Hematopoietic Stem Cell Gene Therapy for Brain Metastases Using Myeloid Cell–Specific Gene Promoters

Abstract Background Brain metastases (BrM) develop in 20–40% of cancer patients and represent an unmet clinical need. Limited access of drugs into the brain because of the blood-brain barrier is at least partially responsible for therapeutic failure, necessitating improved drug delivery systems. Methods Green fluorescent protein (GFP)-transduced murine and nontransduced human hematopoietic stem cells (HSCs) were administered into mice (n = 10 and 3). The HSC progeny in mouse BrM and in patient-derived BrM tissue (n = 6) was characterized by flow cytometry and immunofluorescence. Promoters driving gene expression, specifically within the BrM-infiltrating HSC progeny, were identified through differential gene-expression analysis and subsequent validation of a series of promoter-green fluorescent protein-reporter constructs in mice (n = 5). One of the promoters was used to deliver tumor necrosis factor–related apoptosis-inducing ligand (TRAIL) to BrM in mice (n = 17/21 for TRAIL vs control group). Results HSC progeny (consisting mostly of macrophages) efficiently homed to macrometastases (mean [SD] = 37.6% [7.2%] of all infiltrating cells for murine HSC progeny; 27.9% mean [SD] = 27.9% [4.9%] of infiltrating CD45+ hematopoietic cells for human HSC progeny) and micrometastases in mice (19.3–53.3% of all macrophages for murine HSCs). Macrophages were also abundant in patient-derived BrM tissue (mean [SD] = 8.8% [7.8%]). Collectively, this provided a rationale to optimize the delivery of gene therapy to BrM within myeloid cells. MMP14 promoter emerged as the strongest promoter construct capable of limiting gene expression to BrM-infiltrating myeloid cells in mice. TRAIL delivered under MMP14 promoter statistically significantly prolonged survival in mice (mean [SD] = 19.0 [3.4] vs mean [SD] = 15.0 [2.0] days for TRAIL vs control group; two-sided P = .006), demonstrating therapeutic and translational potential of our approach. Conclusions Our study establishes HSC gene therapy using a myeloid cell–specific promoter as a new strategy to target BrM. This approach, with strong translational value, has potential to overcome the blood-brain barrier, target micrometastases, and control multifocal lesions.

Despite partial disruption of the BBB in BrM, the vessel permeability in experimental BrM reaches only approximately 15% of that seen in other organs (4). Thus, novel approaches for the effective delivery of drugs to BrM are urgently required.
A handful of studies have explored neuronal and mesenchymal stem cells to deliver gene therapy to BrM in preclinical models (6)(7)(8)(9), whereas hematopoietic stem cells (HSCs) have not yet been investigated in this context. We previously observed a substantial homing of macrophages, which are derived from HSCs, to BrM (10). Advantages of HSCs in comparison to other stem cell therapies include the ability to isolate them in large quantities and well-established procedures for their therapeutic use. Recent clinical trials of HSC gene therapy for Wiskott-Aldrich syndrome (11), X-linked severe combined immunodeficiency (12), b-thalassaemia (13), and adrenoleukodystrophy [a severe demyelinating brain disease (14,15)] showed remarkable results. The disadvantage of HSCs in the context of therapies, however, is the wide distribution of their progeny in different tissues, leading to systemic rather than localized delivery of transferred genes and thus potential systemic toxicities. To address this challenge, our goal was to develop a strategy for lentiviral gene transfer into HSCs that would restrict the delivery of transgenes to BrM.

In Vivo BrM Models
Six-to eight-week-old female C57Bl/6J and NSG mice were purchased from Charles River Laboratories, UK. Humanized NSG mice engrafted with human CD34þ cells used in immune cell quantification studies were purchased from JAX. Using stereotaxic apparatus (intracranial implantation model), we injected 1Â10 5 cancer cells into the striatum (16) or into the left internal carotid artery (10,17). Metacam (15 mg) was administered subcutaneously, and an inhalable anaesthetic (Isoflurane) was used during surgery. Bioluminescence imaging was performed using IVIS Spectrum (Perkin Elmer). All surgery and animal care procedures followed recommendations by the University of Leeds Animal Welfare & Ethical Review Committee and were performed under the approved UK Home Office project license in line with the Animal (Scientific Procedures) Act 1986.

