sections using EDTA citrate, pH 7. IHC evaluation of CD34 was evaluated by counting the number of positive vessels on 5 high power fields (HPF) (20x) by two investigators in a blind fashion. A combined scoring system was used for the IHC evaluation of PSMA4, PSMD4, PSME1 and ubiquitin. Specifically, the total score (0-6) was obtained by adding the score associated to the number of positive cells and the score related to the signal intensity. The score associated to the number of positive cells was defined as follow: 0 (1 ≤ Positive cells/HPF), 1 (2 ≤ x ≤ 10 Positive cells/HPF), 2 (11 ≤ x ≤ 20 Positive cells/HPF), 3 (≥ 21 Positive cells/HPF). The score related to the signal intensity was defined as follow: 0 (absent/very low Intensity/HPF), 1 (low Intensity/HPF), 2 (moderate Intensity/HPF), 3 (high Intensity/HPF). The interobserver reproducibility was > 95%. Sections were observed and photographed by Olympus BX53 light microscope or Zeiss Axioscope 5 98,106,108 .
Cell extraction from mice tumor and spleen tissues and flow cytometry assay. At 5 weeks after SALTO transplantation, four hours after the last Bortezomib treatment, tumors from three animals from each group of mice were collected. To extract tumor-infiltrating lymphocytes (TIL), mouse tumors were mechanically dissociated in PBS 2% FBS onto a MACS SmartStrainers (70 µm) in a Petri dish, and a single cell suspension was obtained. Then leukocytes were enriched through 40/80 Percoll (GE Healthcare) density gradient, collecting cells at the interface between 40 and 80% Percoll solutions. Splenocytes were obtained by mechanical dissociation, followed by incubation with Red Blood Cell Lysis Buffer (Roche) for 10 min at room temperature for erythrocyte lysis 109 . Cells (10 6 ) were stained with Fixable Viability Dye eFluor780 (eBioscience), and the following antibodies were used: CD4 FITC (clone GK1.5, eBioscience), CD49b PE (clone DX5, BD Biosciences), CD69 PECy7 (clone H1.2F3, BD Biosciences), CD3 AF647 (clone 17A2, BD Biosciences), CD8a BrilliantVio-let785 (clone 53-6.7, BioLegend www.nature.com/scientificreports/ in PBS pH 7.4, and the samples were processed for ultrastructural analysis following routine procedures and observed by a Morgagni 268D transmission electron microscopy 110 . Statistical analysis. The percentage of cell survival, different phases of the cell cycle and of cell death were preliminarily verified using the Kolmogorov-Smirnov test, and the data sets were analyzed by one-way analysis of variance (ANOVA) followed by the Newman-Keuls test. Differences in the intensity of immunoreactive bands were evaluated by a two-tailed Student's t-test or one-way ANOVA followed by Tukey's post-hoc significance test. Values with p ≤ 0.05 were considered significant. Survival curves and tumor volumes were analyzed using the Kaplan-Meier method and compared with a log-rank test (Mantel-Cox). Differences in tumor volumes were regarded as significant when the p-value was ≤ 0.05. Differences in the IHC score were evaluated by a two-tailed Student's t test. Values with p ≤ 0.05 were considered significant.