Abstract

   Carcinomas are associated with metastasis to specific organs while
   sparing others. Breast cancer presents with lung metastasis but rarely
   kidney metastasis. Using this difference as an example, we queried the
   mechanism(s) behind the proclivity for organ specific metastasis. We
   employed spontaneous and implant models of metastatic mammary carcinoma
   coupled with inflammatory tissue fibrosis, single cell sequencing
   analyses, and functional studies to unravel the causal determinants of
   organ-specific metastasis. Here we show that lung metastasis is driven
   by Angiopoietin-2 (Ang2) mediated suppression of lung specific
   endothelial tight junction protein Claudin 5, which is augmented by the
   inflammatory fibrotic microenvironment and prevented by anti-Ang2
   blocking antibodies; while kidney metastasis is prevented by non-Ang2
   responsive Claudin 2 and Claudin 10. Suppression of Claudin 2/10 was
   sufficient to induce emergence of kidney metastasis. This study
   illustrates the influence of organ specific vascular heterogeneity in
   determining organotropic metastasis, independent of cancer cell
   intrinsic mechanisms.

Introduction

   Metastasis is the primary cause of morbidity and mortality in breast
   cancer, and accounts for about 90% of breast cancer-related deaths
   ^[62]1. Clinically, breast cancer metastasizes to different secondary
   sites, including bone, lung, liver, and brain, but very rarely to
   organs such as the kidney ^[63]2, despite the fact that 20% of total
   blood volume is associated with the renal parenchyma at any given time.
   The preferential spread of metastases to specific organs (organotropic
   metastasis) has been hypothesized to depend on interactions between
   ‘seed’ (cancer cells and their intrinsic properties) and ‘soil’ (or the
   host microenvironment) ^[64]3. Cancer cells may acquire novel genetic
   and/or epigenetic features to overcome microenvironmental pressures,
   with selection of clones with distinct fitness to different
   microenvironments at distant metastatic site(s). The parenchymal
   turnover rate, immune surveillance, metabolic fitness, angiogenic
   potential, biomechanical properties, and other features of the
   metastatic sites may contribute to the survival and adaptation of
   cancer cells, but may also present with restrictive inherent features
   that limit successful formation of metastatic disease. Organotropic
   metastasis is likely regulated by multiple factors, including
   organ-specific metastatic niche formation involving tumor cell-secreted
   factors and exosomes ^[65]4–[66]8. The vascular and perivascular niches
   may also play a role establishing metastasis and needs mechanistic
   elucidation ^[67]9–[68]13. The rate limiting function of specific
   parenchymal remodeling events in organotropic metastasis remains
   unknown. Given vascular remodeling is critical for hematological
   dissemination of cancer cells for metastasis to occur, we aimed to
   ascertain the mechanistic underpinning of organotropic metastasis, with
   a specific focus on the permissive and restrictive properties to blood
   vessels in this process. Studies suggest that tissue alterations
   associated with organ injury and fibrosis may promote metastatic
   outgrowth^[69]14–[70]17. While vascular remodeling is a hallmark of
   tissue repair in cancer and non-cancer pathologies, the functional and
   mechanistic significance of organ specific vascular heterogeneity in
   metastasis is yet to be defined.

Results

Lung fibrosis forms a pro-metastatic niche for lung metastasis

   Fibrosis is considered a precursor for future emergence of cancer and
   potentially serves as a metastatic niche for secondary spread of cancer
   ^[71]14,[72]18,[73]19. To evaluate the impact of the pro-metastatic
   milieu in the recruitment of lung metastasis, we induced lung fibrosis
   in 4T1 mammary tumor bearing mice to generate an inflammatory tissue
   microenvironment. Emergence of bleomycin-induced lung fibrosis was
   associated with prototypical parenchymal damage and remodeling that
   includes immune infiltration, deposition of extracellular matrix
   proteins such as collagens, accumulation of smooth muscle αactin
   expressing (αSMA^+) myofibroblasts, and transcriptional upregulation of
   fibronectin (Fn1) and type I collagen (Col1a1) ([74]Figure 1A-[75]B,
   [76]Extended Data Figure 1A-[77]D). Lung fibrosis was induced in 4T1
   orthotopic tumor bearing mice when total tumor volume reached
   approximately 500 mm^3 ([78]Figure 1C). Induction of lung fibrosis had
   no significant impact on the growth kinetics of the primary mammary
   tumors ([79]Figure 1C), but significantly enhanced lung metastasis and
   surface lung nodules when compared to non-fibrotic lungs of tumor
   bearing mice ([80]Figure 1D-[81]E). An increase in GFP^+ cancer cells
   was noted in the fibrotic lungs compared to the non-fibrotic control
   lungs ([82]Extended Data Figure 1E), in agreement with previous studies
   ^[83]20. Similar results were observed in the spontaneous murine
   MMTV-PyMT metastatic mammary carcinoma model ([84]Figure 1F-[85]H).
   Lung fibrosis in this model, initiated with intratracheal bleomycin
   administration when the tumor burden reached approximately 500 mm^3,
   also failed to impact the growth kinetics of primary mammary tumors but
   significantly enhanced lung metastatic disease ([86]Figure 1F-[87]H).
   Despite enhanced lung metastasis observed in both models of mammary
   carcinoma with lung fibrosis, the metastasis to the liver, evaluated by
   histology and immunolabeling for GFP^+ cancer cells, was unchanged
   ([88]Extended Data Figure 1F-[89]H). Cancer cells were not detected in
   the heart and bowel ([90]Extended Data Figure 1I-[91]J). Collectively,
   these results provide evidence for the specific pro-metastatic
   potential of an inflammatory lung fibrotic microenvironment.

Figure 1. Fibrosis enhances tropic metastasis with no impact on non-tropic
sites.

   Figure 1.
   [92]Open in a new tab

   (A-B) Representative MTS of lung (A) and kidney (B) fibrosis. Scale
   bar, 100 μm. (C) Orthotopic 4T1 tumor volume over time +/− lung
   fibrosis, n=6 mice per group. (D) Representative lung H&E of 4T1
   orthotopic tumor bearing mice +/− lung fibrosis. Scale bar, 25 μm. (E)
   Quantification of metastatic area and surface lung nodules in 4T1
   orthotopic tumor bearing mice +/− lung fibrosis, n=6 mice per group.
   (F) MMTV-PyMT tumor volume over time +/− lung fibrosis, n=5 to 7 mice
   per group. (G) Representative lung H&E of MMTV-PyMT tumor bearing mice
   +/− lung fibrosis. Scale bar, 25 μm. (H) Quantification of metastatic
   area and surface lung nodules in MMTV-PyMT tumor bearing mice +/− lung
   fibrosis, n=5 to 7 mice per group. (I) Orthotopic 4T1 tumor volume over
   time +/− kidney fibrosis, n=6 mice per group. (J) Representative lung
   H&E of 4T1 orthotopic tumor bearing mice +/− kidney fibrosis. Scale
   bar, 25 μm. (K) Quantification of metastatic area and surface lung
   nodules in 4T1 orthotopic tumor bearing mice +/− kidney fibrosis, n=6
   mice per group. (L) MMTV-PyMT tumor volume over time +/− kidney
   fibrosis, n=6 mice per group. (M) Representative lung H&E of MMTV-PyMT
   tumor bearing mice +/− kidney fibrosis. Scale bar, 25 μm. (N)
   Quantification of metastatic area and surface lung nodules in MMTV-PyMT
   tumor bearing mice +/− kidney fibrosis, n=5 to 6 mice per group. (O)
   Representative lung H&E of the lung of mice following 4T1 i.v.
   injection +/− lung fibrosis and +/− kidney fibrosis. Scale bar, 25 μm.
   (P) Quantification of lung colonization following 4T1 i.v. injection
   +/− lung fibrosis or kidney fibrosis, n=6 mice per group. Data are
   presented as mean values +/− SEM. ctrl: control, LF: lung fibrosis, KF:
   kidney fibrosis. For B, E, H, K, N, P: unpaired two-tailed t test with
   Welch’s correction applied for unequal variances (determined by
   F-test); for C, I: two-way ANOVA with Sidak’s multiple comparisons
   test; for F, L: multiple unpaired t tests of tumor measurements. P
   values are listed, ns: not significant.

Kidney fibrosis does not promote kidney metastasis but enhances lung
metastasis

   Breast cancer is rarely associated with metastasis to the kidney in
   humans or in mouse models of disease ^[93]21,[94]22. Based on the
   observation of increased lung metastasis induced by inflammatory lung
   fibrotic environment (vide supra), we decided to test the contribution
   of an inflammatory and pro-tumorigenic kidney fibrotic microenvironment
   on potential emergence of metastasis to the kidney. We induced fibrosis
   in the kidneys of 4T1 tumor bearing mice. Unilateral ureter obstruction
   (UUO) induced inflammatory kidney fibrosis with a prototypical
   parenchymal remodeling of fibrosis, similar to the changes observed in
   lung fibrosis ([95]Figure 1A-[96]B, [97]Extended Data Figure 1A-[98]D).
   Kidney fibrosis was induced in 4T1 orthotopic tumor bearing mice when
   the total tumor volume reached approximately 500 mm^3 ([99]Figure 1I).
   Kidney fibrosis had no significant impact on the growth kinetics of the
   primary tumor but significantly increased lung metastasis and surface
   lung metastatic nodules in two independently performed studies
   ([100]Figure 1J-[101]K, [102]Extended Data Figure 1K-[103]M). These
   histological findings were corroborated by immunolabeling for GFP^+
   cancer cells, which were increased in frequencies in the lung of mice
   with kidney fibrosis compared to sham controls ([104]Extended Data
   Figure 1N-[105]O). Similar results were observed in the MMTV-PyMT
   transgenic mouse model with inflammatory kidney fibrosis ([106]Figure
   1L-[107]N). Liver metastasis was unchanged with inflammatory kidney
   fibrosis ([108]Extended Data Figure 1P), and cancer cells were not
   detected in any other organs surveyed ([109]Extended Data Figure
   1Q-[110]R). Despite inflammatory tissue remodeling (lung and kidney
   fibrosis) increasing metastasis to the lungs, neither 4T1 orthotopic or
   MMTV-PyMT models revealed metastasis to the parenchyma of kidneys
   ([111]Extended Data Figure 2A-[112]G). In MMTV-PyMT mice, PyMT^+ cancer
   cells were not detected in the parenchyma of kidneys ([113]Extended
   Data Figure 2G). The inflammatory fibrosis status in all experiments
   (4T1 orthotopic, 4T1 i.v., and or MMTV-PyMT models) were confirmed
   histologically ([114]Extended Data Figure 2H-[115]J). Moreover, in both
   the 4T1 orthotopic and 4T1 i.v. models, circulating cancer cells can be
   identified within the blood vessels of the kidney ([116]Extended Data
   Figure 3A). However, cancer cells were not detected in the parenchyma
   of perfused kidneys; but they were detected in kidneys without
   perfusion, confined within the vasculature, unable to exit the blood
   vessels. This was ascertained by flow analysis ([117]Extended Data
   Figure 3B-[118]C, [119]Supplementary Figure 1A-[120]B) and QPCR
   analysis ([121]Extended Data Figure 3D-[122]E). Taken together, these
   results establish that unlike inflammatory lung fibrosis, inflammatory
   kidney fibrosis does not induce or facilitate metastases to the kidney.
   Instead, we observe that inflammatory kidney fibrosis enhances
   metastasis to the lung. The impact of organ fibrosis on colonization of
   4T1 cells administered intravenously via the orbital plexus was further
   validated using intracardiac injections, and in both cases metastatic
   colonization in the kidney was not observed ([123]Extended Data Figure
   3F-[124]I). The use of intracardiac rouse of administration of 4T1
   cancer cells eliminated the potential for first pass interaction of
   cancer cells with the lung capillary bed as a rate limiting factor for
   the lack of colonization in the renal parenchyma.

Fibrosis from cutaneous wound injury does not impact metastasis

   We next examined whether inflammatory cutaneous dorsal wound injury and
   healing impacts metastasis. 4T1 metastatic mammary tumor bearing mice
   with cutaneous wounds did not display measurable differences in primary
   tumor growth kinetics or incidence of lung metastasis ([125]Extended
   Data Figure 3J-[126]L). Similarly, MMTV-PyMT mice with cutaneous wounds
   also did not reveal differences in primary tumor growth kinetics or
   incidence of lung metastasis ([127]Extended Data Figure 3M-[128]O).
   Wound closure rate was unchanged between tumor-bearing and
   non-tumor-bearing mice ([129]Extended Data Figure 3P). Taken together,
   these results indicate that the fibrotic remodeling associated with the
   cutaneous wound healing response has no impact on lung metastasis in
   this setting and does not induce metastasis to other organs.

Cancer cells form tumor in the kidney parenchyma with lung metastasis

   Mice with either inflammatory lung fibrosis, kidney fibrosis, or
   cutaneous wounds were injected with 4T1 cancer cells intravenously to
   bypass the influence of primary tumor in metastatic colonization. In
   this setting, we observed increased colonization of cancer cells in
   fibrotic lungs when compared to control lungs ([130]Figure 1O-[131]P).
   Increased lung colonization was also observed in mice with kidney
   fibrosis compared to controls ([132]Figure 1O-[133]P). Lung
   colonization was not impacted by cutaneous wounds ([134]Extended Data
   Figure 3Q-[135]R). To define whether lack of metastasis or colonization
   of cancer cells in the kidneys was reflective of intrinsic,
   cell-autonomous properties of cancer cells, we tested their ability to
   establish tumors when implanted directly into the kidney parenchyma.
   The direct injection of 4T1 cancer cells into the kidney parenchyma
   resulted in tumor formation in the kidney, and inflammatory fibrosis
   further enhanced the ability to form mammary carcinoma tumors in the
   kidney ([136]Figure 2A-[137]F).