In Vivo Survival Study
HSCs were lentivirally transduced with MMP14: GFP or MMP14: tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) construct (MOI 20) and injected intravenously into lethally irradiated (8.45 Gy) C57Bl/6J mice (6-week-old females). Following bone marrow reconstitution (7 weeks later), 1Â10 5 PyMT cells were injected intracranially. Mice were examined daily for tumor growth-related symptoms, and symptoms were recorded according to the scoring sheet in our Home Office project license. At the onset of morbidity, the mice were culled, and brain tissue isolated for quantitative polymerase chain reaction (qPCR) analysis. Three mice were excluded from statistical analyses of the survival study: one mouse from each group because of non-tumor growth-related symptoms necessitating the animals be euthanized at an early time point and one mouse from the MMP14: GFP group because of failure to initiate tumor growth.

Microarray Gene-Expression Analysis
Biological replicates used for analysis were as follows: n ¼ 4 for PyMT BrM and the spleens of PyMT BrM-bearing mice; n ¼ 1 for the bone marrow of PyMT BrM-bearing mice; n ¼ 2 for EO771 BrM, naïve spleens, and naïve bone marrow. Toward the experimental endpoint, reduced size of spleens with black areas suggesting cell death was observed in mice with intracranial EO771 tumors. Consequently, we were unable to isolate viable spleen and bone marrow cells from this model despite repeating the experiment three times. Because of these difficulties, spleens and bone marrow isolated from naïve mice were used instead.

Human Tissue
Human breast cancer BrM tissue and matched blood were obtained from the Leeds General Infirmary, the Leeds Teaching Hospitals Trust. Written

Statistical Analyses
Data were analyzed using a two-tailed t test or one-way analysis of variance with multiple comparisons, as stated in the figure legends (statistical significance cutoff ¼ 0.05). Error bars represent SDs. Statistical significance in the survival study was determined via a two-sided log-rank test using GraphPad Prism 8. Further methods are provided in the Supplementary Methods (available online).

Interaction of Myeloid Cells with BrM
To allow for in vivo detection, PyMT (10,19) and EO771 (20) murine breast cancer cell lines were tagged with Discosoma sp. red fluorescent protein and Firefly luciferase (EO771-DF and PyMT-DF). Both models formed large single lesions following intracranial implantation in C57Bl/6J mice (16) ( Figure 1A Supplementary Figure 1A, available online). They also efficiently colonized the brain after administration into the internal carotid artery (10,17), resulting in multiple cancer lesions including micro-(few cells) and macrometastases ( Figure 1B; Supplementary Figure 1B, available online), which grew in a close association with vasculature (Supplementary Figure 1C, available online). Thus, our models home to and colonize the brain, mimic multifocal metastatic lesions, and allow for the study of micro-and macrometastases, thereby recapitulating key aspects of the clinical disease.
As previously reported (10), CD11bþ myeloid cells were abundantly infiltrating large intracranial tumors and micrometastases in EO771-DF and PyMT-DF models ( Figure 1C). Moreover, we detected substantial infiltration of F4/80þ microglia/macrophages within macrometastases in the brain in a spontaneous melanoma model (21) (Supplementary Figure 1F, available online). To determine the proportion of BrM-infiltrating cells originating from the HSCs as opposed to the yolk sackderived brain-resident microglia (22), we generated bone marrow chimeras through transplantation of GFPþ HSCs into irradiated mice, resulting in mean (SD) ¼ 73.4% (6.6%) GFPþ cells in the blood ( Figure 1D; Supplementary Figure 1D, available  Within the F4/80þCD11bþ population infiltrating normal brain, mean (SD) ¼ 66.6% (4.4%) of the cells were CD45 low , and only mean (SD) ¼ 8.9% (6.0%) within this population were HSCderived GFPþ cells (Supplementary Figure 2E, available online). Interestingly, CD45 low GFPþ cells retained low P2RY12 expression comparable to CD45 high GFPþ macrophages (Supplementary Figure 2F, available online), suggesting they are distinct from brain-resident microglia with high P2RY12 expression, which is in line with other studies (28).
Notably, a proportion of myeloid cells in tumors could originate directly from a subpopulation of transplanted HSCs without hematopoiesis as shown previously for normal brain (28,29). Regardless, based on their capability to deliver transgenes to BrM, and based on their high proportion in intracranial tumors, we reasoned that HSC-derived myeloid cells could be used as cellular vehicles for the delivery of therapeutic genes to BrM.

