Human Endogenous Retrovirus Type K Promotes Proliferation and Confers Sensitivity to Antiretroviral Drugs in Merlin-Negative Schwannoma and Meningioma

Deficiency of the tumor suppressor Merlin leads to the development of multiple tumors of the nervous system such as schwannomas, meningiomas, and ependymomas, which occur spontaneously or as part of the hereditary disease neurofibromatosis type 2 (NF2; ref. 1). NF2 commences in childhood/early adolescence, and it is common that patients develop multiple tumors simultaneously. Current treatments for Merlin-deficient tumors are restricted to surgery or radiosurgery, which have limitations when tumors occur at multiple sites or are situated where resection would risk neurological complications. An international meeting of researchers, clinicians, pharmaceutical companies, and patient advocates has stressed the urgent need to accelerate clinical trials (1). The fastest way toward clinical trials is drug repurposing, an approach pursued by the NF2 research community (2).

In this study, we investigate Merlin-deficient schwannomas and meningiomas, and a potential therapeutic target called human endogenous retrovirus (HERV) type K lineage HML2 (HERV-K).

HERVs are the results of ancient germline retroviral infections that have been transmitted over the generations in a Mendelian fashion, and in total they comprise 8% of the human genome, with HERV-K consisting of approximately 100 individual proviruses (3). HERV-K proteins—Env, Rec, and Np9—are linked to tumorigenesis (4–8) and are upregulated in a variety of cancers (9). Several FDA-approved HIV protease inhibitors appear to both affect HERV-K (10, 11) and show promise as anticancer drugs via other mechanisms (12), such as inhibition of retinoblastoma (RB1) and phospho-AKT (pAKT) and downregulation of S phase genes (13).

We use an in vitro model for Merlin-deficient tumors consisting of patient-derived tumor cells that are cultured up to five passages. This model, in contrast to immortalized cell lines, more closely represents the in vivo tumor, thus facilitating the translation of in vitro studies to phase 0 clinical trials (14–17). We demonstrate that (i) HERV-K Env, Gag, Rec, and Np9 proteins are overexpressed in human Merlin-negative schwannoma (Sch-NF2^−/−) and in all meningioma grades, (ii) ectopic Env overexpression in Schwann cells (Sch-NF2^+/+) upregulates mitogenic pERK and c-Jun protein levels and increases proliferation, (iii) HERV-K Env is upregulated by Merlin deficiency via the CRL4^DCAF1 [cullin 4 (CUL4) RING-type E3 ubiquitin ligase (CRL4)/DNA damage-binding protein 1 (DDB1) and CUL4-associated factor 1 (DCAF1)] and YAP/TEAD (Yes-associated protein/TEA domain transcription factor) pathway, and (iv) three FDA-approved antiretroviral drugs ritonavir, atazanavir, and lopinavir decrease both Sch-NF2^−/− and Merlin-negative grade I meningioma (MN-GI-NF2^−/−) cell proliferation.

Schwannoma and meningioma tissues were obtained from Derriford Hospital (Plymouth, UK) and Southmead Hospital (Bristol, UK) under local research and development (R&D) approval (Plymouth Hospitals NHS Trust: R&D No. 14/P/056 and North Bristol NHS Trust: R&D No. 3458). Normal peroneal nerve tissues (NNT) and frozen/paraffin-embedded schwannoma and meningioma tissues were obtained from BRAIN UK (Neuropathology Department, Derriford Hospital), normal Schwann cells (sural nerve) from postmortem donors (Derriford Hospital), normal meningeal tissues (NMT) from Novus Biologicals and Analytical Biological Services Inc., and human meningeal cells (HMC) were from ScienCell Research Laboratories. All meningiomas were graded by a neuropathologist. Participants provided written informed consent, and the study was conducted in accordance with the Declaration of Helsinki under Institutional Review Board approval. Patient data are given in Supplementary Tables S1 and S2. Schwannomas and Schwann cells were cultured as previously described (14). All schwannoma cells and tissues used in this study are Merlin-negative, and all Schwann cell cultures are S100 positive. Grade I meningioma cells were cultured as previously described (18). All experiments except for some IHC were performed in Merlin-negative meningiomas. The human embryonic kidney (HEK) 293T cell line was grown in DMEM supplemented with 10% FBS and 100 U/mL penicillin/streptomycin at 37°C (5% CO₂).

Ritonavir (cat# SML0491), atazanavir (cat# SML1796), lopinavir (cat# SML1222), sorafenib (cat# SML2653), and selumetinib (cat# AMBH2D6F1825) were from Sigma-Aldrich. Verteporfin (cat# 5305) was from Tocris Bioscience and Bio-Techne, and VT107 from Vivace Therapeutics.