Figure 2. 4T1 cells form tumors when injected under the renal capsule.

   Figure 2.
   [138]Open in a new tab

   (A) Experimental design for the evaluation of tumor formation in mice
   injected with 4T1 cells under the renal capsule +/− kidney fibrosis
   induced immediately post injection. (B) Representative kidney H&E of
   mice with 4T1 cells injected under the renal capsule and +/− kidney
   fibrosis. Scale bar, 100 μm. (C-D) Kidney weight (C) and quantification
   of primary tumor area (D) in kidneys of the indicated groups, n=7 mice
   per group. (E-F) Representative kidney H&E of mice with 4T1 cells
   injected under the renal capsule together without (E) or with kidney
   fibrosis (F). Scale bar, 20 μm. (G-H) Representative images for GFP
   immunolabeling (G) and quantification of percent of the GFP positive
   area (H) in the indicated groups, n=4 mice per group. Scale bar, 50 μm.
   (I) Representative lung H&E images of mice with 4T1 cells injected
   under the renal capsule +/− kidney fibrosis. Scale bar, 25 μm. (J-K)
   Quantification of metastatic area/whole lung (J) and surface lung
   nodules (K) in the indicated groups, n=7 mice per group. (L-M)
   Representative H&E images of kidney in mice with 4T1 i.v.
   administration and with lung (L) or kidney fibrosis (M). Scale bar, 25
   μm. (N) Number of mice with microscopic kidney tumors in the indicated
   groups, n=6 mice per group. Data are presented as mean values +/− SEM.
   For D, H, J, K: unpaired two-tailed t test, with Welch’s correction
   applied for unequal variances (determined by F-test). For C: one-way
   ANOVA with Tukey multiple comparisons test. P values are listed, ns:
   not significant.

   Evaluation of kidneys for GFP^+ cancer cells further confirmed the
   increased tumor burden in fibrotic kidneys compared to non-fibrotic
   control kidneys ([139]Figure 2G-[140]H). These experiments demonstrate
   that inflammatory kidney fibrosis has the capacity to enhance local
   tumor growth, similar to inflammatory lung fibrosis. Mammary tumors
   induced in the kidney resulted in lung metastasis, despite not
   originating from the mammary fat pads ([141]Figure 2I-[142]K). It also
   suggests that the permeability of blood vessels associated with the
   primary tumor in the kidney may be different when compared to the
   endogenous normal kidney vasculature. Colonization studies with
   intravenous administration of cancer cells revealed no metastasis to
   the kidney in mice with either inflammatory lung fibrosis or
   inflammatory kidney fibrosis ([143]Figure 2L-[144]N), unlike the
   increase in lung colonization associated with inflammatory lung
   fibrosis ([145]Figure 1O-[146]P).

   Taken together, these results suggest that inflammatory lung fibrosis
   and kidney fibrosis exhibit equal potential to aid local tumor growth
   mediated by pro-tumor growth mediators and that circulating factors
   from the inflammatory fibrotic kidney, but not healing cutaneous
   wounds, might play a role in the lung tissue to increase metastasis.

Lung or kidney fibrosis induces lung specific vascular remodeling

   To explore the underlying mechanism(s) associated with the enhanced
   lung metastasis induced by lung and kidney fibrosis, we evaluated
   global transcriptomic changes in healthy lungs (HL), healthy kidneys
   (HK), fibrotic lungs (LF), fibrotic kidney (KF) and lungs from mice
   with kidney fibrosis (LoKF). Fibrotic remodeling in both lung and
   kidneys showed similar patterns of transcriptional changes associated
   with extracellular matrix remodeling, fibroblast proliferation,
   fibronectin binding, collagen signaling, and metabolism ([147]Extended
   Data Figure 4A-[148]B, [149]Supplementary Table 1). Analyses of
   differentially expressed genes (DEGs) revealed an increased expression
   of 2830 genes and decreased expression of 3235 genes in fibrotic lung
   when compared to the healthy lung ([150]Figure 3A-[151]B). Increased
   expression of 1173 genes and decreased expression of 999 genes in lungs
   from mice with kidney fibrosis was observed when compared to healthy
   lung ([152]Figure 3C-[153]D). The gene ontology enrichment of the
   comparative analyses revealed vascular ([154]Figure 3B, [155]D;
   [156]Supplementary Tables 2-[157]3) and immune ([158]Extended Data
   Figure 4C-[159]D; [160]Supplementary Tables 2-[161]3) remodeling in
   lungs with fibrosis and lungs from mice with kidney fibrosis.
   Overlapping upregulated or downregulated genes between LF vs HL and
   lungs from mice with kidney fibrosis vs HL confirmed significant
   vascular remodeling in both LF and LoKF compared to the HL
   ([162]Extended Data Figure 4E-[163]F, [164]Supplementary Table 4).
   Inflammatory and vascular remodeling cytokines ar also upregulated in
   both LF and KF ([165]Extended data Figure 4G), and collectively, these
   data suggest a role for a potential regulator of lung vascular
   remodeling that is induced by kidney fibrosis.

Figure 3. Ang2 mediates fibrosis-induced lung metastasis.

   [166]Figure 3.
   [167]Open in a new tab

   (A, C) Volcano plots depicting the number of differentially regulated
   genes in healthy lung vs. lung fibrosis (A) and healthy lung vs. lung
   from mice with kidney fibrosis (C). (B, D) Gene ontology (GO)
   enrichment ratio and false discovery rate (FDR) values for deregulated
   genes implicated in vascular remodeling in lung with fibrosis (B) or
   lung of mice with kidney fibrosis (D). The numbers of deregulated genes
   out of all the genes in the pathways are indicated in parenthesis with
   a star denoting common pathways in B and D. (E) Heatmap of relative
   mean levels of the listed cytokines in the serum of mice in the
   indicated group. Healthy, n=2; wound, n=2; lung fibrosis, n=3; kidney
   fibrosis, n=3 mice. (F) ELISA quantification of circulating Ang2 levels
   in the serum of mice in the indicated groups. Healthy, n=5; wound, n=3;
   4T1 orthotopic tumor bearing mice (T), n=4; control for lung fibrosis
   (ctrl LF), n=6; lung fibrosis, n=6; control for kidney fibrosis (ctrl
   KF), n=6; kidney fibrosis (KF), n=5 mice. (G-H) Representative images
   and respective quantification of dextran extravasation (green) and CD31
   immunolabeling (red) in the lung (G) and kidney (H) after rAng2 or PBS
   control treatment (individual images in [168]Supplementary Figure 2).
   Scale bar, 20 μm. Quantification of FITC-Dextran area per CD31 positive
   area, n=4 to 5 mice per group. (I-J) Representative lung H&E of mice
   with PBS or rAng2 treatment and with (J) and without (I) i.v. injection
   of 4T1 tumor cells. Scale bar, 200 μm. Quantification of pleural edema
   and thickness in the lungs, n=5 to 6 mice per group (I), and
   quantification of lung colonization and surface lung nodules (J) in the
   indicated groups, n=5 to 6 mice per group. Data are presented as mean
   values +/− SEM. Unpaired two-tailed t test, with Welch’s correction
   applied for unequal variances (determined by F-test). P values are
   listed, ns: not significant.

   Flow cytometry was conducted to further evaluate the immune
   microenvironment of the lung in response to lung fibrosis (LF) and
   kidney fibrosis (KF). The overall number of CD45^+ leukocytes was not
   changed upon LF and KF, but the percent of CD3^+ T cells were
   significant reduced. A significant increase in CD11b^+ myeloid cells
   and CD11b^+Ly6G^+Ly6C^- neutrophils were observed and similar immune
   cell alterations were also observed in the context of renal fibrosis,
   suggesting that neutrophils may be involved in the increased lung
   metastasis. However, effector T cells (CD4^+FoxP3^-), cytotoxic T cells
   (CD8^+), gamma delta (γδ^+) T cells, regulatory T cells (CD4^+FoxP3+),
   dendritic cells (CD11c^+) and B cells (CD19^+) were specifically
   altered in the lung of mice with LF but not KF. No change on CD11b^+
   Ly6C^+ Ly6G^- and natural killer (NK1.1^+) cells ([169]Extended Data
   Figure 5A, [170]Supplementary Figure 1C-[171]D). We similarly evaluated
   the immune infiltrate of the kidney in response to LF and KF and found
   that a significant upregulation of CD45^+ cells, regulatory T cells
   (CD4^+FoxP3+), natural killer (NK1.1^+) cells, and B cells (CD19^+),
   CD11b^+ myeloid cells and CD11b^+ Ly6C^+ Ly6G^- cells was observed in
   the kidneys of mice with KF but not LF. A significant down-regulation
   of CD3^+ T cells and effector T cells (CD4^+FoxP3^-) were found in the
   kidneys of mice with KF but not LF. No change could be found on
   cytotoxic T cells (CD8^+), gamma delta (γδ^+) T cells, dendritic cells
   (CD11c^+) and CD11b^+Ly6G^+Ly6C^- neutrophils in the kidney of mice
   with KF ([172]Extended Data Figure 5B, [173]Supplementary Figure
   1C-[174]D). The analyses further support that neutrophils may play
   collaborative role in aiding lung metastasis but by themselves they do
   not determine kidney metastasis. Taken together, these findings suggest
   that the immune microenvironment in fibrotic lungs could promote the
   proliferation of mammary carcinoma cells in lung, but it fails to
   facilitate the growth of cancer cells in the kidney, despite similar
   changes in immune cell composition.

   To evaluate the cytokines/chemokines secreted from the fibrotic kidney
   that may enhance vascular remodeling and favor metastasis, we performed
   targeted cytokine array using the serum from healthy mice, mice with
   cutaneous wounds (wound), mice with lung fibrosis (LF), and mice with
   kidney fibrosis (KF). Several cytokines were uniquely upregulated in KF
   and LF compared to healthy mice and mice with cutaneous wounds,
   suggesting their involvement in the vascular remodeling of the lungs.
   Induced inflammatory cytokines LIX/Cxcl5 and Reg3G, neutrophil products
   lipocalin-2/NGAL, myeloperoxidase, endothelial cell regulator
   angiopoietin-2 (Ang2), resistin, and osteopontin were identified as
   candidate effectors ([175]Figure 3E). To ascertain the cytokines
   specifically upregulated in LF and KF compared to healthy and cutaneous
   wound carrying mice, the data was normalized and ranked. Amongst the
   cytokines implicated in vascular remodeling, Ang2 was the most
   differentially upregulated in kidney fibrosis compared to healthy
   kidney and lung ([176]Figure 3E, [177]Extended Data Figure 4G). Ang2 is
   associated with perivascular remodeling in the primary tumor and its
   impact on endothelial cell signaling plays a rate limiting role in
   metastatic dissemination ^[178]13,[179]23–[180]29. While our prior
   studies demonstrate the importance of vascular remodeling by Ang2 at
   the primary tumor site ^[181]13,[182]25, the impact of vascular
   heterogeneity and permeability in determining organotropic metastasis
   is unknown. Validation studies using quantitative ELISA showed an
   increase in the levels of Ang2 in the serum of mice with kidney
   fibrosis when compared to control mice, with the levels of Ang2 being
   similar to those found in mice with 4T1 mammary carcinoma (T, tumor
   bearing mice, [183]Figure 3F). These results suggest that Ang2 is
   upregulated in the serum of mice with lung metastasis associated with
   kidney fibrosis, whereas its impact may be locally driven in the
   context of lung fibrosis. Interestingly, circulating Ang2 has also been
   reported as being increased in patients with breast cancer where
   elevated levels are associated with reduced survival, higher
   histological grade, more advanced clinical stage, and presence of lymph
   node metastasis^[184]30. Moreover, breast cancer patients with three or
   more metastatic sites demonstrate increased circulating Ang2 compared
   to patients with less metastatic sites^[185]31. Taken together, the
   results implicate Ang2 as a potential mediator of both kidney fibrosis
   and lung fibrosis-induced enhancement of lung metastasis.

Ang2 compromises lung vascular permeability in fibrosis-induced lung
metastasis

   We next evaluated whether the observed increase in circulating Ang2
   plays a role in vascular permeability at the metastatic site to promote
   colonization of lungs by cancer cells. We treated healthy mice with
   recombinant mouse Ang2 protein (rAng2) and assessed the lungs, kidneys,
   and livers from these mice. rAng2 was administered intraperitoneally
   before intravenous infusion of FITC-conjugated dextran. Enhanced
   extravasation of FITC-dextran was observed in the lungs after rAng2
   administration compared to controls, indicating that rAng2 increases
   vascular permeability in the lungs ([186]Figure 3G, [187]Supplementary
   Figure 2). In contrast, no increase in vascular permeability was noted
   in the kidney and liver following rAng2 administration ([188]Figure 3H,
   [189]Extended Data Figure 5C, [190]Supplementary Figure 2). The lung
   parenchyma of mice administered rAng2 showed significant edema and
   increased pleural thickness compared to controls ([191]Figure 3I,
   [192]Extended Data Figure 6A). This is similar to the histopathological
   changes noted in fibrotic lungs and lungs of mice with kidney fibrosis
   ([193]Extended Data Figure 6B-[194]E). There was no appreciable change
   in the kidney parenchyma following rAng2 administration ([195]Extended
   Data Figure 6F) or in the kidneys of mice with lung or kidney fibrosis
   ([196]Extended Data Figure 6G-[197]H).

