Prophylactic In Vivo Hematopoietic Stem Cell Gene Therapy with an Immune Checkpoint Inhibitor Reverses Tumor Growth in Syngeneic Mouse Tumor Models

Women who have at least one first-degree relative diagnosed with breast cancer before the age of 50 or with ovarian cancer at any age, are now referred to genetic testing. Using targeted capture and massively parallel genomic sequencing, a series of multigene tests have been established that detect germline mutations and predict the risk of cancer onset. Among these test platforms is BROCA (1, 2). Using BROCA, it has been established that more than one-fifth of ovarian and breast carcinomas are associated with inherited risk (3). The problem is that the current options for prevention in high-risk carriers lag behind the constantly improving genetic diagnostics. Side effects of prophylactic salpingo-oophorectomy and mastectomy, like infertility, cardiovascular disease, osteoporosis, menopausal symptoms, and psychologic effects, are expected throughout the woman's life. Use of serum markers such as CA125 and HE4 did not show significant reduction of ovarian cancer mortality (4). Prophylactic vaccines against tumor-associated antigens like Her2/neu, HIF1α, or MUC1 rely on the presence of these antigens on all tumor cells, and are plagued by the development of antigen-loss mutants (5).

Our goal is to develop a long-lasting and technically simple approach that allows for the immunoprophylaxis of cancer in patients with high-risk for tumor recurrence and, ultimately, in carriers of cancer-predisposing inherited mutations. During tumor progression, malignant cells secrete a number of specific chemokines that activate and mobilize hematopoietic stem/progenitor cells (HSPC), so that they enter the blood circulation and localize to the tumor where they are differentiated into tumor-supporting cells (6, 7). HSPC-derived myeloid and lymphoid cells are present in early stages of cancer development (8–10), for example, in serous tubal intraepithelial carcinoma (11). Our approach is based on the genetic modification of hematopoietic stem cells. Because these cells are capable of self-renewal, a one-time intervention should have a life-long therapeutic effect. We developed a minimally invasive and cost-efficient technology for in vivo gene delivery into HSPCs without leukapheresis, myeloablation, and transplantation (12, 13). The central idea of our approach is to mobilize HSPCs from the bone marrow using G-CSF/AMD3100, and while they circulate at high numbers in the periphery, transduce them with an intravenously injected HSPC-tropic helper-dependent adenovirus HDAd5/35++ gene transfer vector system. These vectors use CD46, a receptor that is expressed on primitive hematopoietic stem cells. Transduced cells return to the bone marrow where they persist long term. The novel features of the HDAd5/35++ vector system used in this study are: (i) CD46-affinity enhanced fibers that allow for efficient transduction of primitive HSPCs while avoiding infection of nonhematopoietic tissues after intravenous injection (including liver), (ii) a SB100X transposase-based integration system that functions independently of cellular factors and mediates random transgene integration without a preference for genes with one to two integrated vector copies per cell (Fig. 1A), and (iii) a MGMT(P140K) expression cassette mediating selective survival and expansion of progeny cells without affecting the pool of transduced primitive HSPCs by short-term treatment with low-dose O⁶BG/BCNU (14). We have recently demonstrated the efficacy and safety of our in vivo HSPC gene therapy method in mouse models for hemoglobinopathies (13, 15). Here we use this approach for prevention of cancer growth.

HDAd-SB has been described previously (12). The construction of the HDAd-GFP/mgmt and HDAd-αPDL1γ₁-miR423 vectors is described in Supplementary Information. The mouse αPD-L1-γ₁ transgene was provided by Dr. Guy Ungerechts (Department of Translational Oncology, National Center for Tumor Diseases and German Cancer Research Center, Heidelberg, Germany; ref. 16) The production of HDAd5/35++ vectors in 116 cells was described previously (17). Helper virus contamination levels were found to be <0.05%. Titers were 6–12 × 10¹² vp/mL. All HDAd vectors used in this study contain chimeric fibers composed of the Ad5 fiber tail, the Ad35 fiber shaft, and the affinity-enhanced Ad35++ fiber knob (18). All of our HDAd preparations had less than one copy wild-type virus in 10¹⁰ vp measured by qPCR using the primers described elsewhere (19).