CRL4^DCAF1 shRNA and scramble shRNA lentiviruses were provided by J. Lyons-Rimmer (Plymouth University, Plymouth, UK) or purchased from Santa Cruz Biotechnology (cat# sc-76898-V and cat# sc-108080). A HERV-K Env–expressing lentiviral vector and the empty vector were a kind gift from M. Dewannieux (Gustave Roussy Institute, Villejuif, France; ref. 7). Lentiviral particles were produced by cotransfection of HEK 293T cells with the lentiviral vector, packaging plasmids (pCMV-DR8.2; pVSV-G) in combination with MegaTran 1.0 (cat# TT200005, Origene) mixed in Opti-MEM (cat# 31985062, Thermo Fisher Scientific). Cells were incubated with lentiviral particles and 16 to 20 μg/mL protamine sulfate (cat# 107689; Sigma-Aldrich) for 72 hours, followed by selection with either 63.2 μg/mL Hygromycin B (cat# 10687010; Thermo Fisher Scientific; Env overexpression) or 4.0 μg/mL puromycin (cat# P9620; Sigma-Aldrich; CRL4^DCAF1 knockdown).

Western blotting (WB) was performed as previously described (19) using anti–HERV-K Env (cat# HERM-1811-5; AMSBIO), anti–HERV-K Gag (cat# HERM-1841-5; AMSBIO), anti-phospho ERK (cat# V803A; Promega), anti-ERK (cat# 4695; New England Biolabs), anti-CRL4^DCAF1 (cat# 11612-1-AP; Proteintech), anti-CTGF (cat# ab6992; Abcam), anti-YAP (cat# 14074; New England Biolabs), anti-Pan TEAD (cat# 13295; New England Biolabs), anti-CD63 (cat# 10628D; Thermo Fisher Scientific), and anti-Merlin (cat# 6995; New England Biolabs) antibodies. For detection, secondary horseradish peroxidase (HRP)–conjugated antibodies (cat# 170-6516 and cat# 172-1019; Biorad) and Pierce ECL or Pierce ECL Plus substrates (cat# 32209 and cat# 32132 × 3; Thermo Fisher Scientific) were used. Anti-GAPDH (cat# MAB374; Merck Millipore) and anti-tubulin α (ab4074; Abcam) antibodies were used for loading controls. WB band densities were quantified using ImageJ software.

Immunocytochemistry (ICC) was performed as previously described (19) using anti-Rec and anti-Np9 polyclonal sera (kindly provided by F. Grässer, Universitätsklinikum des Saarlandes, Homburg, Germany) and anti–HERV-K Env (cat# HERM-1811-5; AMSBIO), anti–HERV-K Gag (cat# HERM-1841-5; AMSBIO), anti-CD63 (cat# 10628D; Thermo Fisher Scientific), anti-CD9 (cat# orb235075; Biorbyt), and anti–c-Jun (cat# 9165; New England Biolabs) antibodies.

Proliferating or apoptotic cells were detected using anti-Ki67 (cat# M7420; Agilent) and anti–Cleaved Caspase 3 Asp 175 (cat# 9661; Cell Signaling Technology) antibodies, respectively. Secondary goat-anti-mouse Alexa fluor 488 or 594 (cat# A11001 and cat# A11005, Thermo Fisher Scientific) and goat-anti-rabbit Alexa fluor 488 or 568 (cat# A11008 and cat# A11011; Thermo Fisher Scientific) were used for detection. DAPI (cat# D9542, Thermo Fisher Scientific) counterstained nuclei were used for visualization.

The 5-μm paraffin-embedded tissue sections were deparaffinized, pretreated in Tris/EDTA buffer (2.4 mg/mL Tris, 0.2 mg/mL EDTA, 2 mmol/L HCl, pH9.0) for anti–HERV-K Gag and in citrate buffer (2.1 mg/mL citric acid, 10 mmol/L NaOH, pH 6.0) for anti–HERV-K Env, and heated for 30 minutes. Tissue preparations were then incubated overnight with the primary antibodies (1:50). VECTASTAIN Elite ABC HRP Kit Universal (cat# PK-6200; Vector Laboratories) was used for detection.

Merlin wild-type (NF2-Ad) and control GFP (GFP-Ad–containing adenovirus vectors were a kind gift from J. Testa (Fox Chase Cancer Center, Philadelphia, PA). Cells were infected for 24 hours, followed by incubation in fresh culture medium for an additional 48 hours. Successful infection was determined by the presence of GFP, and Merlin expression was quantified by WB.

Cell lysates were treated with 5% sodium dodecyl sulfate, 1 mol/L dithiothreitol, 0.5 mol/L sodium phosphate buffer (pH 7.5), 10% Triton X-100, and PNGase F (cat# V4831; Promega) at 37°C for 1 to 3 hours.

Cells were cultured for 7 days in medium containing exosome-depleted FBS (cat# EXO-FBS-250A-1; System Biosciences). Exosomes were isolated using Total Exosome Isolation Reagent (cat# 4478359; Thermo Fisher Scientific). Exosomes were lysed with RIPA buffer supplemented with protease and phosphatase inhibitors (18). HERV-K Env and CD63 (exosome marker; ref. 20) levels were assessed by WB. Tubulin was used as a cytoplasmic marker, and Colloidal Gold Total Protein Stain (cat#1706527; Biorad) was used as a loading control.