   Exogenous administration of Ang2 (rAng2) was sufficient to
   significantly enhance lung colonization of intravenously administered
   4T1 cancer cells ([198]Figure 3J, [199]Extended Data Figure 6I), with
   no localization observed in the kidney ([200]Extended Data Figure 6J).
   These results are consistent with the increased 4T1 lung colonization
   observed in the context of lung or kidney fibrosis ([201]Figure
   1O-[202]P). Collectively, these results support the effect of Ang2 on
   lung vascular permeability and its consequent impact on lung metastasis
   without measurable impact in the kidney.

Ang2 neutralization reverses kidney fibrosis-induced enhanced lung metastasis

   To determine the rate limiting role of Ang2 on lung metastasis, we
   evaluated primary tumor growth kinetics and metastatic disease in mice
   treated with anti-Ang2 neutralizing antibodies (anti-Ang2) ^[203]25.
   Evaluation of 4T1 primary tumors revealed similar growth kinetics in
   control IgG treated mice with or without kidney fibrosis, while
   anti-Ang2 treatment significantly reduced tumor burden in both groups
   (with and without kidney fibrosis) by approximately 30% ([204]Figure
   4A). These results are consistent with previous reports ^[205]25 and
   likely reflect the impact of anti-Ang2 on intratumoral vascular
   permeability. In the absence of fibrosis, mice with 4T1 orthotopic
   tumors treated with either control IgG or anti-Ang2 presented with
   similar amounts of lung metastasis ([206]Figure 4B-[207]D). Anti-Ang2
   treatment reduced the lung metastasis area and the number of
   microscopic lung nodules in mice with kidney fibrosis to the baseline
   levels observed in control (no kidney fibrosis) groups ([208]Figure
   4B-[209]D). Histopathological analyses of the kidneys revealed that
   anti-Ang2 treatment did not impact kidney fibrosis when compared to the
   lgG control group ([210]Figure 4E-[211]G). Our results support a
   rate-limiting role for systemic circulating Ang2 released by kidney
   fibrosis, with an impact on lung metastasis associated with mammary
   carcinoma.

Figure 4. Ang2 neutralization is sufficient to revert lung metastasis induced
by kidney fibrosis.

   Figure 4.
   [212]Open in a new tab

   (A) Orthotopic 4T1 tumor volume over time +/− kidney fibrosis (KF) and
   IgG or anti-Ang2 treatment, n=9 mice per group. Arrows indicate a ~30%
   reduction in tumor volume at experimental endpoint with anti-Ang2
   compared to IgG (ctrl groups). (B-C) Quantification of metastatic area
   (B) and number of microscopic lung nodules (C) in the indicated groups,
   n=9 mice per group. (D) Representative lung H&E of mice in the
   indicated groups. Scale bar, 25 μm. (E) Representative images of
   kidneys with H&E, MTS, and Sirius Red staining from the indicated
   groups. Scale bar, 25 μm. (F-G) Quantification of positive MTS (F) and
   Sirius Red stained area (G) in the indicated groups, n=9 mice per
   group. (H) Relative mRNA expression levels of Cldn2, Cldn3, Cldn4,
   Cldn5, Cldn7, Cldn10, Cldn17, Cldn19, Jam2, Jam3, Occludin and Tjp1 in
   healthy kidney, lung, and liver tissues. Expression value is normalized
   to endogenous Pecam1 (CD31) levels and is presented as 2^-ΔCt, n=3 mice
   per group. Data are presented as mean values +/− SEM. For A: two-way
   ANOVA with Tukey’s multiple comparisons test. For B, C, F and G:
   one-way ANOVA test with Tukey’s multiple comparisons test. P values are
   listed, ns: not significant.

Ang2–Foxo1 signaling suppresses Claudin 5 in lung endothelial cells

   Ang2 has been previously implicated in breast cancer metastasis via the
   impairment of tight junction (TJ) integrity ^[213]32. Claudins (Cldns)
   are critical structural and functional components of TJ intercellular
   adhesion complexes ^[214]33. Evaluation of key TJ components, including
   Cldn2, Cldn3, Cldn4, Cldn5, Cldn7, Cldn10, Cldn17, Cldn19, Jam2, Jam3,
   Occludin and Tjp1, showed that Cldn5 was uniquely expressed at higher
   levels in healthy lungs when compared to healthy liver and kidney
   ([215]Figure 4H). Single-cell RNA transcriptome sequencing (scRNA-seq)
   of murine endothelial cells across 11 tissues ^[216]34, revealed that
   Cldn5 was highly expressed in endothelial cells from the lung, brain,
   and testes, and could also be found in kidney and liver endothelial
   cells ([217]Figure 5A-[218]B). Notably, Cldn5 was uniquely enriched in
   endothelial cells of both murine and human scRNA-seq analyses of
   healthy lungs ^[219]35 ([220]Figure 5C-[221]E, [222]Extended Data
   Figure 7A-[223]C), whereas the expression of Claudins in the kidney
   endothelium revealed more diversity, with not only Clnd5, but also
   Cldn2, 3, 4, 6, 7, 8, 10, 12, 16, and 19 noted in murine kidney
   endothelial cells ^[224]36 ([225]Figure 5F-[226]G, [227]Extended Data
   Figure 7D), and CLDN2–5, 7, 8, 10, 12, 15 and 23 in human kidney
   endothelial cells ^[228]37 ([229]Extended Data Figure 7E).
   Immunolabeling experiments confirmed the expression of Cldn5 in
   CD31-expressing endothelial cells in the liver, brain, kidney, and lung
   tissues ([230]Figure 5H, [231]Supplementary Figure 3A-[232]D).
   Immunolabeling studies showed Cldn5 was more frequently associated with
   CD31^+ vascular endothelial cells rather than Lyve1^+ lymphatic
   endothelial cells ([233]Figure 5 H-[234]I, [235]Extended Data Figure
   8A-[236]C, [237]Supplementary Figures 4A-[238]B). In addition to Cldn5,
   other Claudins are expressed in murine kidney endothelial cells
   ([239]Figure 5G, [240]Extended Data Figure 8D-[241]G). Cldn2 and Cldn10
   were also expressed in both endothelial and epithelial TJs in the
   kidney and brain parenchyma ^[242]38. Immunolabeling for Cldn2 and
   Cldn10 co-localized with CD31^+ vascular endothelial cells in both the
   cortex and medulla ([243]Extended Data Figure 8H-[244]I,
   [245]Supplementary Figure 5A-[246]D). scRNA-seq data also pointed to
   diverse Cldn genes expressed epithelial and endothelial cells in murine
   kidney, whereas lung and brain endothelial cells show specific Cldn5
   expression ([247]Extended Data Figure 8D-[248]G). Collectively, these
   studies support a role for Cldn5 in the lung (and brain) endothelial
   TJs, and a role for multiple Claudins, including Cldn5, Cldn2 and
   Cldn10 in kidney endothelial TJs. Given the enriched Cldn5 expression
   in lung endothelial TJs, we sought to determine whether the increased
   lung vascular permeability induced by Ang2 was, in part, a result of
   Cldn5 downregulation as previously supported in other studies
   ^[249]39,[250]40. The lungs of mice with lung fibrosis, kidney
   fibrosis, or rAng2 treatment consistently revealed a reduced
   endothelial Cldn5 expression when compared to control lungs
   ([251]Extended Data Figure 8J-[252]K, [253]Supplementary Figure
   6A-[254]D). Inhibition of Ang2 signaling in kidney fibrosis with
   neutralizing anti-Ang2 administration failed to show a reduction in
   endothelial Cldn5 ([255]Extended Data Figure 8J-[256]K,
   [257]Supplementary Figure 6A-[258]D). Taken together, the results
   support Ang2 downregulate Cldn5 in vivo in murine lung endothelial
   cells.

Figure 5. Differential Claudin expression levels in lung and kidney.

   [259]Figure 5.
   [260]Open in a new tab

   (A-B) t-SNE plots (A) and violin plots (B) of scRNA-seq evaluation of
   Cldn5 mRNA levels in endothelial cells from the listed organs. (C-D)
   t-SNE plot of sRNA-seq of murine lung cells (C) with a focus on
   endothelial cells (EC) showing the relative mRNA levels for Cldn2,
   Cldn5 and Cldn10 (D). (E) tSNE plots and analysis of Claudin 5 and CD31
   colocalization percentage in mouse lung endothelial cells (EC). (F-G)
   t-SNE plot of sRNA-seq of murine kidney cells (F) with a focus on
   endothelial cells (EC) showing the relative mRNA levels for Cldn2,
   Cldn5 and Cldn10 (G). (H) Colocalization of CD31 (green) and Claudin 5
   (red) in WT liver, brain, and kidney. Scale bar, 25 μm (individual
   images in [261]Supplementary Figures 3, [262]4A). (I) Colocalization of
   Lyve-1 (green) and Claudin 5 (red) in lung. Scale bar, 25 μm
   (individual images in [263]Supplementary Figure 4B). Data are presented
   as mean values +/− SEM.

   To define the signaling events associated with Ang2 mediated
   downregulation of Cldn5, we evaluated FOXO1 in lung endothelial cells
   treated with rAng2. Previous studies reported that, in an inflammatory
   state with down regulation of Tie1, Ang2 antagonist action on Tie2
   leads to FOXO1 activation in endothelial cells ^[264]40. FOXO1
   activation and nuclear translocation results in downregulation of Cldn5
   in endothelial cells ^[265]39. Murine lung microvascular endothelial
   cells showed transcriptional downregulation of Cldn5 upon Ang2
   treatment ([266]Figure 6A-[267]B). Cldn5 protein levels were also
   downregulated in lung endothelial cells following rAng2 treatment
   ([268]Figure 6C-[269]E), and this was associated with a nuclear
   enrichment in FOXO1 ([270]Figure 6F). Jam2, Cldn3 and Cldn4 mRNA levels
   were unaltered ([271]Figure 6G). Decreased expression of Cldn5 was also
   noted in brain endothelial cell following rAng2 treatment ([272]Figure
   6H), corroborating previous findings ^[273]32.

Figure 6. Ang2 downregulates Claudin 5 via nuclear enrichment of FOXO1.

   [274]Figure 6.
   [275]Open in a new tab

   (A) Experimental design for the evaluation of Ang2 mediated Claudin 5
   downregulation in the lung endothelial cells. (B) Relative Cldn5 mRNA
   levels in lung endothelial cells treated with the rAng2 (PBS control
   normalized to 1) for the indicated timepoints, n=3 biologically
   independent experiments (n=3 PBS and n=3 rAng2 with each 2 technical
   replicates). (C-D) Relative Claudin 5 protein levels in lung
   endothelial cells treated with rAng2 for the indicated timepoints, n=5
   biologically independent experiments (n=5 PBS and n=5 rAng2). Uncropped
   blots are shown in the Image Source Data File. (E) Representative
   immunofluorescence staining of Claudin 5 in lung endothelial cells
   after treatment with rAng2 for 24h. n=1 biologically independent
   experiment. Scale bars, 25 μm. (F) FOXO1 protein levels in nuclear and
   cytoplasmic fractions at the indicated timepoints, with alpha-tubulin
   and Lamin A/C as markers for cytoplasmic and nuclear fraction,
   respectively. n=1 biologically independent experiment. Uncropped blots
   are shown in the Image Source Data File. (G) Relative Jam2, Cldn3 and
   Cldn4 mRNA levels of in lung endothelial cells after treatment with
   rAng2 for 24h, n=3 biologically independent experiments (with each 3
   technical replicates). (H) Relative Cldn5 mRNA levels of in brain
   endothelial cells after treatment with the rAng2 for 24h, n=3
   biologically independent experiments (with each 2 technical
   replicates). Data are presented as mean values +/− SEM. Unpaired
   two-tailed t test, with Welch’s correction applied for unequal
   variances (determined by F-test). P values are listed, ns: not
   significant.

Claudin 5 tight junction compromised by Ang2 enhances lung metastasis

   To determine the direct contribution of Cldn5 reduction by Ang2 in lung
   metastasis, we performed rescue experiments where Cldn5 was
   overexpressed via the intratracheal administration of Cldn5
   overexpressing adenovirus (Ad-Claudin 5) before the intravenous
   infusion of 4T1 cancer cells, with or without rAng2 administration
   ([276]Figure 7A). Following the in vitro validation of the reagents
   ([277]Extended Data Figure 9A-[278]B; Ad-ctrl: control adenovirus), in
   vivo studies revealed a decrease in Cldn5/CD31 double positive area
   after rAng2 treatment by immunofluorescence labeling, while
   overexpression of Cldn5 induced by intratracheal Ad-Claudin 5 prevented
   such downregulation ([279]Figure 7B-[280]C, [281]Supplementary Figure
   7). Consistent with our previous findings ([282]Figure 3I-[283]J),
   rAng2 treatment significantly increased 4T1 lung colonization
   ([284]Figure 7D-[285]E). Concurrent overexpression of Cldn5 in this
   context, however prevented the increased 4T1 lung colonization
   ([286]Figure 7D-[287]E).