Mouse mammary carcinoma (MMC) cells were established from a spontaneous tumor in a neu/CD46-tg mouse. MMC cell authentication was performed by immunofluorescence using the Neu-specific mAb 7.16.4 (5). TC-1 cells were from the ATCC. TC-1 cells are immortalized murine epithelial cells that stably express HPV-16 E6 and E7 proteins. C57Bl/6-derived ovarian cancer ID8 p53^−/− brca2^−/− cells were described previously (20). This cell line was generated by CRISPR/Cas9 knockout of p53 and brca2 in ID8 cells. The gene knockout was verified by sequencing of genomic DNA. MMC and TC-1 cells were maintained in RPMI1640 supplemented with 10% FCS, 1 mmol/L sodium pyruvate, 10 mmol/L HEPES, 2 mmol/L l-glutamine, 100 units/mL penicillin, and 100 mg/mL streptomycin. ID8 p53^−/− brca2^−/− cells were cultured in DMEM supplemented with 4% FCS, 100 μg/mL penicillin, 100 μg/mL streptomycin, and ITS (5 μg/mL insulin, 5 μg/mL transferrin, and 5 ng/mL sodium selenite). Absence of mycoplasma was confirmed using the PCR Mycoplasma Detection Kit from abm. For amplification, cryopreserved cells were thawed and passaged four times.

Ovarian cancer biopsies were provided by the Pacific Ovarian Cancer Research Consortium. We obtained written informed consent from the patients. The studies were conducted in accordance with recognized ethical guidelines (e.g., Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common Rule). The studies were approved by the Fred Hutchinson Cancer Research Center Institutional Review Board. Specimen Repository without any confidential information was used to identify a patient (Fred Hutchinson Cancer Research Center IRB protocol #6289). Tumor tissue from biopsies was dissected into 4 mm pieces and digested for 2 hours at 37°C with collagenase/dispase (Roche) as described previously (21). Leukocytes were isolated by magnetic activated cell sorting using human CD45 microbeads (Miltenyi Biotec, catalog no. 130-045-801). Tumor-associated leukocytes from two high-grade serous ovarian cancer biopsies were pooled and RNA was analyzed by miRNA-seq in comparison with matching peripheral blood mononuclear cell (PBMC) RNA by LC Sciences, LLC.

Small RNA-sequencing in the MMC and TC-1 models was performed as described previously (23). RNA was extracted using a miRNeasy Mini Kit (Qiagen, catalog no. 1071023). One microgram of RNA per sample was ligated to a 3′ Universal miRNA Cloning Linker (New England Biosciences, catalog no. S1315) using T4 RNA Ligase 1 (New England Biosciences, catalog no. M0204) in the absence of ATP. Ligated samples were run on a 15% urea-polyacrylamide gel. Fragments corresponding to small RNAs (17–28nt) were cut from the gel and ligated to 5′ barcodes, again using T4 RNA ligase 1. Barcoded samples were then multiplexed and sequenced on an Illumina MiSeq machine obtaining 50-bp single-end reads, at the University of Washington Center for Precision Medicine. The barcodes and adaptors were trimmed from the sequence and subsequently aligned to mouse miRNAs on miRBase using Bowtie version 0.12.7, allowing for two mismatches (22).

The protocol described previously (23) with the following ³²P-γ-ATP–labeled probes was used: for miRNA 423-5p: 5′ AAA GTC TCG CTC TCT GCC CCT CA; for U6 snRNA: 5′GAA TTT GCG TGT CAT CCT TGC GCA GGG GCC ATG CTA A. Radioactive RNA molecular weight markers were from Ambion.