TEAD-binding motifs (TGGAAT) were searched within HERV-K promoter regions using an alignment of proviruses (3) followed by chromatin immunoprecipitation sequencing (ChIP-seq; CD Genomics).

Images were acquired with Zeiss LSM510 (Zeiss) and Leica SPE (Leica Microsystems) confocal units attached to Zeiss Axiovert and Leica IM8 microscopes, respectively. Colocalization was performed using z-stacks. Proliferation assays used a 20× air objective, and all other experiments were imaged using a 40× oil pH2 objective. Zeiss image manipulation software (ZEN) was used for editing.

Except for the IHC data, unpaired Student two-tailed t tests and one-way ANOVA with post hoc Tukey statistical tests were used. Experiments were performed using samples from at least three different individuals. Mann–Whitney U tests were used on IHC data (GraphPad Prism). Scoring of IHC staining was done blind and was as follows: 0 = negative, 1 = weakly positive, 2 = moderately positive, and 3 = strongly positive. Statistical values are as follows: ns (not significant), P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001. In figures, the mean ± SEM is given.

First, we determined HERV-K protein expression in human schwannoma tissues and primary cells, as well as control Schwann cells, by IHC, ICC, and WB. In IHC, 9 of 10 Merlin-negative schwannoma (Sch-NF2^−/−) tissues and 4 of 10 Merlin-positive (NF2^+/+) normal nerve tissues (NNT) stained positively for HERV-K Env, with the staining intensity in schwannomas usually higher (Mann–Whitney U test, P = 0.009; Fig. 1A). Correspondingly, ICC revealed strong HERV-K Env staining (in the cytoplasm and at the cell membrane) in both permeabilized and nonpermeabilized Sch-NF2^−/− cells in vitro, but staining was negligible in normal Schwann cells (Sch-NF2^+/+; Fig. 1B). Increased nuclear staining of HERV-K Rec and Np9 was also observed in Sch-NF2^−/− compared with Sch-NF2^+/+ cells (Fig. 1C).

The upregulation of HERV-K Env was verified by WB. HERV-K Env is translated as a full-length Env (Env-FL) precursor that is cleaved into a surface unit (Env-SU) and a transmembrane unit (Env-TM), and the anti–HERV-K Env antibody used binds to the latter. WB in Sch-NF2^+/+ cells after ectopic overexpression of Env (Supplementary Fig. S1A) confirmed the presence of both Env-FL and Env-TM proteins at the expected sizes of approximately 98 and 36 kDa, respectively (21). We also confirmed that Env-FL is glycosylated by using Peptide N-glycosidase (PNGase) treatment in Sch-NF2^−/− cells (Supplementary Fig. S1B; ref. 21). WB in Sch-NF2^−/− tissues demonstrated significant upregulation of both Env-TM and Env-FL (Fig. 1D–F), as well as Rec and Np9 (Fig. 1D, G, and H) proteins in Sch-NF2^−/− tissues compared with very weak expression in NNT. WB in Sch-NF2^−/− primary cells confirmed the above, revealing significant Env-FL and Env-TM upregulation in tumor cells compared with normal Sch-NF2^+/+ cells (Fig. 1I and J).

Merlin reintroduction into Sch-NF2^−/− cells using a Merlin-expressing adenovirus (NF2-Ad) significantly reduced the expression of HERV-K Env-FL by approximately 30% and Env-TM by approximately 60% (Fig. 1K and L).

In addition to the potentially tumorigenic proteins—Env, Rec, and Np9—we confirmed that HERV-K is upregulated in Sch-NF2^−/− cells and tissues using another HERV-K protein, namely Gag. IHC demonstrated moderate or strong expression of HERV-K Gag in 13 of 15 Sch-NF2^−/− tissues compared with only 3 of 15 NNT (Mann–Whitney U test, P = 0.0002; Supplementary Fig. S1D), which was confirmed by ICC in primary Sch-NF2^−/− cells (Supplementary Fig. S1E). WB also demonstrated significant HERV-K-Gag-FL (full-length Gag) overexpression in Sch-NF2^−/− cells (the increase in mean HERV-K p15+CA/CA+NC (Gag sub-products) expression was not significant; Supplementary Fig. S1F and S1G).

We selected c-Jun and pERK as mitogenic/tumorigenic markers in schwannoma. The transcription factor c-Jun is a master regulator of Schwann cell (Sch-NF2^+/+) differentiation, and its expression strongly increases following nerve injury, which results in reactivation of proliferation (22). C-Jun and its upstream activator, phosphorylated c-Jun N-terminal kinase (pJNK), are highly expressed in Merlin-negative schwannoma (Sch-NF2^−/−) cells compared with normal Merlin-positive (Sch-NF2^+/+) cells and contribute to increased cell proliferation and survival (23, 24). pERK is also strongly activated in Sch-NF2^−/− cells and, in addition to pJNK, is a component of a key mitogenic pathway (15, 17).

We demonstrate that ectopic overexpression of Env in Sch-NF2^+/+ cells (Fig. 2A) did indeed increase proliferation (percentage of total cells monitored by DAPI that were Ki67+; Fig. 2B and C), upregulated c-Jun expression (Fig. 2D and E), and increased levels of active pERK (Fig. 2F and G).