Homing of Genetically Modified HSC Progeny to BrM
To validate HSCs and their progeny as cellular vehicles, we used GFP as a model gene to be delivered to BrM. Murine HSCs were transduced with a lentiviral vector pFUGW, expressing GFP under the Ubiquitin C (UBC) promoter (18,30) and transplanted into lethally irradiated C57Bl/ 6J mice. BrM were generated by intracranial EO771-DF implantation ( Figure 2A). The progeny of GFP-transduced HSCs efficiently homed to BrM ( Figure 2B). As expected, GFPþ HSC progeny were also present in other organs to variable extent ( Figure 2C).
In line with experiments using HSCs from GFP: UBC transgenic mice, the F4/80þ population contained almost exclusively macrophages (Supplementary Figure 1E The progeny of GFP-transduced HSCs also efficiently tracked down micrometastases and closely associated with small EO771 lesions (Figure 2, F and G). The majority of micrometastasesassociated cells were CD11bþ and F4/80þ ( Figure 2H), with 19.3-53.3% expressing GFP ( Figure 2I). The ratio of cancer cells to CD11bþ cells (1:1 to 1:2) was similar to the intracranial model ( Figure 2H). All micrometastases were associated with CD11bþ cells, and more than 90% contained CD11bþGFPþ cells (Supplementary Figure 3B, available online). Thus, we demonstrated that the genetically engineered HSC progeny, mostly consisting of macrophages, efficiently homed to large BrM and to micrometastases and can deliver genetically expressed molecules to the close proximity of cancer cells.

Validation of Human HSCs and Their Progeny as Cellular Vehicles
With clinical translation in mind, we next validated human HSCs (hHSCs). Engraftment of CD34þ hHSCs in sublethally irradiated NOD/SCID/IL2rcKO (NSG) mice, previously shown to result in mature and functional human hematopoietic cells (31), was followed by intracranial implantation of the human brainhoming breast cancer cell line MDA-MB-231/brain (32) ( Figure 3A Figure 3I). In summary, these data demonstrated strong parallels between human tissue and our preclinical models.

Identification of Gene Promoters for BrM-Specific Delivery of Therapeutic Molecules within Myeloid Cells
To enable predominant delivery of gene therapy to BrM, we sought to identify gene promoters that are upregulated in BrM-infiltrating myeloid cells. GFPþCD11bþ cells were isolated from brain tumors, bone marrow, and the spleen of mice with chimeric GFPþ bone marrow ( Figure 4A), bearing EO771 and PyMT tumors, respectively, or from naïve mice and subjected to a genome-wide gene-expression analysis. Data from both cancer models were initially combined, and BrM were compared with the pooled spleen/bone marrow group. Differential gene-expression analysis identified 5972 statistically significantly differentially expressed probes (False Discovery Rate [FDR] < 1%; Figure 4B), including chemokines, matrix metalloproteinases, and genes associated with macrophage activation/polarization (10,34,35). Expression of macrophage polarization markers (36)(37)(38)(39) differed strongly between the intratumoral and the spleen/bone marrow-derived myeloid cells (Figure 4, C and D). Notably, 119 probes showed greater than 10-fold and 9 probes greater than 100-fold upregulation in BrM. However,