Tissue lysates were separated by SDS-PAGE and blots were incubated with chicken anti-HA-tag-HRP (Abcam, ab1190). Chemiluminescence detection on X-ray films was performed after treatment with Pierce ECL Plus Western Blotting Substrate (Thermo Fisher Scientific, catalog no. 34029). CDKN1A antibodies (STJ23069, St John's Laboratory) cross-reacted with the mouse and human protein.

qRT-PCR for αPD-L1-γ₁ mRNA is described in Supplementary Materials and Methods

Recombinant mouse PD-L1 protein (Sino Biological Inc., catalog no. 50010-M08H) at 2 μg/mL were used to coat ELISA plates. Serum from test animals was added at a 1:10 dilution and αPD-L1-γ₁ was measured using chicken anti-HA-tag-HRP (Abcam, ab1190).

Flow cytometry, immunofluorescence analyses, T-cell assays, qRT-PCR, and CFY assays are described in Supplementary Information.

All experiments involving animals were conducted in accordance with the institutional guidelines set forth by the University of Washington (Seattle, WA). The studies were approved by the Institutional Animal Care and Use Committee of the University of Washington (protocol no. 3108-01).

C57Bl/6-based transgenic containing the human CD46 genomic locus and expressing CD46 at a level and in a pattern similar to humans were described earlier (24). They were used in transplantation studies with C57Bl/6 derived TC-1 cells.

Neu-transgenic (neu-tg) mice [strain name: FVB/N-Tg(MMTVneu)202Mul] were obtained from The Jackson Laboratory. These mice harbor nonmutated, nonactivated rat neu under control of the mouse mammary tumor virus promoter (one transgene copy per genome). For in vivo transduction studies, CD46tg and neu-tg mice were crossed to obtain CD46^+/+/neu⁺ mice.

Please see legend of Fig. 1A.

CD8-T cells were depleted using intraperitoneal injection of 200 μg rat anti-mouse CD8 IgG (169.4; ATTC). Injection was repeated every 3 days to maintain the depletion.

Statistical significance of in vivo data was analyzed by Kaplan–Meier survival curves and log-rank test (GraphPad Prism Version 4). Statistical significance of in vitro data was calculated by two-sided Student t test (Microsoft Excel). P values >0.05 were considered not statistically significant (n.s.).

We employed two human CD46 transgenic mouse models with syngeneic tumors (CD46 is required for HSPC transduction with HDAd5/35++ vectors). The first model were human CD46/rat neu-transgenic mice that overexpress rat Neu in breast tissue from a mouse mammary tumor virus promoter. Neu-tg mice develop active immune tolerance towards Neu, which is dependent on Tregs and is similar to what is observed in patients with breast cancer (25). Mouse MMC are a Neu-positive breast cancer cell line derived from a spontaneous neu/CD46-transgenic mouse tumor (Supplementary Fig. S1). We mobilized HSPCs in neu/CD46-tg mice and injected an integrating GFP-expressing HDAd5/35++ vector (Fig. 1A, bottom). Similar to previous studies (14), three rounds of low-dose treatment with O⁶BG/BCNU resulted in stable GFP expression in ∼80% of PBMCs (Fig. 1B). At week 17 after in vivo HSPC transduction, we implanted syngeneic MMC cells into the mammary fat pad and monitored tumor growth. When tumors reached a volume of 700 mm³, animals were sacrificed and GFP expression was analyzed. In agreement with earlier studies, there was no significant transduction of nonblood cells (12). About 80% of bone marrow cells, splenocytes, PBMCs, and tumor-infiltrating leukocytes (TIL) expressed GFP (Figs. 1B). In the tumor, GFP⁺ cells were found predominantly in tumor stroma (Fig. 1C). Immunophenotyping showed that GFP⁺ tumor-infiltrating cells were lymphocytes (predominantly Tregs), neutrophils, DCs/MDSCs, and macrophages (Fig. 1D; Supplementary Fig. S2). This pattern differed from that of GFP⁺ cells in peripheral blood (Fig. 1D), bone marrow, and spleen (Supplementary Fig. S3), indicating that tumors actively differentiate HSPCs into specialized protumor cells. Efficient recruitment of in vivo–transduced HSPCs to the tumor was further confirmed in a second model consisting of CD46-tg mice and TC-1 cells, a HPV16 E6/E7-postive mouse lung cancer cell line (Supplementary Fig. S4).