The involvement of HERV-K Env in Sch-NF2^−/− cell proliferation is further supported by our observation that the monoclonal anti–HERV-K Env antibody added to cell culture medium significantly reduced proliferation (Fig. 2H and J) and decreased the activity of pERK (Fig. 2L and M) in Sch-NF2^−/− cells. Importantly, no effect of the antibody on normal Sch-NF2^+/+ cells proliferation was observed (Fig. 2I and K).

The observation that the anti–HERV-K Env antibody reduced proliferation in Merlin-negative schwannoma (Sch-NF2^−/−) cells suggested that HERV-K Env action may also involve autocrine or paracrine signaling. Retroviral Env proteins are responsible for attachment of the viral particles to host cells, and the physiologically important Env proteins of another HERV family, Syncytin-1 and Syncytin-2, are crucial for exosome binding and internalization (25). We therefore investigated whether HERV-K Env is present in Sch-NF2^−/− exosomes. ICC and confocal microscopy showed some HERV-K Env colocalization in the cytoplasm with the late endosome/exosome marker CD9 (Fig. 3A). WB of exosomes extracted from Sch-NF2^−/− cell culture medium collected after 7 days of culture, using the additional late endosome/exosome marker CD63, revealed that the HERV-K Env protein is released from Sch-NF2^−/− cells via exosomes and that the release is significantly increased in Sch-NF2^−/− cells compared with Sch-NF2^+/+ cells (Fig. 3B and C). Note, the levels of Env-FL could not be measured due to a comigrating band seen in negative control samples (which is the exosome fraction from culture medium not exposed to cells; Supplementary Fig. S1C). No Env protein was detected in exosome-free supernatant fractions collected after exosome isolation (Fig. 3B).

To understand why HERV-K is overexpressed in Merlin-deficient tumors, we first investigated NF-κB, which is strongly overexpressed in Merlin-negative schwannoma (Sch-NF2^−/−) cells due to Merlin-deficiency and activates mitogenic signaling pathways (17). NF-κB is known to bind to and stimulate HERV-K expression (26). However, by using the NF-κB inhibitor SN50 at a concentration that inhibits NF-κB translocation into the nucleus and activation, and target gene expression (Supplementary Fig. S2A–S2D; ref. 7), we demonstrated that NF-κB is not involved in HERV-K expression in Sch-NF2^−/− cells.

Next, we investigated if the CRL4^DCAF1 and YAP/TEAD pathways, both of which are activated in Merlin-deficient tumors (27), are involved in the increased expression of HERV-K. Depletion of CRL4^DCAF1 by shRNA knockdown significantly decreased Env-FL expression (Fig. 4A, B, and K; by ∼20%–30% with one construct and by ∼50%–60% with a second construct). Inhibition of the downstream YAP/TEAD interaction, which prevents TEAD-mediated transcription, using either the YAP inhibitor verteporfin (which promotes 14-3-3σ/YAP sequestration in the cytoplasm and its subsequent degradation; ref. 28) or a novel TEAD-specific inhibitor VT107 (Vivace Therapeutics; ref. 29) decreased expression of the HERV-K Env proteins by approximately 40% to 50% (Fig. 4C, D, and I–K). The efficacy of the drugs was demonstrated by decreased expression of CTGF (a target gene of YAP/TEAD transactivation), YAP, and Pan-TEAD (Fig. 4C, I, and J). In addition, verteporfin decreased proliferation (Ki67) in Sch-NF2^−/− cells (Fig. 4E and F) but had no effect on the proliferation of normal Schwann cells (Sch-NF2^+/+; Fig. 4G and H).

Further downstream involvement of this pathway was suggested by the presence of a TEAD-binding site on the HERV-K sequence in silico (Supplementary Fig. S3); however, ChIP-seq detected only a 12% increase in the number of HERV-K matches compared with the experimental control.

Thus, HERV-K upregulation is at least in part triggered by CRL4^DCAF1 and YAP/TEAD Hippo pathway deregulation due to Merlin deficiency.

We tested whether three FDA-approved retroviral protease inhibitors—ritonavir, atazanavir, and lopinavir—have an antiproliferative effect on Merlin-negative schwannoma (Sch-NF2^−/−) cells. All have been reported to have affinity for the HERV-K protease (10, 11).