Figure 7. Critical role of Claudin 5 in 4T1 cell extravasation in lung.

   Figure 7.
   [288]Open in a new tab

   (A) Experimental design for the evaluation of Cldn5 overexpression
   (Ad-Claudin 5) in mice with 4T1 i.v. injection +/− rAng2 treatment. (B)
   Representative colocalization of Claudin 5 and CD31 in the lung of mice
   in the indicated groups (individual images in [289]Supplementary Figure
   7). Scale bars, 50 μm. (C) Quantification of percentage of Claudin
   5/CD31 double positive area per visual field in the lung of mice in the
   indicated groups, n=4 to 5 mice per group. (D) Representative lung H&E
   of mice in the indicated groups. Scale bar, 25 μm. (E) Quantification
   of lung colonization area of mice in the indicated groups, n=4 to 5
   mice per group. (F) Experimental design for the evaluation of the
   downregulation of Claudin 5 (AAV-shRNA) in mice with 4T1 orthotopic
   tumors. Dox: doxycycline in the drinking water (Dox on). (G)
   Representative colocalization of Claudin 5 and CD31 in the lung of mice
   in the indicated groups individual images in [290]Supplementary Figure
   8). Scale bar, 50 μm. (H) Quantification of percentage of Claudin
   5/CD31 double positive area per visual field in the lung of mice in the
   indicated groups, n=4 to 5 mice per group. (I) Representative lung H&E
   of mice in the indicated groups. Scale bar, 25 μm. (J) Quantification
   of lung metastatic area in mice in the indicated groups, n=4 to 5 mice
   per group. Data are presented as mean values +/− SEM. One-way ANOVA
   with Tukey ‘s multiple comparisons test. P values are listed.

   Next, the downregulation of Cldn5, using doxycycline controlled Cldn5
   shRNA AAV administered intratracheally was performed to determine
   whether the suppression of Cldn5 was sufficient to increase lung
   metastasis of 4T1 mammary carcinoma ([291]Figure 7F, [292]Extended Data
   Figure 9C). A decrease in Cldn5/CD31 double positive area was confirmed
   using immunolabeling ([293]Figure 7G-[294]H, [295]Supplementary Figure
   8). The downregulation of Cldn5 in the lung was found to be sufficient
   to increase lung metastatic disease when compared to controls
   ([296]Figure 7I-[297]J), with no impact on the growth kinetics of the
   primary 4T1 tumor growth ([298]Extended Data Figure 9D). Additionally,
   reduced CLDN5 mRNA expression in metastatic lungs was found in patients
   with breast cancer compared to healthy lung ([299]Extended Data Figure
   9E). Taken together, our results support a role for vascular Cldn5 in
   the lung as one of the gatekeepers of metastasis.

Claudin 2/10 suppression enables metastasis to the kidney

   We next assessed whether the loss of Cldn2 and Cldn10, along with Ang2
   mediated suppression of Cldn5, would be sufficient to generate a
   metastasis-permissive vascular microenvironment due to compromised
   vascular permeability in the kidney. We employed CRISPR/Cas9 knockdown
   of Cldn2 and Cldn10 expression in kidneys ^[300]41, and confirmed that
   this approach yields a significant reduction in Cldn2 and Cldn10 in
   mouse kidney endothelial cells ([301]Extended Data Figure 9F-[302]J).
   Fluorescence-activated cell sorting (FACS) purified CD31^+ kidney
   endothelial cells revealed lower Cldn2 and Cldn10 mRNA levels compared
   to E-Cadherin^+ epithelial cells ([303]Extended Data Figure 9H).
   Twenty-six days after administration of CRISPR/Cas9 constructs directly
   into the renal cortex to suppress Cldn2 or Cldn10 in mice with 4T1
   tumors, we confirmed a reduction in Cldn2 and Cldn10 in the FACS
   purified CD31^+ endothelial cells, albeit with limited downregulation
   in E-Cadherin^+ epithelial cells ([304]Extended Data Figure 9I-[305]J,
   [306]Supplementary Figure 9A). 4T1 tumors showed similar primary tumor
   growth kinetics ([307]Extended Data Figure 9K-[308]L) and lung
   metastasis when Cldn2/10 were downregulated in the kidney
   ([309]Extended Data Figure 9M-[310]O). Importantly, kidney metastasis,
   evaluated by histopathology and immunolabeling for GFP^+ cancer cells,
   was detected in mice with Cldn2/10 downregulation compared to controls
   ([311]Extended Data Figure 10A-[312]E), effectively demonstrating the
   formation of metastasis to a non-organotropic location for 4T1 mammary
   carcinoma. This experiment was repeated with inclusion of additional
   experimental arm, in which mice were also subjected to kidney fibrosis
   ([313]Figure 8A). In this context, lung metastasis was increased with
   kidney fibrosis without impact from the suppression of Cldn2/10, and no
   impact on primary tumor growth was also observed ([314]Figure
   8B-[315]C, [316]Extended Data Figure 10F). Suppression of Cldn2/10 in
   the kidney revealed kidney metastasis, which was further increased with
   kidney fibrosis ([317]Figure 8D-[318]G, [319]Extended Data Figure 10G,
   [320]Supplementary Figure 9B).

Figure 8. Knockdown of Claudin 2/10 in kidney fibrosis enables renal
metastasis.

   Figure 8.
   [321]Open in a new tab

   (A) Experimental design for the evaluation of the downregulation of
   Claudin 2 and Claudin 10 in kidney together with induction of kidney
   fibrosis in mice with 4T1 orthotopic tumors. (B) Representative lung
   H&E of mice in the indicated groups. Scale bar, 25 μm. (C)
   Quantification of lung metastatic area. LentiCRISPRv2 control vector
   without fibrosis, n=14; LentiCRISPRv2 control vector with fibrosis,
   n=11; LentiCRISPRv2 Cldn2 and Cldn10 without fibrosis, n=12 mice;
   LentiCRISPRv2 Cldn2 and Cldn10 with fibrosis, n=10 mice. (D)
   Representative kidney H&E of mice in the indicated groups. Scale bar,
   50 μm (upper), 25 μm (lower). (E-F) Representative lung (E) and kidney
   (F) images for GFP immunolabeling in the indicated groups and
   associated quantification. E: scale bar, 50 μm, F: scale bar, 25 μm.
   (G) Percentage of mice with kidney lesion in the indicated groups. (H)
   Graph abstract. Data are presented as mean values +/− SEM. For C, E:
   one-way ANOVA with Tukey’s multiple comparisons test. For F:
   Kruskal–Wallis with Dunn’s multiple comparison test. For G: two-sided
   Chi-square. P values are listed, ns: not significant.

   Intravascular cancer cells were observed in the kidney ([322]Extended
   Data Figure 10H, [323]Supplementary Figure 10A-[324]B) but the
   metastatic growth in the kidney was detected in the parenchyma only
   upon vascular tight junction compromise ([325]Extended Data Figure 10I,
   [326]Supplementary Figure 10C). CD31 staining revealed that the vessel
   numbers were unchanged and suppression of Cldn 2/10 in kidney had no
   impact on primary mammary tumor vasculature ([327]Extended Data Figure
   10J-[328]K). Collectively, our data indicates that destabilization of
   kidney endothelial cell TJs and the compromised vascular permeability
   resulting from suppressing Cldn2/10 enables the atypical formation of
   metastasis to the kidney, which is enhanced with inflammatory fibrosis.
   Our findings show that Cldn5 in the lung is uniquely susceptible to
   Ang2 mediated vascular destabilization and lung metastasis, and
   organotropism of metastasis, at least in part, is driven by endothelial
   TJ tissue specific composition ([329]Figure 8H).

Discussion

   Organ specific tropism of metastasis associated with different cancers
   has long been a subject of intrigue and scientific investigation
   ^[330]42. The question as to why certain cancers rarely result in
   metastasis to certain organs but do so with high frequency to other
   organs remains under explored. Here, we employed mouse models of
   metastatic mammary carcinoma to evaluate the molecular underpinnings of
   frequent breast cancer metastasis to the lung but not to the kidney.
   Collectively, our studies demonstrate that when the extravasation step
   is bypassed, mammary carcinoma cells become equally well adapted to
   grow in the lungs and kidneys, and form established tumors. This
   finding suggests that cancer cell intrinsic mechanisms are not
   sufficient to enable their growth into tumors in the kidney, as shown
   with the models studied here. In different spontaneous metastasis
   models of mammary carcinoma that depend on extravasation of cancer
   cells, it remains clear how growth of secondary tumors in the kidney is
   prevented despite robust growth of metastatic tumors in the lung. In
   this experimental setting, the absence of metastasis to the kidney
   seems to depend on the vasculature-mediated barrier against
   extravasation of cancer cells, rather than on cancer cell intrinsic
   mechanisms and pre-metastatic niche. Previous studies that conducted
   extravasation and colonization assays using i.v. infusion of melanoma
   cells and ectopic tissue implantation revealed the critical role of the
   local environment in influencing lung and kidney metastasis ^[331]43.
   However, this approach involved the injection of several-fold higher
   numbers of cultured melanoma cells directly into circulation, unlike
   the number of circulating cancer cells originating from a primary tumor
   ^[332]43. Hence, the rationale behind this selective pattern of tumor
   growth is likely more intricate than tissue damage alone that creates a
   conducive microenvironment. This also suggests the involvement of
   different vascularization patterns on influencing metastasis^[333]43.

   Ang2 expression is upregulated by hypoxia, a known feature of the
   primary tumor microenvironment and inflammatory fibrotic
   microenvironment (IFM). In this study, the induction of IFM in either
   the lung or kidney leads to secretion of Ang2 along with the
   simultaneous activation of inflammatory signaling networks. Notably,
   vasculature response played a critical role in amplifying these
   inflammatory signals, leading to the formation of a pro-metastatic
   vascular niche ^[334]44,[335]45. Our results support that Ang2
   increases vascular permeability in the lung but not the kidney, leading
   to efficient extravasation of the cancer cells into the lung and
   enabling metastatic colonization in the lung but not the kidney.
   Despite the findings reported by Chang et al. that Ang2 inhibition
   using a peptide-Fc fusion (L1–10) attenuated kidney fibrosis ^[336]46,
   our experiments using anti-Ang2 neutralizing antibodies did not yield
   any significant impact on renal fibrosis. This discrepancy in results
   could be attributed to variations in treatment modality, dosing,
   timing, and readouts. Robust mammary carcinoma cell growth and tumor
   formation upon direct injection into the renal tissue support that a
   mechanism maintaining vascular integrity, even when endothelial cells
   are exposed to Ang2, is a major factor underlying the lack of
   metastasis to the kidneys. Importantly, many host factors can
   contribute to emergence of metastasis in a given tissue and here we
   demonstrate that molecular features of vasculature also plays a causal
   role in restricting metastasis^[337]43.

   Single cell analyses of endothelial cells from the lung and the kidney
   revealed that the endothelial TJ in the lung predominantly contains
   Cldn5, which can be downregulated by Ang2, while the endothelial TJ in
   the kidney contains low levels of Cldn5 along with two non-Ang2
   responsive Claudins–Cldn2 and Cldn10. Ang2, a known inducer of lung
   vascular permeability ^[338]47, regulates Cldn5 downregulation via
   engagement of FOXO1 signaling ^[339]39,[340]40. A positive correlation
   has been reported between pulmonary fibrosis and circulating Ang2
   levels ^[341]48, and Ang2 can directly provoke lung vascular
   hyperpermeability and induce pulmonary edema ^[342]47,[343]49,[344]50.
   Ang2 is a known ligand of the receptor tyrosine kinase Tie2, and the
   activation of Tie1/2 signaling by angiopoietins is not only important
   for vascular homeostasis, but also critical in the tumor angiogenic
   switch and metastasis ^[345]51. Specifically, circulating Ang2 levels
   are associated with disease progression and survival in metastatic
   malignant breast cancer ^[346]30,[347]52, and a positive correlation of
   Ang2 and a pro-metastatic outcome has been reported
   ^[348]25,[349]32,[350]53. Our data support a direct role of Ang2 in
   lung vascular permeability, making this organ specifically permissive
   to metastasis. Direct suppression of Cldn5 in the lung vasculature
   results in increased lung metastasis. Rescue studies with
   overexpression of Cldn5 in the context of elevated Ang2 further
   implicates Cldn5 amongst the gatekeepers of lung metastasis.

   CRISPR mediated suppression of Cldn2 and Cldn10, along with Ang2
   mediated suppression of Cldn5 in the kidney vasculature, enabled
   extravasation of the cancer cells and resulted in mammary carcinoma
   metastasis to the kidney. These studies directly implicate a role for
   vascular endothelial cells as potential rate-limiting regulators of
   organ specific metastasis. Our proof-of-concept studies support that
   heterogeneity within endothelial cell junctions in different organs is
   likely a key contributor of organotropism associated with mammary
   carcinoma and support further testing of Ang2 targeting to control lung
   metastasis.