Figure 1B and Supplementary Fig. S4C illustrate that GFP (under the control of the ubiquitously active EF1α promoter) is not only expressed in TILs but also in other tissues including bone marrow, spleen, PBMCs, and resident macrophages. To minimize autoimmune reactions, our therapy approach requires that the therapeutic transgene (i) be predominantly expressed in the tumor, (ii) automatically activate only when the tumor begins to develop, and (iii) cease when the tumor disappears. These requirements can be met through miRNA regulation. During hematopoiesis, the miRNA profile changes depending on the differentiation stage and cell lineage (26). Tumor-associated myeloid cells have distinct mRNA and miRNA expression profiles (27). Finally, there is a high degree of conservation of miRNAs in myeloid and lymphoid cells found in different tumor types in humans (27). The principle of miRNA regulation of transgene expression is shown in Fig. 2A. Using the in vivo HSPC-transduced mouse models, we sorted GFP⁺/CD45⁺ cells from bone marrow, spleen, PBMCs, and tumor (see Fig. 1B; Supplementary Fig. S4C) and analyzed their miRNA expression profile. Our goal was to find miRNAs that were expressed at high levels in bone marrow, blood, and spleen cells, but were absent in tumor-associated leukocytes. We subjected total RNA (pooled from 5 mice) to next-generation miRNA sequencing (Fig. 2B and C). We identified a series of miRNAs that fulfilled the above criteria. We focused on miR423-5p, a miRNA that was on the top of the list, both in the neu/CD46tg-MMC (Fig. 2B) and in the CD46tg-TC-1 model (Fig. 2C). The expression profile of miRNA-423-5p in GFP⁺ fractions from in vivo transduced mice with MMC and TC-1 tumors was validated by miRNA array and Northern blot analysis (Supplementary Fig. S5). Furthermore, miRNA array analysis showed miR-423-5p at the top of the list of miRNAs that are expressed at high levels in blood, bone marrow, and spleen, but are absent in tumors.

To assess whether miR-423-5p regulation could also be used in humans, we examined levels of miR-423-5p in a published dataset that evaluated miRNAs across a series of human tissues (28). We found that miR-423-5p is in the top 20% of expressed miRNAs and has even distribution across tissues, including in the bone marrow and spleen (Supplementary Fig. S6A). We obtained matching PBMCs and tumor biopsies from 2 patients with high-grade serous ovarian cancer. We performed miRNA-seq on RNA from tumor-infiltrating (CD45⁺) leukocytes versus RNA from matching PBMCs (Supplementary Fig. S6B). This analysis confirmed high-level expression of miR423-5p in PBMCs and low-level expression in TILs. These data increase our confidence that the results we observe in mice have the potential to be translated to human studies.