Ritonavir decreased Sch-NF2^−/− cells proliferation (Ki67) with an IC₅₀ of 2.9 μmol/L (Fig. 5A and C; Supplementary Table S3). This is 7.6-fold lower than the clinically reported peak plasma concentration (Cmax) of 22 μmol/L, and 3.6-fold lower than the trough plasma concentration (Cmin) of 10.4 μmol/L observed in patients with HIV, without side effects (30). Importantly, ritonavir had no effect on normal Schwann cells (Sch-NF2^+/+) proliferation (Fig. 5B and C). In addition, ritonavir significantly downregulated two major proliferation markers in schwannoma: pERK, with IC₅₀ = 1.35 μmol/L, and cyclin D1 (17), with IC₅₀ = 2.31 μmol/L (Fig. 5D; Supplementary Fig. S4A; Supplementary Table S3). As expected, ritonavir appears to inhibit the HERV-K protease in Sch-NF2^−/− cells, causing an increased expression of the uncleaved Gag-FL precursor protein and decreased expression of the second main band, which we interpret as representing cleaved p15+CA (15 kDa protein + capsid) and/or cleaved CA+NC (capsid + nucleocapsid) proteins (our antibody binds to CA) with IC₅₀ = 1.31 μmol/L (Fig. 5D; Supplementary Fig. S4B; Supplementary Table S3; ref. 31). In contrast to Gag, the Env-FL precursor protein is cleaved by a human furin protease rather than the retroviral protease (32), and ritonavir treatment decreased the expression of both Env-FL (IC₅₀ = 1.23 μmol/L) and Env-TM (IC₅₀ = 0.55 μmol/L; Fig. 5D; Supplementary Fig. S4C; Supplementary Table S3).

Lopinavir was as effective as ritonavir and decreased Sch-NF2^−/− cells proliferation with IC₅₀ = 3.66 μmol/L, which is approximately 4.7-fold lower than the plasma Cmax (∼17 μmol/L) and approximately 2.6-fold lower than Cmin (∼9.4 μmol/L) assessed by a pharmacokinetics study (Fig. 5E and G; Supplementary Table S3; ref. 33). In addition to inhibiting cell proliferation, lopinavir decreased pERK with IC₅₀ = 1.26 μmol/L (Fig. 5H; Supplementary Fig. S4D; Supplementary Table S3), inhibited Gag-FL cleavage and decreased cleaved p15+CA and/or CA+NC with IC₅₀ = 1.38 μmol/L (10), and decreased the levels of Env-TM (IC₅₀ = 8.78 × 10^–3 μmol/L; Fig. 5H; Supplementary Fig. S4E and S4F; Supplementary Table S3). Atazanavir appears to be less effective than ritonavir and lopinavir. This drug decreased Sch-NF2^−/− cell proliferation with IC₅₀ = 7.38 μmol/L, which is approximately 1.8-fold higher than the Cmax (∼4.1 μmol/L) and approximately 10-fold higher than Cmin (∼0.7 μmol/L; Fig. 5F and G; Supplementary Table S3; ref. 33).

Ritonavir, in addition to its antiretroviral properties, can also inhibit CYP3A4 and thereby boost efficacy of drugs that are metabolized by CYP3A4 (34). We therefore investigated whether ritonavir would increase the efficacy of the MEK inhibitor selumetinib and the PDGFR/Raf inhibitor sorafenib, both of which reduce proliferation of Merlin-negative schwannoma (Sch-NF2^−/−) cells in vitro (15) and are known CYP3A4 substrates (35; 36). Sorafenib has been tested in phase 0 clinical trials in NF2 patients (37). Treating Sch-NF2^−/− cells with ritonavir in combination with selumetinib or sorafenib showed that both ritonavir + sorafenib and ritonavir + selumetinib combinations have additive effects (Fig. 5I–K).

An increase of HERV-K expression was demonstrated by WB in all grades of Merlin-negative meningioma tissues (grade I, MN-GI-NF2^−/−; grade II, MN-GII-NF2^−/−; grade III, MN-GIII-NF2^−/−) and in MN-GI-NF2^−/− primary cells compared with control NMT and HMC, respectively.

Env-FL and Np9 expression in MN-GI-NF2^−/− and MN-GII/III-NF2^−/− tissues were increased (Env-FL average ∼7-fold for G-I and average ∼10-fold for GII/III; Np9 average ∼4-fold for G-I and average ∼12-fold for G-II/III), although not significantly compared with NMT (Fig. 6A, C, D, and G). Env-TM was significantly increased in MN-GI-NF2^−/− tissues (average ∼7-fold) but not in MN-GII/III-NF2^−/− tissues (Fig. 6A and E). Rec was observed only in MN-GI-NF2^−/− and MN-GII/III-NF2^−/− biopsies, but not in NMT (Fig. 6B and F). Gag-FL expression was higher in MN-GI-NF2^−/− and MN-GII/III-NF2^−/− tissues, although not significantly (Supplementary Fig. S5A and S5B). Gag products p15+CA/CA+NC were significantly increased in MN-GI-NF2^−/− but not in MN-GII/III-NF2^−/− tissues compared with NMT (Supplementary Fig. S5A and S5C).

The expression of Env-FL (Fig. 6H and I) and Gag-FL (Supplementary Fig. S5D and S5E) was significantly increased in MN-GI-NF2^−/− primary cells (Env-FL average ∼3-fold; Gag-FL average ∼6-fold) compared with HMC. The difference in the expression of Env-TM (Fig. 6H and I) and p15+CA/CA+NC (Supplementary Fig. S5D and S5F) was however not significant.