Methods

   All research presented here complies with ethical regulation, namely
   the Institutional Animal Care and Use Committee of MDACC. Further
   information on research design is available in the Nature Research
   Reporting Summary linked to this article.

Cell Culture

   4T1-GFP cells (4T1-turbo GFP and 4T1-luc-GFP) were modified using 4T1
   BALB/c mammary carcinoma epithelial cells (Expasy RRID: CVCL 0125, from
   the Characterized Cell Line Core laboratory at MD Anderson Cancer
   Center), transduced to stably express TurboGFP (Origene, TR300007), or
   infected to stably express eGFP and luciferase (as previously described
   ^[351]16. 4T1 cells were maintained in RPMI 1640 medium supplemented
   with 10% heat-inactivated FBS (Gemini), penicillin (100 U/mL) and
   streptomycin (100 μg/mL) (Corning, VWR45000–652). When needed, culture
   medium was also supplemented with 1 μg/mL puromycin (Sigma-Aldrich,
   P8833). HEK293T/17 cells were provided by Dr. Lynda Chin, MD Anderson
   Cancer Center, and maintained in DMEM medium supplemented with 10%
   heat-inactivated FBS (Gemini), streptomycin (100 μg/mL) solution
   (Corning, VWR45000–652) and penicillin (100 U/mL) at 37°C and 5% CO[2]
   as described before^[352]41. Mouse kidney endothelial cells (MKECs) and
   NIH-3T3 cells were kindly provided by Dr. Isaiah J. Fidler, MD Anderson
   Cancer Center. MKECs were evaluated for microscopic morphology without
   further characterization, and they were maintained in endothelial cell
   medium (Cell Biologics, M1168) together with the endothelial cell
   medium supplement kit (Cell Biologics, M1168-Kit). NIH-3T3 cells were
   maintained in DMEM medium supplemented with 10% heat-inactivated FBS
   (Gemini), penicillin (100 U/mL) and streptomycin (100 μg/mL) solution
   (Corning, VWR45000–652). C57BL/6 mouse primary lung microvascular
   endothelial cells (Cell Biologics, C57–6011) and C57BL/6 mouse primary
   brain microvascular endothelial cells (Cell Biologics, C57–6023) were
   maintained in complete mouse endothelial cell medium (Cell Biologics,
   M1168) on gelatin (Cell Biologics, 6950) coated culture dishes. The
   cells were incubated in 0.25% FBS for 24h before treatment with murine
   recombinant Ang2 (rAng2, R&D Systems, 7186-AN-025, 100 ng/mL) or PBS
   (control) for the indicated timepoints (6h, 12h, and 24h) and harvested
   for RNA isolation based on previous studies ^[353]47. For the primary
   lung microvascular endothelial cells, the cells were treated with rAng2
   (100 ng/mL) or PBS for different timepoints before starvation in 0.25%
   FBS for 24h and harvested into lysis buffer for western blot.
   HEK293T/17 cells were validated by STR analysis at the UT MD Anderson
   Cytogenetics and Cell Line Authentication Core. 4T1 and NIH-3T3 cells
   identity were validated via CellCheck (IDEXX BioResearch). All the
   cells were routinely screened to confirm the absence of mycoplasma
   contamination. C57BL/6 mouse primary lung microvascular endothelial
   cells (Cell Biologics, C57–6011) were found positive for mycoplasma and
   were treated, and subsequently found negative for mycoplasma prior to
   use in experiments.

Virus

   For the overexpression of Cldn5, Ad-CMV-CLDN5v2 with His tag (Vigene
   Biosciences, Catalog number VH809267, Lot 100001) and Adeno-CMV-luc
   control (Vigene Biosciences, Catalog number 1000, Lot A1022t#6M13) were
   purchased from the listed vendor. For the downregulation of Cldn5 gene
   expression, AAV9-mCldn5-o4-shRNA (Vector Biolab, custom production,
   shAAV-180716#17) and AAV9-anti-Luc-Control-shRNA (Vector Biolab, custom
   production, shAAV-Lot180716#18) were purchased from the listed vendor.
   For the overexpression of Cldn5, NIH-3T3 (5 × 10^4) cells were
   transduced with 10 μL (1 × 10^7 PFU/μL, total 1× 10^8 PFU)
   Ad-CMV-CLDN5v2 or 10 μL (1 × 10^7 PFU/μL, total 1 × 10^8 PFU)
   Adeno-CMV-luc for 48 hours followed by RNA isolation and protein
   extraction. Control parental cells were not transduced. For the
   downregulation of Cldn5, NIH-3T3 cells were transduced with
   AAV9-mCldn5-shRNA. NIH-3T3 (5 × 10^4) cells were seeded in complete
   medium with or without doxycycline (1 μg/mL), parental cells without
   any treatment were used as control. After 24 hours, the cells were
   transduced with 5 μL (1.1 × 10^10 genome copies (GC)/μL, with 10-fold
   dilution, total 5.5 × 10^9 GC) AAV9-mCldn5-o4-shRNA or 5 μL (1.1 ×
   10^10 GC/μL, with 10-fold dilution, total 5.5 × 10^9 GC)
   AAV9-anti-Luc-Control-shRNA. RNA was isolated 72 hours after the
   transduction.

   To downregulate Cldn2 and Cldn10 gene expression using CRISPR-Cas9,
   Cldn2 or Cldn10 targeting sgRNAs were designed using an online tool
   ([354]http://crispr.dfci.harvard.edu/SSC/). The sequences used are
   shown in [355]Supplementary Table 5. LV/Cas9-Cldn2-sgRNA1,2 and
   LV/Cas9-Cldn10-sgRNA1,2,3 were generated by employing BsmBI to insert
   each sgRNA into the RNA scaffold of lentiCRISPRv2 vector (Addgene
   #52961). The generation of lentiviruses and concentration measurements
   were performed as previously described ^[356]41. The titration of
   lentivirus was determined by utilizing the Lenti-X qRT-PCR Titration
   Kit (LV900, Applied Biological Materials Inc) in accordance with
   manufacturer’s instructions. For the transduction in vitro, mouse
   kidney endothelial cells (MKEC) were maintained in endothelial cell
   medium (Cell Biologics, M1168) and incubated with 5 μg/mL polybrene
   (EMD Millipore TR-1003-G, CA, US) and 5 μL (2 × 10^8 IU/μL) viral
   solution together with for 60 hours. Fresh RPMI medium supplemented
   with puromycin (1 μg/mL) was replaced every two days post-transduction
   before the establishment of LV/Cas9-Cldn2-sgRNA1,2-MKEC and
   LV/Cas9-Cldn10-sgRNA1,2,3-MKEC stable cell lines. After 5 to 7 days,
   RNA was isolated for subsequent quantitative real-time PCR analyses
   (described in a separate section below).

Mice

   Source data is listed for all tumor experiments in the associated
   source data file for each Main and Extended Data Figure in this study.
   For orthotopic implantation, BALB/c virgin female mice (Charles River,
   strain code 028) between the ages of 6 to 12 weeks were inoculated with
   4T1 cells in each mammary pad (5 × 10^5 4T1 cells in the right or left
   mammary fat pad). Tumor burden was assessed using Vernier calipers and
   volumes were calculated as previously described ^[357]13. Surface lung
   nodules were counted at autopsy. When the sum of averaged tumors on
   each flank reached 500 mm^3, fibrotic insults were induced in the lung
   via intratracheal administration of bleomycin (25 μL of 0.15U in saline
   solution, Sigma-Aldrich, B5507–15), and in the kidney via surgical
   manipulation (unilateral ureteral obstruction surgery, UUO) ^[358]54.
   Full-thickness cutaneous wound surgery was performed as previously
   described ^[359]55.

   For the cancer cell colonization studies, 5 × 10^5 4T1 cells were
   injected intravenously (4T1 i.v.) into the retro-orbital venous sinus
   of BALB/c virgin female mice aged between 6 to 12 weeks. 4T1 i.v.
   injection was performed on the same day as the intratracheal bleomycin,
   UUO surgery, and cutaneous wound surgery.

   MMTV-PyMT (Pymt) transgenic mice backcrossed on the BALB/c background
   spontaneously develop tumors in their mammary fat pads between the ages
   of 8 to 12 weeks ^[360]56. Fibrotic events were induced in the lung and
   kidney when the cumulative tumor volumes reached 500 mm^3, and the
   cutaneous wounding was generated as described above.

   For both orthotopic implantation and MMTV-PyMT transgenic mice, all the
   mice were euthanized before or upon reaching a tumor volume of
   approximately 2,000 mm^3, as permitted by the Institutional Animal Care
   and Use Committee of MDACC (with a goal of a maximum of 2,000 mm^3
   using regular evaluation of tumor size with calipers and prompt
   euthanasia if volume measurement were found to exceed this limit).
   Following i.v. cancer cell injection, the mice were euthanized on day
   10 after lung and kidney fibrosis induction, and on day 14 after wound
   surgery.

   For the ultrasound-guided intracardiac injection, BALB/c virgin female
   mice aged between 6 to 12 weeks were separated into 2 groups receiving
   either UUO or sham surgery (Day 0). The next day, 4T1 cancer cells (5 ×
   10^5) were injected into the left cardiac ventricle via using
   ultrasound machine (VisualSonics Vevo-2100, Day 1). All the mice were
   euthanized after 10 days (Day 10).

   4T1 (5 × 10^5) cells were injected under the renal capsule of BALB/c
   virgin female mice aged between 6 to 12 weeks. The mice were separated
   into 2 cohorts receiving either UUO or sham surgery to the injected
   kidney. All the mice were euthanized after 10 days. The mice were
   euthanized before or at early signs of moribundity.

   In some experiments, mice received recombinant murine Ang2 (100 μg/mL
   in 100 μL PBS) or PBS (100 μL) control intraperitoneally on Day 1 and
   Day 3 (2 times) and subsequently euthanized on Day 4 (24h after the
   last rAng2 administration). In mice with 4T1 i.v., rAng2 or PBS was
   intraperitoneally injected one day before and one day after i.v.
   injection of 4T1 cells.

   In experiments in which Ang2 was neutralized, kidney fibrotic events
   were induced by UUO when mice presented with an averaged tumor burden
   of 500 mm^3 and mice were treated with control isotype-matched IgG (10
   mg/kg BW in 200 μL PBS) or anti-Ang2 neutralizing antibody (murinized
   LC06, Roche Diagnostics, 10 mg/kg BW in 200 μL PBS) once a week (2
   times) as previously reported ^[361]25.

   Overexpression of Claudin 5 was induced in the lung via intratracheal
   administration of 20 μL (1 × 10^7 PFU/μL) Ad-CMV-CLDN5v2 and 20 μL (1 ×
   10^7 PFU/μL) Adeno-CMV-luc. After 24 hours (day 1), the mice were
   treated with intraperitoneal injection of rAng2 (as described above) or
   PBS. On day 3, the mice were injected i.v. with 4T1 cells (5 × 10^5)
   cells. On day 4, the mice were treated once more with rAng2 or PBS.

   Down-regulation of Claudin 5 in the 4T1 orthotopic model was achieved
   by intratracheal administration of AAV9-mCldn5-o4-shRNA. On day 1, the
   mice were implanted orthotopically with 4T1 cells (5 × 10^5) cells.
   After 10 days, the mice were intratracheally administered 20 μL (1.1 ×
   10^10 GC/μL) AAV9-mCldn5-o4-shRNA or 20 μL (1.1 × 10^10 GC/μL)
   AAV9-anti-Luc-Control-shRNA. Doxycycline (0.2% doxycycline with 4%
   sucrose in water) was administered from the day of tumor implantation
   until the mice were euthanized.

   For the knockdown of Claudin 2 and Claudin 10 in vivo,
   LV/Cas9-Cldn2-sgRNA1, 2 and LV/Cas9-Cldn10-sgRNA1, 2, 3 virus solutions
   (4 μL each, with a concentration ranging from 3 × 10^8 to 4 × 10^9
   IU/μL) were combined in 20 μL (for a final combined concentration of
   approximately 3 × 10^9 IU/μL in 20 μL) and injected into the renal
   cortex. LV/Cas9-control sgRNA (3 × 10^9 IU/μL in 20 μL) was used as
   control. After 26 days, the kidneys of some of the mice were harvested
   for analyses. Other mice were subsequently implanted orthotopically
   with 4T1 cells (5 × 10^5) as described above. Some of the mice were
   also subjected to UUO or sham surgery 10 days after orthotopic cancer
   cells inoculation. Although rare, mice were excluded from the
   experimental cohort if UUO was found to have failed (no fibrosis due to
   failed ligation of the ureter).

   To assess vascular permeability, mice were administered with
   FITC-conjugated 70 kDa dextran via retro-orbital venous sinus injection
   5 min prior to euthanasia.

   All the mice involved in the study were maintained under standard
   housing conditions at UT MD Anderson Cancer Center (MDACC) animal
   facilities: mice were housed in individually ventilated cages with a
   12h light/12 h dark cycle (06:00/18:00), a temperature range of at 21°C
   to 23°C and humidity levels between 40 to 60%. The procedures conducted
   on mice were reviewed and approved by Institutional Animal Care and Use
   Committee of MDACC.