Validation of BrM-Specific Myeloid Promoter Constructs in Preclinical Models
Respective fragments of MMP14, SPP1 (two lengths), and DAB2 promoters were cloned into lentiviral vectors upstream of GFP ( Figure 6A). HSCs transduced with these vectors were transplanted into lethally irradiated mice, followed by generation of intracranial EO771 tumors ( Figure 6B). As expected, UBC promoter-driven GFP expression could be detected in CD45þ cells in all tissues without a tissue-specific pattern. By contrast, MMP14 and SPP1 promoters displayed statistically significantly higher GFP mean fluorescence intensity (MFI; a measure of promoter strength; see Supplementary Table 2, available online for P values) in CD45þ cells infiltrating BrM as compared with other tissues ( Figure 6, C and D). Promoter strength (GFP MFI) was higher for the MMP14 promoter construct. To account for differences in the viral copy number (VCN) between different promoter constructs and tissues, we normalized MFI to VCN (MFI/VCN). As expected, VCN was higher in the bone marrow and spleen as compared with the brain tumors, whereas MFI/VCN was highest in the brain tumors. This confirmed an increased activity of MMP14 and SPP1 promoters in BrM-infiltrating vs spleen/bone marrow-infiltrating HSC progeny. In line with the MFI analysis, the MFI:VCN ratio was also the highest for the MMP14 promoter . Identification of genes specific for brain metastases-infiltrating myeloid cells. A) Tumors in the brain were generated by intracranial implantation of cancer cells (untagged) into chimeric mice with green fluorescent protein (GFP)-tagged bone marrow (derived from Ubiquitin C [UBC]: GFP mice). CD11bþGFPþ cells were isolated from the brain tumors, spleens, and the bone marrow and subjected to RNA microarray analysis. Spleens and bone marrow isolated from naïve mice were used instead of those isolated from EO771 brain metastases (BrM)-bearing mice. B) A heat map of the top 150 probes differentially expressed between bone marrow-derived myeloid cells (CD11bþGFPþ) isolated from BrM (Br) and myeloid cells isolated from the spleen (S) bone marrow (BM); n ¼ 4 for PyMT BrM and the spleens; n ¼ 1 for the PyMT bone marrow; n ¼ 2 for EO771 BrM, naïve spleens, and naïve bone marrow. C) Differential expression of macrophage polarization-associated genes between BrM-and the spleen/bone marrow-derived myeloid cells. D) Functional protein interaction network of 33 murine macrophage polarization-associated genes, as identified using Search Tool for the Retrieval of Interacting Genes (STRING) (33). Individual nodes represent genes connected by color-coded lines of interaction according to software predictions (confidence score set to 0.4). E) Expression of top 10 genes upregulated in myeloid cells within brain metastases was validated by semiquantitative polymerase chain reaction. GADPH was used as a control. Representative DNA gels are shown. The quantification of the data is shown in Supplementary Figure S4B (available online). HSC ¼ hematopoietic stem cell. Figure 8A, available online). With the exception of the lungs, the percentage of GFPþ cells was higher in BrM than in other tissues for MMP14 and SPP1 promoter-reporter constructs ( Figure 6D; Supplementary Table 1, available online). Based on these data, we focused on the MMP14 promoter.

ARTICLE (Supplementary
Using mice that have received GFPþ HSCs from UBC: GFP transgenic mice, the expression of MMP14 in GFPþ cells in the brain was further analyzed by using immunofluorescence. GFPþ cells within normal brain were mainly located at the choroid plexus and ventricle walls, with lower numbers found within the brain parenchyma, cortex, cerebellum, and at the brain surface (Supplementary Figure 9, available online). In contrast to strong MMP14 expression in tumor-infiltrating GFPþ cells, GFPþ cells within normal brain were MMP14-negative (Supplementary Figure 9, available online). In line with this, analysis of previously published gene-expression data (25,27,28) revealed a statistically significantly higher expression of Mmp14 in HSC-derived tumor-infiltrating macrophages vs HSC-derived normal brain-engrafting microglia-like cells (Supplementary Figure 8B, available online). In summary, this demonstrated an increased specificity of MMP14 expression in tumor-infiltrating hematopoietic cells.
Notably, CD45þ cells infiltrating BrM in a spontaneous melanoma model (21) also expressed MMP14 (Supplementary Figure 6B, available online), confirming that the MMP14 promoter is also active in BrM-infiltrating hematopoietic cells in a more physiological model and in the absence of irradiation.