IHC demonstrated moderate or strong staining for HERV-K Env in all eight MN-GI-NF2^−/− tumors (Supplementary Fig. S5G and S5I). Eight of 10 control NMT tissues were also positive for HERV-K Env but the average staining intensity was significantly weaker (Supplementary Fig. S5G and S5I; Mann–Whitney U test, P = 0.0015). HERV-K Gag staining was also moderate or strong in all eight MN-GI-NF2^−/− tissues, and the average staining intensity was significantly higher compared with NMT (Supplementary Fig. S5H and S5J; Mann–Whitney U test, P = 0.0005). We observed similar results with the higher meningioma grades for HERV-K Env (Supplementary Fig. S5G and S5I; Mann–Whitney U test, P = 0.0015 for grade II, and P < 0.0001 for grade III) but not for HERV-K Gag (Supplementary Fig. S5H and S5J; Mann–Whitney U test, P = 0.6667 for grade II, and P = 0.0561 for grade III). However, the Merlin status of some of these tissues used in IHC was not recorded (not determined, ND; Supplementary Fig. S5G–S5J).

The specificity of the anti–HERV-K Env antibody in control HMC cells was confirmed by ectopic overexpression of Env (Supplementary Fig. S5K). Env overexpression in MN-GI-NF2^−/− cells was also reversed by Merlin reintroduction (Fig. 6J and K) and was significantly decreased by the TEAD-specific inhibitor VT107 (Fig. 6L and M), confirming the involvement of CRL4^DCAF1 and YAP/TEAD Hippo pathway in the regulation of HERV-K overexpression in meningioma.

Although some intracellular colocalization of Env with the late endosome/exosome marker CD63 was observed (Supplementary Fig. S5L), no exosome-mediated Env release was detected in either HMC or MN-GI-NF2^−/− cells (Supplementary Fig. S5M).

Ritonavir, atazanavir, and lopinavir all strongly decreased proliferation (Ki67) of Merlin-negative grade I meningioma (MN-GI-NF2^−/−) cells, displaying even stronger inhibition than in Sch-NF2^−/− cells (Fig. 7A–D; Supplementary Table S4). Ritonavir decreased the number of proliferating cells with IC₅₀ = 0.61 μmol/L (∼36-fold lower than plasma Cmax and ∼17-fold lower than plasma Cmin in patients with HIV; ref. 30; Fig. 7A and D; Supplementary Table S4), atazanavir with IC₅₀ = 0.14 μmol/L (∼29-fold lower than plasma Cmax and ∼5.2-lower than plasma Cmin; ref. 33; Fig. 7B and D; Supplementary Table S4), and lopinavir with IC₅₀ = 0.88 μmol/L (∼19-fold lower than plasma Cmax and ∼9-fold lower than plasma Cmin; ref. 33; Fig. 7C and D; Supplementary Table S4). Cell viability was not affected at drug concentrations 1, 5, and 10 μmol/L (Supplementary Fig. S6A). However, at Cmax concentrations, ritonavir (22 μmol/L; ref. 30) and lopinavir (17 μmol/L; ref. 33) induced death of almost 100% of tumor cells (Supplementary Fig. S6B). Atazanavir at its Cmax concentration (4.1 μmol/L; ref. 33) was not toxic.

In addition, ritonavir, atazanavir, and lopinavir all significantly decreased active pERK (ritonavir, IC₅₀ = 8.73 μmol/L; atazanavir, IC₅₀ was not calculable, and lopinavir, IC₅₀ = 4.07 μmol/L) and cyclin D1 (ritonavir, IC₅₀ = 4.41 μmol/L; atazanavir, IC₅₀ = 6.3 μmol/L; and lopinavir, IC₅₀ = 3.05 μmol/L; Fig. 7E; Supplementary Fig. S7; Supplementary Table S4). They also inhibited the retroviral protease, causing increased expression of Gag-FL and decreased expression of p15+CA/CA+NC (ritonavir, IC₅₀ = 186.82 μmol/L, ∼35% decrease at 10 μmol/L; atazanavir, IC₅₀ = 12.51 μmol/L; and lopinavir, IC₅₀ = 6.19 μmol/L; Fig. 7E; Supplementary Fig. S7; Supplementary Table S4). All three drugs reduced the expression of both Env-FL and Env-TM [ritonavir (Env-FL IC₅₀ = 57.22 μmol/L, ∼40% decrease at 1 μmol/L; Env-TM IC₅₀ = 1.13 × 10³ μmol/L, ∼30% decrease at 10 μmol/L), atazanavir (Env-FL, IC₅₀ = 2.95 μmol/L; Env-TM IC₅₀ = 2.3 × 10³ μmol/L, ∼35% decrease at 1 μmol/L), and lopinavir (Env-FL IC₅₀ = 0.89 μmol/L, Env-TM IC₅₀ = 0.54 μmol/L; Fig. 7E; Supplementary Fig. S7; Supplementary Table S4)].

This report suggests the use of antiretroviral protease inhibitors to treat patients with Merlin-deficient schwannomas and meningiomas. In both Merlin-negative schwannoma (Sch-NF2^−/−) cells and Merlin-negative meningioma (MN-GI-NF2^−/−) cells, ritonavir and lopinavir (and, for meningioma, atazanavir) decreased proliferation, with an IC₅₀ lower than the Cmin in patients with HIV. In addition, ritonavir had no effect on proliferation of normal Merlin-positive Schwann (Sch-NF2^+/+) cells, suggesting that it is tumor selective (we lacked sufficient samples to test lopinavir and atazanavir). Interestingly, the effect of all three drugs was much stronger against meningiomas.