Histology

   Formalin-fixed mouse lung, liver, kidney and other tissues (5 μm) were
   cut for different staining, including hematoxylin and eosin (H&E),
   Picrosirius red, and Masson Trichrome staining (MTS, using Gomori’s
   Trichrome Stain Kit (Leica Biosystems, 38016SS2) or Trichrome, Masson,
   Aniline Blue Stain Kit (Newcomer Supply, 9179B). Images of an entire
   lung lobe(s) or sagittal kidney sections were obtained using the
   Panoramic 250 Flash III slide scanner or ZEISS Axio Scan.Z1 AxioScan.
   Metastatic area is defined as the metastatic disease burden in the lung
   of a mouse with a primary tumor. Colonization area refers to the cancer
   disease burden in the lung of a mouse following a systemic injection of
   cancer cells. Lung metastasis or colonization were determined through
   histopathological analysis, and the metastatic area was quantified as a
   percentage of total lung area via using Panoramic Viewer software or
   Image J. One or two lung H&E sections per mouse was evaluated, as
   detailed in the source data. Liver metastases were identified using
   histopathological analysis from H&E staining and liver metastatic
   burden was quantified by counting the average number of nodules from
   randomly selected visual fields images captured by Leica DM 1000 LED
   microscope equipped with a DFC295 microscope camera (Leica) with LAS
   version 4.4 software (Leica). The kidney tumor burden was measured
   using histopathological analysis of H&E staining and the tumor area was
   quantified using Panoramic Viewer software and Image J software. The
   tumor area was measured as a percentage of the total kidney area.

   Pleural thickness was measured at 8 random points at the circumference
   of the lung lobe by H&E staining. Pulmonary edema was identified from
   H&E staining in both septal and alveolar area, characterized by
   noticeable interstitial and intra-alveolar fluid accumulation
   throughout the lungs as previously reported ^[362]57. Edema area was
   quantified by determining the percentage of total lung area affected
   using Panoramic Viewer software.

Immunofluorescence and Immunohistochemistry

   Kidney tissues were fixed in 4% paraformaldehyde at 4°C for overnight.
   After fixation, they were transferred to 30% sucrose solution for 24
   hours before being embedding in OCT compound. Deparaffinized and
   rehydrated FFPE sections (5 μm) followed antigen retrieval with heated
   citric acid buffer (pH 6.0) or TE buffer (pH 9.0) for 15 minutes at
   98°C, were blocked with blocking solution (1% BSA in PBS) for 1 hour
   before the incubation with different primary antibodies (as detailed in
   the Nature Research Reporting Summary linked to this article) overnight
   at 4°C. Samples were incubated with conjugated secondary antibodies (as
   detailed in the Nature Research Reporting Summary linked to this
   article), followed by DAPI (Vectashield, H1200) or Hoechst 33342
   (Invitrogen, H3570, 1:2,000) staining. Random visual fields were
   selected, and representative pictures were acquired with Zeiss Axio
   Observer.Z1 or LSM800 confocal laser scanning microscope using ZEN
   software (Zeiss). We evaluated the entire kidney tissue for GFP signal
   and analyzed kidneys from all the mice in the study and did not observe
   any GFP positive cells in kidneys. The percentage of Claudin 5 CD31
   double positive area out of total CD31 positive area were also
   quantified by ImageJ software. Randomly selected visual fields images
   were captured and the quantification was performed on 200x or 400x
   images, as listed in the figure legends.

   Vascular leakage assay was performed and quantified as previously
   described ^[363]25. Briefly, OCT-embedded tissues were stained with
   CD31 (Dianova, Dia310M, 1:40) before the incubation with Alexa Fluor
   647 anti-rat antibody (Invitrogen A21472, 1:1000). Representative
   images were captured via Zeiss Axio Observer.Z1 and quantified as the
   total FITC positive field normalized to CD31 field by Image J.

   For the visualization of Claudin 5 expression in lung endothelial
   cells, cells were cultured on gelatin-coated 8-well chambers (Thermo
   Scientific^™ 154534) for overnight. The cells were then starved in
   0.25% FBS for 24 hours prior to the treatment with rAng2 (100 ng/mL) or
   PBS (control) for 24 hours. Afterwards, cells were fixed with 100%
   methanol for 5 minutes and permeabilized with PBST (+0.5% Triton
   X-100/PBS) for an additional 5 minutes. Subsequently, the cells were
   incubated with blocking buffer (5% BSA in PBS+0.5% Triton X-100/PBS)
   for 1 hour at RT before overnight incubation with Claudin 5 antibody
   (Abcam 131259, 1:100) at 4°C. Following the primary antibody
   incubation, the cells were treated with Anti-Rabbit 488 secondary
   antibody (Jackson ImmnoResearch, 111546047, 1:200) before staining with
   Hoechst 33342 (Invitrogen, H3570, 1:2,000) for 10 minutes. The media
   chamber was carefully removed, and representative pictures were
   acquired with LSM800 confocal laser scanning microscope using ZEN
   software (Zeiss).

   For the immunohistochemistry staining of CD31, it was performed as
   described previously ^[364]41. FFPE mouse tumor tissues (5 μm) were
   processed before antigen retrieval with heated citric acid buffer (pH
   6.0) for 15 minutes at 98°C. After incubation with 3% H[2]O[2] in PBS
   for 15 minutes, the samples were blocked with 1% BSA in PBST before
   incubation with CD31 (Dianova, Dia310M, 1:40) antibody overnight at
   4°C. Then the samples were incubated with 4plus Biotinylated Goat
   Anti-Rat IgG (Biocare GR607H) for 30 minutes and a subsequent 30-minute
   incubation with 4plus Streptavidin HRP Label (Biocare HP604H) at room
   temperature. Subsequently, the slides were stained with DAB (Life
   Technologies 34065) and counterstained with hematoxylin before being
   cover-slipped. Randomly selected visual fields images (200X) imaged
   with the Leica DM 1000 LED microscope were captured and quantified with
   ImageJ (NIH, Bethesda, MD).

Microarray analysis

   Total RNA was extracted from healthy kidney (n=3 mice), fibrotic kidney
   (n=2 mice), healthy lungs (n=3 mice), fibrotic lungs (n=3 mice), and
   lungs from mice with fibrotic kidneys (n=3 mice) using TRIzol reagent
   (Life Technologies, 15596026) according to the manufacturer’s
   instructions. An equal amount of RNA was submitted from each cohort to
   the Advanced Technology Genomics Core at MD Anderson. The microarray
   data was deposited into Gene Expression Omnibus ([365]GSE199178). Gene
   expression analysis was performed using Mouse Ref6 Gene Expression
   BeadChip (Illumina), and the Limma package (version 3.44.3) from R
   (version 4.0.2) was used to identify differentially expressed genes for
   each comparison (p<0.05, abs (fold change) >=1.5). Pathway enrichment
   analysis was conducted by WebGestalt 2019
   ([366]http://www.webgestalt.org/) ^[367]58. Benjamini and Hochberg’s
   procedure was used for multiple test correction ^[368]59. Significant
   pathways were defined as those with an FDR< 0.05.

Quantitative real-time PCR analyses

   RNA from cells were extracted using TRIzol reagent (Life Technologies
   15596026) according to the manufacturer’s instructions. cDNA was
   synthetized using high-capacity cDNA reverse transcription kit (Applied
   Biosystems 4368814). Gene expression was determined using Power SYBR
   Green PCR Master Mix (Applied Biosystems) on QuantStudio^™ 7 Flex
   Real-Time PCR System (ThermoFisher Scientific). Primer sequences for
   the target genes are shown in [369]Supplementary Table 6. Gapdh, Actb
   or Cd31 were used to normalize the data. The relative mRNA expression
   is presented as fold change (2^-ΔΔCt), normalizing the control to 1, or
   presented as 2^-ΔCt. Statistical analyses were performed on the ΔCt
   values.

Western blot analysis

   NIH3T3 cells were transduced with Ad-CMV-CLDN5v2 (with His tag) or
   Adeno-CMV-luc for 48 hours. The cells were lysed with RIPA buffer
   (Thermo Scientific, 89900) with EDTA free protease inhibitor (Roche,
   4693159001) and the lysates were resolved on 4–12% Bis-Tris protein
   gels (Thermo Fisher Scientific, NP0321PK2) as previously
   described^[370]60. Control parental cells without any treatment were
   used. The proteins were transferred with a semidry trans-blot turbo
   transfer system (Bio-Rad, 1704150) into methanol activated PVDF
   membrane (Bio-Rad 1704272). The membranes were incubated with 5 %
   non-fat milk (LabScientific M0841) before incubation with different
   primary antibodies (as detailed in the Nature Research Reporting
   Summary linked to this article) overnight at 4°C. The membranes were
   washed and incubated with mouse (R&D system, HAF007, 1:2000) or Rabbit
   IgG (Abcam, ab16284, 1:2000) horseradish peroxidase-conjugated
   secondary antibody for 1 hour at room temperature. Amersham^™
   Hyperfilm^™ ECL (GE Healthcare, 28906835) and West-Q Pico ECL Solution
   (GenDEPOT, W3652020) were utilized for protein detection as described
   previously ^[371]41. The band density from different independent
   experiments were quantified using ImageJ as previously described
   ^[372]41, and statistical analyses were performed using difference in
   intensity between β-actin and Claudin 5 on three distinct experiments.
   For the evaluation of Claudin 5 protein levels in lung microvascular
   endothelial cells treated with rAng2, the cells were lysed with RIPA
   buffer and the lysates were resolved as described above. For the
   evaluation of Foxo1, the cells were lysed with subcellular protein
   fractionation kit for cultured cells (Thermo Scientific^™ 78840). The
   cytoplasmic and nuclear protein fractions were isolated based on
   manufacturer’s instructions and the lysates were separated on 4–12%
   Bis-Tris protein gels (Thermo Fisher Scientific, NP0321PK2) before
   transfer onto a PVDF membrane (Bio-Rad 1704272). The membranes were
   incubated with 5 % non-fat milk (LabScientific M0841) before incubation
   with different primary antibodies (as detailed in the Nature Research
   Reporting Summary linked to this article) overnight at 4°C. The
   membranes were washed and incubated with Rabbit IgG (Abcam, ab16284,
   1:2000) horseradish peroxidase-conjugated secondary antibody for 1 hour
   at room temperature and proteins were detected as described above.
   β-actin served as the internal control for data normalization. Relative
   protein expression is reported as fold change, with the control set to
   1. Statistical analyses were conducted on the ratio of target protein
   band intensity relative to β-actin band intensity. Uncropped blots are
   shown in the Image Source Data File.

Cytokine profiling and Ang2 ELISA

   Blood was collected at the time of euthanasia and centrifuged at 300g
   for 10 minutes. The serum was extracted and stored at −80°C for further
   use. Cytokine profiling was performed using the Mouse XL Cytokine Array
   Kit (R&D, ARY028) based on the manufacturer’s instructions. Cytokine
   array scans are shown in the Image Source Data File.Ang2 protein levels
   in serum samples were quantified by Angiopoietin 2 Mouse ELISA kit
   (Abcam, ab171335) according to the manufacturer’s protocol.

Flow cytometry

   Fresh kidneys were harvested and minced into small pieces prior to
   digestion with collagenase IV (Gibco, 17104019, 4 mg/mL) and dispase II
   (Gibco, 17105041, 4 mg/mL) in RPMI medium for 1 hour at 37°C. The
   mixture was then filtered through a cell strainer (70 μm) and the
   collected cells were resuspended in PBS/2% FBS. Afterwards, the cells
   were stained with anti-CD31-PE (BioLegend, 102508, 1:100) and
   anti-E-cadherin-eFluor^™ 660 (eBioscience, 50–3249-82, 1:100) in PBS/2%
   FBS on ice for 30 min. Samples were filtered through a 40 μm mesh and
   sorted with BD FACS AriaTM II sorter (South Campus Flow Cytometry Core
   Lab of MD Anderson Cancer Center). Singlets were then gated for CD31
   and E-cadherin positivity, with endothelial cells captured as CD31^+
   E-cadherin^- and epithelial cells captured as CD31^- E-cadherin^+.
   Cells were collected in RPMI medium with PBS/10% FBS before RNA
   isolation.

   For the evaluation of immune cells infiltration in the lung and kidney
   after induction of fibrosis, analysis was performed as previously
   reported ^[373]54. Lung fibrosis was induced via intratracheal
   administration of bleomycin, and kidney fibrosis was induced via UUO as
   described above. Lung and kidney were minced and digested in 10 mL
   Liberase TL (Roche 05401020001, 0.1 mg/mL) and DNase I (Roche,
   10104159001, 0.2 mg/mL) in RPMI-1640 medium for 30 minutes at 37°C. The
   tissue lysate was filtered through cell strainer (100 μm) and washed
   with PBS/2% FBS buffer prior to immunostaining. The resulting single
   cell suspension was incubated with different antibodies (as detailed in
   the Nature Research Reporting Summary linked to this article) on ice
   for 30 minutes in dark. For the staining of intracellular marker, cells
   were fixed and permeabilized using the Foxp3/Transcription Factor
   Staining Buffer Set (eBioscience 00552300) according to manufacturer’s
   instructions before staining with anti-Foxp3-Alexa Fluor^™700
   (eBioscience 56577382, 1:50) in permeabilization buffer. Data were
   acquired on BD LSR Fortessa X-20 Cell Analyzer (BD Biosciences) and the
   CD45^+ cells were analyzed for the expression of the different immune
   populations using FlowJo.