miRNA-423-5p is expressed in all normal tissues and therefore, most likely, involved in the regulation of gene expression. A search of target mRNAs for miR-423-5p in “mirtarbase” identified the cyclin-dependent kinase inhibitor 1A (CDKN1A) mRNA as the primary target (http://mirtarbase.mbc.nctu.edu.tw/php/detail.php?mirtid=MIRT000589#target). Other target mRNAs include transcription elongation factor A like 1 (TCEAL1), bcl2 like 11 (bcl2L11), and proliferation-associated 2G4 (PA2G4). To assess whether added expression of miR-423-5p target sites from HDAd vectors influences the expression of CDKN1A, we constructed two HDAd-GFP vectors with and without the target sites linked to a GFP containing mRNA (Fig. 3A). We infected mouse and human HSPCs, that is, cell types with high level miR-423-5p expression, at MOIs that would result in the transduction of the vast majority of cells (29) and analyzed CDKN1A protein levels 3 days later by Western blot analysis (Fig. 3B). We did not find a significant difference between the two HDAd vectors in both cell types. Furthermore, no detrimental effects of miR-423-5p target site overexpression were observed in progenitor colony assays (Fig. 3C). As outlined later, in vivo HSPC transduction with a therapy vector that contained the miR423-5p target sites did not cause abnormalities in hematopoiesis. Taken together, this suggests that our miR-423-5p-based regulation system is safe in HSPCs.

In hereditary breast and ovarian cancer, genetic variants disrupt DNA repair mechanisms, resulting in higher mutational burden and neoantigen presence. This makes the tumors more amenable to immunotherapies than nonheritable breast and ovarian cancers, which are often characterized by aberrant copy number and low immunogenicity (27). Here, we choose the checkpoint inhibitor αPD-L1-γ₁ as our immunotherapeutic transgene. Previously, it was shown that intratumoral αPD-L1-γ₁ expression after viral gene transfer resulted in tumor growth attenuation (16, 30). In MMC cell cultures, we observed strong PD-L1 expression (Fig. 4A), which should make MMC tumors susceptible to αPD-L1-γ₁ therapy. We incorporated four copies of miR423-5p target sites into a globin 3′ untranslated region (UTR) linked to the αPD-L1-γ₁ gene (Fig. 4B). The experimental scheme was the same as shown in Fig. 1A. In mice that were in vivo transduced with the control HDAd-GFP-miR423 vector, implanted MMC tumors grew rapidly and reached the endpoint volume by day 35 after tumor cell transplantation (Fig. 4C, left). In the αPD-L1-γ₁ model, after initial tumor growth, 6 out of 7 tumors regressed and did not recur within the observation period (100 days; Fig. 4C, right). Treated mice rejected another challenge of MMC cells given 11 weeks after the first injection. Depletion of CD8 cells by anti-CD8 mAb injections abolished the therapeutic effect. Antitumor T-cell responses were measured at the end of the observation period (day 100). Analysis of splenocytes by flow cytometry showed a significant higher percentage of IFNγ-producing CD4 and CD8 cells as well as a higher frequency of CD8 cells that stained positive with a Neu-tetramer (Fig. 4D). Splenocytes from HDAd-αPDL1_γ1-miR423-treated animals exhibited approximately 30-fold greater IFNγ secretion upon stimulation with (Neu-positive) MMC cells, compared with Neu-negative cells (Fig. 4E). As expected, naïve CD46/neu-tg mice possessed Neu-specific T cells, which however could not control tumor growth due to the presence of immunosuppressive T cells in the tumor (25).