The pleiotropic anticancer effects of antiretroviral drugs (not just protease inhibitors) have attracted interest recently (38). For example, there have been at least 24 clinical trials involving ritonavir in a broad range of cancer therapies, five of which also involved lopinavir. Most trials have not yet reported, but a ritonavir plus lopinavir phase II trial in high-grade gliomas found that the drugs were well tolerated but did not significantly improve 6-month progression-free survival among 19 patients (39). In antiretroviral therapy, ritonavir is now used primarily in combination with atazanavir or lopinavir to boost the latter's bioavailability, e.g., by its CYP3A4-inhibitory properties mentioned above (34). We found ritonavir to have an additive effect in combination with sorafenib or selumetinib.

Treatment of tumors such as schwannomas and meningiomas, especially when occurring as part of the NF2 disease, must be for prolonged periods of time, which risks the development of long-term adverse effects. The low IC₅₀ observed in treatment of Sch-NF2^−/−cells and MN-GI-NF2^−/− cells in vitro suggests one could use a low dose in the medication regimen. Also, both schwannoma and meningioma tumors are located outside the blood–brain barrier, and thus, drug delivery should not be problematic. Our recent phase 0 clinical trial of orally administered sorafenib in NF2 patients achieved high intratumoral concentration of the drug (37).

Another possible treatment is immunotherapy with a humanized anti–HERV-K Env antibody (40). Although HERV-K Env is upregulated in Sch-NF2^−/− and MN-GI-NF2^−/− cells and tissues, it is expressed at lower levels in normal Sch-NF2^+/+ cells and meningeal cells. An immunotherapy approach would therefore need careful safety testing.

Further testing with a mouse model is impossible. There is an NF2 mouse model (41), but HERV-K occurs only in humans (3, 9). Implantation of human schwannoma into mice vestibular nerves is surgically impossible, and subcutaneous xenograft would not recapitulate the unique intraneural microenvironment for schwannoma growth.

The mechanism at play in the antiretroviral drug-driven proliferation inhibition is probably multifaceted and involves both HERV-K–dependent and –independent pathways. For example, in addition to having an inhibitory effect on HERV-K protease, ritonavir is known to inhibit proteasome activity, although only at higher concentrations (>10 μmol/L; ref. 42). In a study using human glioblastoma–derived cells GL15, ritonavir inhibited chymotrypsin-like activity of the proteasome with IC₅₀ = 50 μmol/L and significantly induced cell-cycle arrest at concentration of 100 μmol/L (42). In our study, however, ritonavir decreased proliferation of Merlin-negative schwannoma (Sch-NF2^−/−) and Merlin-negative grade I meningioma (MN-GI-NF2^−/−) cells at IC₅₀ = 2.9 μmol/L and IC₅₀ = 0.61 μmol/L, respectively. These concentrations are much lower than needed for effective proteasome inhibition and perhaps indicate that ritonavir's proteasome inhibitory effect is a minimal contribution to the inhibition of HERV-K levels.

Moreover, both ritonavir and lopinavir inhibited pERK activity, which we think may be partly due to overexpression of HERV-K. We confirmed the effect of both drugs on the HERV-K protease, which cleaves the viral Gag protein. Less expectedly, the drugs also inhibited the expression of HERV-K Env, which is cleaved by a cellular furin-like endoprotease (32). Because levels of both the uncleaved Env-FL precursor and cleaved Env-TM were decreased, the drugs did not affect the cleavage efficiency as with Gag but rather the overall expression of HERV-K Env. This is perhaps consistent with these drugs having a broad range of effects.

We demonstrated that HERV-K Env expression is dependent on the tumor suppressor Merlin. There is evidence of another HERV, HERV-E, being upregulated by inactivation of another tumor suppressor: von Hippel–Lindau protein (43). Downstream of Merlin, we suggest for the first time that HERV-K Env expression in both Merlin-negative schwannoma (Sch-NF2^−/−) cells and Merlin-negative grade I meningioma (MN-GI-NF2^−/−) is regulated by the transcription factor TEAD via binding to YAP. We observed that TEAD possesses a binding domain on HERV-K long terminal repeats (LTR) in silico. However, we were unable to confirm binding with our ChIP analysis. This failure to confirm binding might be caused by the approximately 1,000 fragments of HERV-K scattered across the genome (3).

Blocking different elements in the Hippo pathway, as well as CRL4^DCAF1, reduced but did not completely block HERV-K expression, suggesting the involvement of additional factors. One such factor could be Src, which activates YAP (44) and has been previously shown by us to be involved in increased Sch-NF2^−/− cell proliferation downstream of PDGFRβ (14) and integrin β1 (45). Other possible factors are that HERV-K loci are often silenced by methylation and have binding sites for many transcription factors, at least eight of which have been demonstrated experimentally to upregulate HERV-K (although our study allows us to exclude one, NF-κB, in schwannomas; ref. 46).