Detection of circulating cancer cells

   To detect circulating tumor cells in lung and kidney, 5 × 10^5 4T1
   cells were injected intravenously (4T1 i.v.) into the retro-orbital
   venous sinus of BALB/c virgin female mice. After 10 days, mice were
   euthanized with CO[2] and a perfusion with 150 mL of PBS was conducted
   via insertion into the left ventricle of the heart. Subsequently, fresh
   lung and kidney tissues were collected for flow cytometry analysis and
   genomic DNA isolation. For flow cytometry, fresh lung and kidneys were
   minced into small pieces prior to digestion with collagenase IV (Gibco,
   17104019, 4 mg/mL) and dispase II (Gibco, 17105041, 4 mg/mL) in
   RPMI-1640 medium at 37°C for 1h, followed by treatment with DNase I
   (Roche 10104159001, 0.2 mg/mL). After the filtration of the cells via
   cell strainer (70 μm), the cells were resuspended in PBS/2% FBS and
   10mM EDTA before the filtration through a 40 μm mesh before the
   centrifugation at 500g for 5 minutes. Cells were then incubated with 3
   mL ACK lysis buffer (Gibco A1049201) at RT for red cell lysis and wash
   with PBS/2% FBS and 10mM EDTA for twice before centrifugation at 500g
   for 5 minutes. Then the cells were stained with eFluor^™ 780 live/dead
   fixable viability dye (eBioscience 65086514, 1:1000) on ice for 10
   minutes before the resuspension with PBS/2% FBS in PBS. Flow cytometry
   analysis for GFP^+ cancer cells was performed using BD LSR Fortessa
   X-20 Cell Analyzer (BD Biosciences). The data were recorded as the
   percentage of GFP^+ cells per 10^6 total cells. For genomic DNA
   isolation, DNA was isolated from lung and kidney tissues using the
   DNeasy Blood & Tissue Kits (Qiagen, 69506), 4T1-luc-GFP cells derived
   DNA was served as the positive control, and fresh lung and kidney
   tissues from WT mice were served as the negative control. For Q-PCR
   analysis of GFP, the DNA was amplified using Power SYBR Green PCR
   Master Mix (Applied Biosystems) on QuantStudioTM 7 Flex Real-Time PCR
   System (ThermoFisher Scientific). Neurofibromatosis type 1 (Nf1) was
   used to normalize the data and presented as 2^-ΔCt.

Single cell RNA sequencing analysis

   The raw counts and metadata of scRNA-seq of murine endothelial cells
   ^[374]34, murine lung tissue ^[375]35, human lung tissue ^[376]35,
   murine kidney tissue ^[377]36, and human kidney tissue ^[378]37, murine
   brain tissue ^[379]34 and human brain tissue ^[380]61 were obtained
   from the database as described in the original studies. The data was
   processed using R (version 4.0.0) package Seurat (version 4.0.3).
   Principal component analysis (PCA) was applied to perform
   dimensionality reduction on the expression matrices. The
   ‘FindNeighbors’ function identified nearest neighbors among cells
   within the PCA space, while the ‘FindClusters’ function grouped cells
   using the Louvain algorithm. Visualization of the clusters resulting
   from t-Distributed Stochastic Neighbor Embedding (tSNE) dimension
   reduction was achieved through the ‘RunTSNE’ function. The cell type
   annotations were adopted from the original studies to label the
   clusters for visualization purpose. ‘DimPlot’ function was used to show
   the tSNE plots of cell clusters. ‘VlnPlot’ function was used for
   displaying the expression of genes in different tissues. ‘FeaturePlot’
   was employed to display the expression of genes in cell clusters.

RNA sequencing analysis

   The raw reads from bulk RNA sequencing, capturing transcriptome data
   from breast cancer samples with lung metastasis and corresponding
   normal lung tissues, and were obtained from the database as described
   in the original studies ([381]GSE193103)^[382]62. The data was
   processed using R software (version 4.2.2) along with the DESeq2
   package (version 1.38.3) for data normalization and differential gene
   expression analysis.

Statistics & Reproducibility

   No statistical methods were used to pre-determine sample sizes but our
   sample sizes are similar to those reported in previous
   publications^[383]13,[384]16,[385]25,[386]54,[387]55. Mice and samples
   were randomized prior to group allocations whenever applicable. No data
   were excluded from the analyses with the following rare exceptions: 1)
   mice in which UUO was found to have failed and 2) sample processing for
   flow cytometry failed quality control assessment. Data collection and
   analysis were not performed blind to the conditions of the experiments.
   Data are represented as mean ± SEM and statistical analysis was
   conducted using GraphPad Software. Detailed information on statistical
   tests used and definition of P values are listed in the Figure Legends.

Extended Data

Extended Data Fig. 1: Lung and kidney fibrosis induction.

   Extended Data Fig. 1:
   [388]Open in a new tab

   (a) Representative H&E staining, Sirius Red staining, and αSMA
   immunolabeling of the lung and kidneys of mice with and without
   fibrosis in the respective organs. Scale bar, 100 μm. (b)
   Quantification of positive Sirius Red and αSMA^+ stained area in the
   indicated organs and groups, n = 5 to 6 mice per group. (c, d) Relative
   Fn1 and Col1a1 mRNA levels in the respective organs and groups, n = 3
   mice per group. KF: kidney fibrosis; LF: lung fibrosis. (e)
   Representative images for GFP immunolabeling and quantification of
   percent of the GFP positive area in lung with or without lung fibrosis.
   Scale bar, 50 μm, n = 6 mice per group. (f) Representative images for
   GFP immunolabeling of tumor in the mice with or without lung fibrosis.
   n = 1 mouse per group. Scale bar, 25 μm. (g) Representative liver H&E
   images in 4T1 tumor orthotopic bearing mice after induction of lung
   fibrosis. Scale bar, 50 μm. Quantification of metastatic liver nodules
   in the indicated groups, n = 6 mice per group. (h) Representative
   images for GFP immunolabeling and quantification of percent of the GFP
   positive area in liver of mice with or without lung fibrosis. Scale
   bar, 50 μm, n = 6 mice per group. (i, j) Representative H&E images of
   heart (i) and intestine (j) in 4T1 tumor orthotopic bearing mice after
   induction of lung fibrosis. Scale bar, 50 μm. (k) Orthotopic 4T1 tumor
   volume over time in mice with or without kidney fibrosis, n = 6 mice
   per group. (l, m) Representative H&E images (l) of the lung of 4T1
   orthotopic tumor bearing mice with or without kidney fibrosis and
   respective quantification (m) of metastatic area. Scale bar, 25 μm, n =
   6 mice per group. (n) Representative images for GFP immunolabeling and
   quantification of percent of the GFP positive area in lung with or
   without kidney fibrosis. Scale bar, 50 μm, n = 6 mice per group. (o)
   Representative images for GFP immunolabeling for tumor in the mice with
   or without kidney fibrosis. Scale bar, 25 μm. (p) Representative liver
   H&E images in 4T1 tumor orthotopic bearing mice after induction of
   kidney fibrosis. Scale bar, 50 μm. Quantification of metastatic liver
   nodules in the indicated groups, n = 6 mice per group. (q)
   Representative images for GFP immunolabeling and quantification of
   percent of the GFP positive area in liver of mice with or without
   kidney fibrosis. Scale bar, 50 μm, n = 6 mice per group. (r)
   Representative H&E images of heart and intestine of 4T1 tumor
   orthotopic bearing mice after induction of kidney fibrosis. Scale bar,
   50 μm. KF: kidney fibrosis; LF: lung fibrosis. Data are presented as
   mean values +/− SEM. For k: two-way ANOVA with Sidak’s multiple
   comparisons test. For all other panels: Unpaired two-tailed t test with
   Welch’s correction applied for unequal variances (determined by
   F-test). P values are listed, ns: not significant.

Extended Data Fig. 2. Fibrosis does not promote non-tropic metastasis.

   Extended Data Fig. 2
   [389]Open in a new tab

   (a) Representative H&E images of the kidneys of 4T1 orthotopic tumor
   bearing mice +/− lung or kidney fibrosis. Scale bar, 25 μm. (b) Number
   of mice with microscopic kidney tumors in the indicated groups (c)
   Representative images for GFP immunolabeling and quantification of
   percent of the GFP positive area in kidney +/− kidney fibrosis. Scale
   bar, 50 μm. (d) Number of mice with microscopic kidney tumors in the
   indicated groups. (e) Representative kidney H&E images in MMTV-PyMT
   transgenic mouse model with lung or kidney fibrosis. Scale bar, 25 μm.
   (f) Number of mice with microscopic kidney tumors in the indicated
   groups. (g) Representative immunofluorescence images of PyMT positive
   cancer area in tumor, lung, and kidney after kidney fibrosis in
   MMTV-PyMT transgenic mice. Scale bar, 25 μm. (h–j) Representative
   images and quantification of positive MTS and Sirius Red staining from
   the indicated groups. Scale bar, 25 μm, n = 6–7 mice per group. Data
   are presented as mean values +/− SEM. Unpaired two-tailed t test, with
   Welch’s correction applied for unequal variances (determined by
   F-test). P values are listed, ns: not significant.

Extended Data Fig. 3. Intracardiac cancer cell injection does not alter
metastatic disease in renal fibrosis & Wound healing does not impact tropic
and non-tropic site metastasis.

   Extended Data Fig. 3
   [390]Open in a new tab

   (a) Representative GFP and CD31 staining of the kidneys from mice with
   4T1 orthotopic inoculation and 4T1 i.v. administration with or without
   kidney fibrosis. Scale bar, 25 μm. (b) Quantification of GFP^+ cancer
   cells from lung after 4T1 i.v. administration for 10 days with or
   without perfusion by flow cytometry. n = 3 mice per group. Gating
   strategy shown in [391]Supplementary Fig. 1A. (c) Quantification of
   GFP^+ cancer cells from kidneys after 4T1 i.v. administration for 10
   days with or without perfusion by flow cytometry, n = 3 mice per group.
   Gating strategy shown in [392]Supplementary Fig. 1B. (d, e) QPCR
   quantification of GFP DNA (expressed 2^-ΔCt after normalization to Nf1)
   from the lung (d) and kidney (e) of mice after 4T1 i.v. administration
   for 10 days with or without perfusion, n = 3 mice per group. (f)
   Representative H&E images and quantification of lung after intracardiac
   injection of 4T1 cancer cells together with induction of kidney
   fibrosis. Control, n = 5 mice; kidney fibrosis, n = 9 mice, Scale bar,
   25 μm. (g) Representative H&E images and quantification of metastatic
   liver nodules after intracardiac injection of 4T1 cancer cells together
   with induction of kidney fibrosis. Control, n = 5 mice; kidney
   fibrosis, n = 9 mice. Scale bar, 50 μm. (h) Representative H&E images
   and number of mice with microscopic kidney tumors in the indicated
   groups after Intracardiac injection of 4T1 cancer cells together with
   induction of kidney fibrosis. Scale bar, 25 μm. (i) Representative
   images for GFP immunolabeling and number of mice with microscopic
   kidney tumors in the indicated groups after Intracardiac injection of
   4T1 cancer cells together with induction of kidney fibrosis. Scale bar,
   50 μm. (j) Orthotopic 4T1 tumor volume over time +/− cutaneous wound, n
   = 6 mice per group. (k) Representative H&E images of the lung of 4T1
   orthotopic tumor bearing mice +/− cutaneous wound. Scale bar, 25 μm.
   (l) Quantification of metastatic area and surface lung nodules in 4T1
   orthotopic tumor bearing mice with and without cutaneous wounds, n = 6
   mice per group. (m) Tumor volume over time in MMTV-PyMT mice +/−
   cutaneous wound, n = 5 to 6 mice per group. (n) Representative lung H&E
   images from in MMTV-PyMT mice +/− cutaneous wound. Scale bar, 25 μm.
   (o) Quantification of metastatic area and surface lung nodules in
   MMTV-PyMT mice +/− cutaneous, n = 5 to 6 mice per group. (p) Wound
   closure rate in 4T1 orthotopic tumor bearing mice (n = 10 mice) and no
   tumor control (n = 4 mice). (q) Representative lung H&E images from
   mice with 4T1 i.v. injection and +/− cutaneous wound. Scale bar, 25 μm.
   (r) Quantification of metastatic area and surface lung nodules in mice
   with 4T1 i.v. injection +/− cutaneous wound, n = 5 mice per group. Data
   are presented as mean values +/− SEM or individual values shown. For j,
   m: two-way ANOVA with Sidak’s multiple comparisons test. For p: mixed
   effect model with Sidak’s multiple comparison test. For other panels:
   Unpaired two-tailed t test, with Welch’s correction applied for unequal
   variances (determined by F-test). P values are listed, ns: not
   significant.

Extended Data Fig. 4. Global gene expression profiling in fibrosis and serum
cytokine analysis.