In a separate group of HDAd-αPDL1γ₁-miR423-treated animals, we harvested tumors at day 17 after implantation before they started to shrink. In these tumors (∼300–400 mm³), we observed about 10-fold higher levels of αPD-L1-γ₁ than in PBMCs, bone marrow, and spleen of the same animal by Western blot analysis (Figs. 5A and B). Preferential expression of αPD-L1-γ₁ mRNA in TILs was confirmed by qRT-PCR (Fig. 5C). This expression pattern suggested that miR-423 regulation suppressed αPD-L1-γ₁ expression in HSPC progeny other than tumor-infiltrating myeloid and lymphoid cells. Serum αPD-L1-γ₁ became detectable after MMC cell injection and declined once tumors had disappeared, indicating a functional autoregulation of αPD-L1-γ₁ expression (Fig. 5D), that is the transgene expression started only once HSPGs differentiated into tumor-associated leukocytes. Starting from week 2 after MMC cell injection, we observed autoimmune reactions reflected by fur discoloration and inflammatory infiltrates in tissues (Supplementary Fig. S7). Importantly, in animals sacrificed 4 weeks after tumor disappearance, the histology of all organs returned to normal. This observation indicates that as long as αPD-L1-γ₁ is expressed and released into the blood stream, transient autoimmune reactions (most likely against neu-expressing tissues/cell types) can occur. Notably, a study with a HDAd-αPDL1γ₁-miR423 vector without miR-423-5p target sites had to be terminated because >20% weight loss occurring in treated animals 2 weeks after the last O⁶BG/BCNU treatment. This underscores the necessity for regulated αPD-L1-γ₁ expression. We are currently refining our miRNA regulation system by incorporating target sites for multiple miRNAs. In future studies, the observed autoimmune reactions could be minimized by physical tethering of αPD-L1-γ₁ to the tumor or by the use of intracellular immunomodulatory effectors (e.g., miRNAs that repolarize tumor-promoting leukocytes into tumor-killing cells). Furthermore, vectors could also contain a truncated EGFR receptor that allows for the destruction of all transduced cells by antibody (Erbitux)-dependent cytotoxicity (31).

The efficacy of the in vivo HSPC αPD-L1-γ₁ gene therapy approach is remarkable considering that in the neu-tg/MMC model, other immunotherapy approaches did not prevent tumor recurrence (25, 32). In this context, we found that four rounds of intraperitoneal injection of an anti-mouse PD-L1 mAb had no significant effect on tumor growth (Supplementary Fig. S8). Our data indicate that intratumoral expression of αPD-L1-γ₁ early during tumor development (as soon as HSPC progeny cells infiltrate the tumor) can tip the balance between suppressor and effector immune cells towards tumor elimination.

C57Bl/6-derived murine ovarian cancer ID8 cells do not contain typical cancer-associated germline mutations (brca1, brca2, p53, Nf1, Rb1, Pten, etc.) and poorly form tumors after intraperitoneal injection (20). Newer improved ID8-derived models, created by CRISPR/Cas9 knockout of tumor-suppressor genes, address these deficiencies (20, 33). Among these models are ID8-p53^−/−-brca2^−/− cells. Intraperitoneal in injection of 5 × 10⁶ ID8-p53^−/−-brca2^−/− cells into CD46-transgenic mice resulted in tumor growth and onset of ascites (or death) within 6–8 weeks (Fig. 6A). Intraperitoneal tumors were widespread along the mesenterium with invasion of other organs (spleen, liver, lymph nodes). Immunophenotyping of TILs in intraperitoneal ID8-p53^−/−-brca2^−/− tumors showed the pronounced presence of Tregs as well as immunosuppressive DCs/MDSCs as well as tumor-associated macrophages (TAM; Fig. 6B). We isolated TILs, macrophages (TAMs), and neutrophils (TAN) from peritoneal ID8 p53^−/− brca2^−/− tumors and analyzed miRNA-423-5p levels by Northern blot. As observed in the MMC and TC-1 models, miR-423-5p was expressed in bone marrow mononuclear cells but not detectable in tumor-infiltrating leukocytes (including TILs, TANs, and TAMs) indicating that all three cell types had been specifically reprogrammed by the tumor (Fig. 6C).