HERV-K Env has been previously demonstrated to contribute to tumorigenesis of melanoma, breast, and pancreatic cancers involving MYC, AKT, and—especially—RAS/RAF/MEK/ERK signaling pathways (4–7). We demonstrate that ectopic HERV-K Env overexpression in normal Schwann cells induces proliferation and is associated with an upregulation of phosphorylated/active ERK1/2 (pERK1/2), which is similar to that observed in schwannoma tumor counterparts. In 293T cells, Env overexpression was also associated with pERK upregulation involving the RAS/RAF/MEK/ERK pathway, a process requiring the presence of Env cytoplasmic tail (7). Which effector allows signal transduction from HERV-K Env to RAF is unknown. The cytoplasmic tail of the Jaagsiekte sheep retrovirus Env, another betaretrovirus envelope glycoprotein, harbors binding motifs for PI3K, which is involved in the PI3K/AKT pathway leading to fibroblast transformation (47). However, the Lemaitre and colleagues' study cited above (7) did not find upregulation of pAKT.

Furthermore, we observed an increase in c-Jun, suggesting that Env may also stimulate the JNK/c-Jun network. C-Jun triggers proliferation of Schwann cells after nerve injury and is overexpressed in schwannoma (22–24). The phosphorylation profile in Env-knockdown cell lines reveals downregulation of several kinases including c-Jun and JNK1/2/3 (6). Therefore, our findings are consistent with HERV-K Env contributing to schwannoma tumorigenesis by the stimulation of the RAS/RAF/MEK/ERK and JNK/c-Jun pathways, leading to increased cell proliferation. We also found increased expression of Rec and Np9 proteins in schwannoma, both of which are linked to tumor growth in other cancer types by altering pERK, Myc, and β-catenin pathways (48). We have previously shown these pathways to be involved in schwannoma development (16).

We attempted HERV-K knockdown using a large set of shRNAs targeting sites across the provirus (LTR, gag, and env). Our set included sequences that reduce Env expression in pancreatic (6) and melanoma cell lines (5). The shRNA used in pancreatic cells reduced growth of cell lines inoculated in mice and reduced pERK expression as well as pAKT, MYC, and RAS. However, we were unsuccessful.

HERV-K Env was detected at the cell membrane and in the exosomal fraction of culture medium from schwannoma cells. We speculate that HERV-K Env, which is expressed on the cell surface and has fusogenic ability (49), contributes to cell-to-cell transfer of growth factors via exosomes. Thus, HERV-K Env would facilitate uptake of exosomes that are transporting protumoral molecules. Evidence for the protumorigenic role of exosomes is accumulating (50), and exosomal release of Env proteins (syncytins) from another HERV lineage (HERV-W) has been reported to significantly increase the uptake of exosomes via receptor-facilitated endocytosis (25). Our observed antiproliferative effect of the anti–HERV-K Env monoclonal antibody might result from blocking this process.

D.A. Hilton is a neuropathologist who performed some of the IHC staining and evaluated all IHC staining in this study. K.M. Kurian runs Brain Tumour Bank South West (BRASH) and provided schwannoma and meningioma tissues used in this study. C.O. Hanemann is Associate Head of Peninsula Medical School (Research) and Director of Brain Tumour Centre at Plymouth University. He contributed with his expertise in neuro-oncology and NF2, and all research presented in this article was performed at his lab. R.D. Belshaw significantly contributed with his expertise in endogenous retroviruses and genetics. S. Ammoun significantly contributed with her expertise in NF2 and brain tumor pathobiology. No disclosures were reported by the other authors.

E.A. Maze: Formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. B. Agit: Formal analysis, investigation, visualization, methodology. S. Reeves: Investigation. D.A. Hilton: Formal analysis, investigation, visualization, methodology. D.B. Parkinson: Resources, writing–review and editing. L. Laraba: Resources, writing–review and editing. E. Ercolano: Resources, investigation. K.M. Kurian: Resources, investigation. C.O. Hanemann: Resources, supervision, writing–review and editing. R.D. Belshaw: Resources, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, writing–review and editing. S. Ammoun: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

The authors thank Jhen Tsang and Marie Dewannieux (Gustave Roussy Institute, Villejuif, France) for providing the HERV-K Env-expressing vector for overexpression experiments, Jade Lyons-Rimmer (Peninsula Medical School, University of Plymouth, UK) for providing the CRL4^DCAF1-specific shRNA-expressing vector, Friedrich Grässer (Universitätsklinikum des Saarlandes, Homburg, Germany) for the anti-Rec and anti-Np9 polyclonal sera, and Joseph Testa (Fox Chase Cancer Center, Philadelphia, PA) for the NF2 and GFP-containing adenovirus vectors.

This study was supported by 2016 Action Medical Research for Children, 2014 Action on Hearing Loss Flexi Grant [S. Ammoun (lead applicant) and R.D. Belshaw (coapplicant)], and Brain Tumour Research (C.O. Hanemann).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.