   [393]Extended Data Fig. 4
   [394]Open in a new tab

   (a, b). Volcano plot and gene ontology (GO) enrichment ratio and false
   discovery rate (FDR) values associated with overlapped up-regulated and
   down-regulated genes. (c, d) GO enrichment ratio and FDR values in the
   lungs of mice with and without lung fibrosis (c) or with and without
   kidney fibrosis (d). (e, f) Overlapping up-regulated (e) or down
   regulated (f) genes in the lungs of healthy mice vs. mice with the lung
   fibrosis, and in the lungs of healthy mice vs mice with kidney
   fibrosis, with overlapped genes involved in vascular (f) and immune
   remodeling (e) pathways. (g) Cytokines profile of serum samples from
   healthy mice, mice with cutaneous wound, mice with lung fibrosis and
   kidney fibrosis, and heat map of the averaged relative cytokine levels
   in the indicated groups after normalization to healthy group (set to
   1). Healthy, n = 2 mice; wound, n = 2 mice; lung fibrosis (LF), n = 3
   mice; kidney fibrosis (KF), n = 3 mice. Cytokines showing minimal
   changes between healthy and wound (wound/healthy (W/H) ratio between
   0.9 and 1.1) were then ranked based on levels in kidney fibrosis.

Extended Data Fig. 5. Immune cells infiltration in lung and kidney after
fibrosis induction.

   [395]Extended Data Fig. 5
   [396]Open in a new tab

   (a) Percentages of CD45^+, CD3^+, CD4^+FoxP3^− effector T (Teff) cells,
   CD8^+, γδ^+, CD4^+FoxP3^+ regulatory T (Treg) cells, NK1.1^+ cells,
   CD11c^+, CD19^+, CD11b^+, CD11b^+Ly6C^+Ly6G^−and CD11b^+Ly6G^+ Ly6C^−
   cells in lungs of mice with lung fibrosis or kidney fibrosis. Healthy
   lung, n = 6 mice; lung fibrosis (LF), n = 5 mice; healthy kidney, n = 6
   mice; kidney fibrosis (KF), n = 5 mice. (b) Percentages of CD45^+,
   CD3^+, CD4^+FoxP3^− effector T (Teff) cells, CD8^+, γδ^+, CD4^+FoxP3^+
   regulatory T (Treg) cells, NK1.1^+ cells, CD11c^+, CD19^+, CD11b^+,
   CD11b^+Ly6C^+Ly6G^−and CD11b^+Ly6G^+Ly6C^− cells in kidneys of mice
   with lung fibrosis or kidney fibrosis. Healthy lung, n = 5 mice; lung
   fibrosis (LF), n = 6 mice; healthy kidney, n = 6 mice; kidney fibrosis
   (KF), n = 4 mice. (c) Colocalization of FITC-dextran (green) and CD31
   (red) in the liver of mice with and without after rAng2 treatment.
   Scale bar, 20 μm. Data are presented as mean values +/− SEM. Unpaired
   two-tailed t test, with Welch’s correction applied for unequal
   variances (determined by F-test). P values are listed, ns: not
   significant.

Extended Data Fig. 6. Recombinant Ang 2 enhances metastasis to the lung but
not kidney.

   Extended Data Fig. 6
   [397]Open in a new tab

   (a) Representative lung H&E of mice treated with PBS (control) or
   recombinant Ang2 (rAng2). Scale bar, 25 μm. (b, c) Representative lung
   H&E of mice after induction of lung (b) or kidney fibrosis (c). Scale
   bar, 25 μm. (d) Quantification of edema area per whole lung area, n = 5
   to 6 mice per group. (e) Quantification of pleural thickening of lung,
   n = 5 to 6 mice per group. (f) Representative kidney H&E images of mice
   treated with PBS (control) or rAng2. Scale bar, 25 μm. (g, h)
   Representative kidney H&E images of mice after induction of lung (g) or
   kidney fibrosis (h). Scale bar, 25 μm. (i) Representative lung H&E of
   mice treated with PBS (control) or recombinant Ang2 (rAng2) with 4T1
   cells i.v. injection. Scale bar, 25 μm. (j) Representative kidney H&E
   of mice treated with PBS (control) or recombinant Ang2 (rAng2) with 4T1
   cells i.v. injection. Scale bar, 25 μm. Data are presented as mean
   values +/− SEM. Unpaired two-tailed t test, with Welch’s correction
   applied for unequal variances (determined by F-test). P values are
   listed.

Extended Data Fig. 7. tSNE plots of Claudin genes in scRNA-seq data sets.

   [398]Extended Data Fig. 7
   [399]Open in a new tab

   (a) tSNE plots of different Cldn genes in flow cytometry sorted CD31^+
   endothelial cells from mouse lung tissue[400]34. (b) tSNE plots of
   different Cldn genes in endothelial cells from mouse lung
   tissue[401]35. (c) tSNE plots of Claudin family genes in the
   endothelial cells from human lung scRNA-seq data[402]35. (d) tSNE plots
   of Claudin family genes in the endothelial cells from mouse kidney
   scRNA-seq data[403]36. (e) tSNE plots of Claudin family genes in the
   endothelial cells from human kidney scRNA-seq data[404]37. Colors coded
   for the relative gene expressions.

Extended Data Fig. 8. Claudins expression in distinct cell types.

   [405]Extended Data Fig. 8
   [406]Open in a new tab

   (a) Different Claudin expressions in different endothelial cell types
   of mouse lung[407]1,[408]34. (b, c) tSNE plots of Claudin 5 in
   different cell types of in mouse lung (b) and highly expressed in lung
   endothelial cells (c)[409]35. (d) Different Claudins expression in
   different mouse lung cell types[410]35. (e) Expression of diverse
   Claudins in different mouse kidney cell types[411]36. (f, g) Diverse
   Claudins expression in different mouse[412]34 (f) and human brain cell
   types (g)[413]61. (h) Colocalization of Claudin 2 and CD31 in WT kidney
   cortex and medulla (individual images in [414]Supplementary Fig. 5a,
   [415]b). (i) Colocalization of Claudin 10 and CD31 in WT kidney cortex
   and medulla (individual images in [416]Supplementary Fig. 5c, [417]d).
   Scale bar, 50 μm (upper), 25 μm (lower). (j) Representative
   colocalization of Claudin 5 and CD31 in the lung of mice in the
   indicated groups. Scale bar, 20 μm. (k) Quantification of percentage of
   Claudin 5 and CD31 double positive area per visual field in the
   indicated groups, n = 5–9 mice per group. These images and graphs also
   shown in [418]Supplementary Fig. 6. Data are presented as mean values
   +/− SEM. Unpaired two-tailed t test, with Welch’s correction applied
   for unequal variances (determined by F-test). P values are listed, ns:
   not significant.

Extended Data Fig. 9. Data Figure 9. Ang2 and Claudins signaling and
Claudin2/10 suppression.

   [419]Extended Data Fig. 9
   [420]Open in a new tab

   (a) Relative mRNA expression of Cldn5 in parental NIH-3T3 cells and
   NIH-3T3 cells treated with Adeno-CMV-luc (Ad-Ctrl) or Ad-CMV-CLDN5v2
   (Ad-Claudin 5), n = 3 biologically independent experiments (with each 2
   technical replicates). (b) Representative western blot and
   quantification of his-tagged Claudin 5 (His), Claudin 5, and β-actin in
   parental NIH-3T3 cells and NIH-3T3 cells treated with Adeno-CMV-luc
   (Ad-Ctrl) or Ad-CMV-CLDN5v2 (Ad-Claudin 5), n = 3 biologically
   independent experiments. Uncropped blots are shown in the Image Source
   Data File. (c) Relative mRNA expression of Cldn5 in NIH-3T3 cells with
   the indicated treatment. Dox on/off indicates doxycycline exposure, n =
   4 biologically independent experiments (with each 2 technical
   replicates). (d) Orthotopic 4T1 tumor volume over time in the indicated
   groups, n = 5 mice per group. (e) CLDN5 mRNA levels in metastatic lungs
   of patients with breast cancer (n = 9) and healthy lungs (n = 6)
   ([421]GSE193103)62. (f, g) Relative mRNA expression of Cldn2 (f) and
   Cldn10 (g) in kidney endothelial cells treated with the indicated guide
   RNAs, n = 3 biologically independent experiments (with each 2 technical
   replicates). (h) Relative mRNA expression of Cldn2 and Cldn10 in sorted
   E-cadherin^+ epithelial cells and CD31^+ endothelial cells, n = 3 mice
   per group. Gating strategy is shown in [422]Supplementary Fig. 9. (i,
   j) Relative mRNA expression levels of Cldn2 and Cldn10 in sorted CD31^+
   endothelial cells (i) and E-cadherin^+ epithelial cells (j) after
   treatment with the indicated guide RNAs, n = 3 mice per group. (k)
   Experimental design for the evaluation of the downregulation of Claudin
   2 and Claudin 10 in the kidney of mice with 4T1 orthotopic tumors. (l)
   Orthotopic 4T1 tumor volume over time in the indicated groups. On day
   14, LentiCRISPRv2 control vector, n = 5; LentiCRISPRv2 Cldn2 and
   Cldn10, n = 5; Day 18, LentiCRISPRv2 control vector, n = 5;
   LentiCRISPRv2 Cldn2 and Cldn10, n = 6; Day 22, LentiCRISPRv2 control
   vector, n = 8; LentiCRISPRv2 Cldn2 and Cldn10, n = 6 mice. (m)
   Representative H&E images of the lung of mice in the indicated groups.
   Scale bar, 25 μm. (n) Quantification of lung metastatic area.
   LentiCRISPRv2 control vector, n = 15; LentiCRISPRv2 Cldn2 and Cldn10, n
   = 20 mice. (o) Representative images for GFP immunolabeling and
   quantification of percent of the GFP positive area in lung. Scale bar,
   50 μm, n = 9 mice per group. Data are presented as mean values +/− SEM,
   Min-to-Max in E. For a: one-way ANOVA test with Tukey’s multiple
   comparisons test. For d: two-way ANOVA with Sidak’s multiple
   comparisons test. For b, c, f, g: one-way ANOVA test with Dunnett’s
   multiple comparisons test. For e: two-tailed Mann-Whiney test. For i,
   j, n, o: Unpaired two-tailed t test, with Welch’s correction applied
   for unequal variances (determined by F-test).

Extended Data Fig. 10. Knockdown of Claudin 2/10 redirects metastasis to the
kidney.

   [423]Extended Data Fig. 10
   [424]Open in a new tab

   (a) Representative H&E images of the kidney of mice in the indicated
   groups. Scale bar, 50 μm (upper), 25 μm (lower). (b) Percentage of mice
   with kidney lesion in the indicated groups. LentiCRISPRv2 control
   vector, n = 15; LentiCRISPRv2 Cldn2 and Cldn10, n = 20 mice. (c)
   Representative images for GFP immunolabeling in kidney. Scale bar, 50
   μm (upper), 25 μm (lower). (d) Number of mice with GFP positive kidney.
   (e) Quantification of percent of the GFP positive area in the indicated
   groups, n = 12 mice per group. (f) Orthotopic 4T1 tumor volume over
   time in the indicated groups. LentiCRISPRv2 control vector without
   fibrosis, n = 14; LentiCRISPRv2 control vector with fibrosis, n = 11;
   LentiCRISPRv2 Cldn2 and Cldn10 without fibrosis, n = 12 mice;
   LentiCRISPRv2 Cldn2 and Cldn10 with fibrosis, n = 10 mice. (g) Number
   of mice with GFP positive kidney in the indicated groups. LentiCRISPRv2
   control vector without fibrosis, n = 14; LentiCRISPRv2 control vector
   with fibrosis, n = 11; LentiCRISPRv2 Cldn2 and Cldn10 without fibrosis,
   n = 12 mice; LentiCRISPRv2 Cldn2 and Cldn10 with fibrosis, n = 10 mice.
   (h) Representative GFP and CD31 staining of the kidneys in mice with
   Claudin 2 and Claudin 10 down regulation together with or without
   kidney fibrosis. (i) Representative GFP and CD31 staining in the
   metastatic kidney with Claudin 2 and Claudin 10 downregulation. For
   i-j, individual images are shown in [425]Supplementary Fig. 10A-[426]C.
   Scale bar, 25 μm. (j) Representative images for CD31 and quantification
   of percent of the CD31 positive area in tumor when downregulate Claudin
   2/10 in the kidney. Scale bar, 50 μm. LentiCRISPRv2 control vector, n =
   5; LentiCRISPRv2 Cldn2 and Cldn10, n = 7 mice. (k) Representative
   images for CD31 and quantification of percent of the CD31 positive area
   in tumor when downregulate Claudin 2 and Claudin 10 with or without
   kidney fibrosis. Scale bar, 50 μm. LentiCRISPRv2 control vector without
   fibrosis, n = 5; LentiCRISPRv2 control vector with fibrosis, n = 4;
   LentiCRISPRv2 Cldn2 and Cldn10 without fibrosis, n = 4 mice;
   LentiCRISPRv2 Cldn2 and Cldn10 with fibrosis, n = 6 mice. Data are
   presented as mean values +/− SEM. For b: two-sided Chi-square. For f:
   two-way ANOVA with Sidak’s multiple comparisons test. For j: Unpaired
   two-tailed t test, with Welch’s correction applied for unequal
   variances (determined by F-test). For k: one-way ANOVA with Tukey’s
   multiple comparisons test. P values are listed, ns: not significant.

Supplementary Material

   Supplementary Table
   [427]NIHMS2052537-supplement-Supplementary_Table.docx^ (1.8MB, docx)
   Supplementary Figures
   [428]NIHMS2052537-supplement-Supplementary_Figures.pdf^ (2.7MB, pdf)

Acknowledgments