TRAIL Gene Therapy Delivered under the MMP14 Promoter
Delivery of proapoptotic molecule TRAIL (9) was chosen as a proof of principle of the therapeutic applicability of our approach. After confirming the expression of the TRAIL construct under the UBC promoter ( Figure 7A) and the sensitivity of PyMT cancer cells to the TRAIL in vitro (Figure 7, B and C), the TRAIL was placed under the MMP14 promoter fragment in a lentiviral vector and was used for in vivo efficacy study in the intracranial PyMT model ( Figure 7D). Survival analysis revealed a statistically significantly prolonged survival of mice following TRAIL delivery in HSCs (mean [SD] ¼ 19 [3.4] vs 15 [2.0] days for MMP14: TRAIL vs MMP14: GFP; two-sided P ¼ .006) ( Figure 7E), together with increased Tnsf10 (TRAIL gene) expression in brain tumors as determined by qPCR ( Figure 7F). In summary, our data establish a new platform for the delivery of therapeutic genes to BrM.

Discussion
In this study, we developed a strategy for the delivery of therapeutic genes to BrM within the HSC progeny, using gene promoters specific for tumor-infiltrating myeloid cells. The latter were identified as the optimal cellular vehicles because of their high presence in BrM, their uniform distribution within tumors, and predominant homing to the tumor rather than healthy brain. Moreover, this strategy is expected to circumvent the BBB, control micrometastases, and enable a simultaneous targeting of multifocal BrM that pose a challenge for surgical removal. Use of myeloid cell-specific promoters is expected to limit the delivery of gene therapy mainly to tumors and thereby minimize systemic side effects. Whereas a similar approach using the Tie2-promoter has been previously used to deliver therapy to glioma in Tie2-expressing monocytes (40), here, we chose to focus on the most abundant hematopoietic cell population in BrM. In summary, our study establishes a novel approach for targeting of BrM with HSC gene therapy.
Lentiviral gene transfer demonstrated an excellent safety record (41,42), with promising results in patients (11)(12)(13)(14)(15). We demonstrated a strong potential for the clinical translation of HSC gene therapy for BrM by demonstrating the BrM-homing capacity of hHSCs, the abundant infiltration of myeloid cells, and the activity of identified myeloid promoters in human BrM. Because MMP14 is upregulated in brain-infiltrating macrophages in Alzheimer disease, multiple sclerosis, and stroke (43), this promoter could be also used for gene therapy in noncancerous brain disorders accompanied by strong myeloid cell infiltration (14,(44)(45)(46)(47)(48)(49). Moreover, our approach has the potential for simultaneous targeting of multiorgan metastases.
Several preclinical models were used in this study to address the key clinically relevant features of BrM, including the context of an intact immune system (syngeneic models), validation of human HSCs (humanized NSG mice), and analysis of macrometastases (intracranial model) and micrometastases (carotid artery model). Because approximately 15-60% of BrM remain undiagnosed (50,51), micrometastases in the brain may be quite common. Established BrM have also been shown to invade surrounding tissue (52,53). Thus, targeting of dormant micrometastases and those remaining after surgical removal (54,55) represents an unmet clinical need that could be addressed with our strategy.
In the present proof-of-principle study we used the MMP14 promoter. However, our study has several limitations. First, the initial differential gene-expression analysis was limited to the comparison of myeloid cells between brain tumors, bone marrow, and the spleen. Although we subsequently analyzed additional tissues, several tissues containing mature macrophages (ie, gut, liver, and fat) were not investigated because of technical limitations, and thus a possibility remains that our promoters are active in these tissues. Second, we reasoned that assessing the promoter strength at the protein level (flow cytometry) is superior, as proteins/peptides rather than mRNA represent effector molecules in therapeutic applications. However, use of an additional method (ie, qPCR quantification of transgene expression) would strengthen our findings. Third, we used whole-body irradiation before HSC transplantation. With clinical translation in mind, it will be important to determine in the future which HSC transplantation approach (ie, busulfan or treosulfan [29]) results in the highest engraftment of HSC progeny within tumors while minimizing engraftment within the healthy brain and other organs.