First, we used the ID8-Trp53^−/−-brca2^−/− model in a prophylactic setting (Fig. 6D). After HSPC in vivo transduction/selection with HDAd-αPDL1γ₁-miR423 + HDAd-SB or HAd-GFP-miR423 + HDAd-SB (control), ID8-p53^−/−brca2^−/− cells were injected intraperitoneally and serum αPDL1γ1 levels and onset of morbidity and ascites were monitored. While all control mice reached the endpoint by day 70 after in vivo transduction, 100% of HDAd-αPDL1γ₁-miR423 + HDAd-SB–treated animals were alive at the end of the monitoring period (11 weeks after tumor cell inoculation; Fig. 6E). Elevated serum αPD-L1-γ1 levels around week 6 (post cell injection) suggest that tumors had grown and activated serum αPDL1γ1 expression (Fig. 6F). By week 11, serum αPD-L1-γ1 returned to background levels, indicating that tumors had been cleared. In this study, we did not observe signs of autoimmune reactions (e.g. fur discoloration), most likely due to the absence of antigens shared between the tumor and normal tissues (e.g., Neu). In the context of assessing the safety of our approach, we also showed that in vivo HSPC transduction with HDAd-αPDL1γ₁-miR423 did not cause abnormalities in hematopoiesis (Supplementary Fig. S9).

While an immunoprophylaxis approach has the advantage of commencing automatically at a very early stage of tumor-development, its immediate application in healthy women carrying high-risk mutations will likely face regulatory hurdles in clinical translation. A more realistic goal, therefore, is to use our approach to prevent cancer recurrence after first-line therapy. In this case, in vivo HSPC selection can be directly embedded into the chemotherapy treatment of patients (Fig. 7A). This setting also has the advantage that tumor-specific neoantigens and the immuno-phenotype of the tumor will be known from the analysis of surgical biopsies, which would allow for selecting the adequate immunotherapy effector genes. On the other hand, preventing the recurrence of cancer with “fully fledged” cancer hallmarks (6) is more challenging than targeting a tumor at early stages of development.

To simulate such a “therapeutic” setting, we injected CD46-transgenic mice first with ID8-Trp53^−/−-brca2^−/− cells followed by in vivo HSPC transduction/selection two weeks later (Fig. 7B). While all mice in the control setting (HDAd-GFP-miR423 + HDAd-SB transduced HSPCs) reached the end point by week 12 after tumor cell injection, all mice treated with the αPD-L1-γ1–expressing vector were healthy at week 15 (Fig. 7C). As in the prophylaxis study, elevated serum αPD-L1-γ1 levels at week 11 suggest that tumors initially grew but disappeared once the self-regulated αPD-L1-γ1 mechanism was activated (Fig. 7D). These data indicate that our approach could potentially prevent cancer recurrence after surgery/first-line chemotherapy.

We have performed mRNA profiling/Northern blot analyses for TILs present in TC-1 (mouse lung-cancer) tumors (Supplementary Fig. S4 and S5), MMC (mouse breast cancer) tumors (Fig. 2; Supplementary Fig. S5), and ID8-p53^−/−/brca2^−/− (mouse ovarian cancer tumors; Fig. 6C) and found in all three tumor types that miR423-5p is undetectable, but present at high-levels in normal hematopoietic compartments. Together with our data from human ovarian cancer biopsies (Supplementary Fig. S6), this indicates that the miR423-5p-based system can be broadly used for different tumor types across species for the regulation of effector gene expression.

Considering the limited prophylactic options that are currently offered to women with germline mutations associated with high-risk of cancer onset, and the increasing numbers of these carriers due to population-wide screening, we believe that our in vivo HSPC gene therapy approach is a promising strategy that addresses a major medical problem.

No potential conflicts of interest were disclosed.

Conception and design: A. Lieber

Development of methodology: C. Li, A. Lieber

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Li, M.M. Course, I.A. McNeish, C.W. Drescher, P.N. Valdmanis, A. Lieber

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Li, M.M. Course, P.N. Valdmanis

Writing, review, and/or revision of the manuscript: C. Li, P.N. Valdmanis

We thank Pavel Sova for help with RNA analysis, Sucheol Gil for support in mouse experiments, and Meng Wang for editing the figures. The study was supported by the 2016 Lester and Bernice Smith Challenge Grant Award, a grant from the "Wings of Karen," and a NIH R21 grant (CA193077 to A. Lieber).

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.