Dynamic Patterns of Clonal Evolution in Tumor Vasculature Underlie Alterations in Lymphocyte–Endothelial Recognition to Foster Tumor Immune Escape

Clonal selection of somatic and germline stem cells, through competitive interactions, is a common feature of tissue development (1–6). Although genomic analysis of the clonal origins of tumor cells has begun to map the evolutionary relationships responsible for cancer clone maintenance and propagation, little is known regarding the stepwise progression of the tumor microenvironment. Recent evidence suggests the tumor microenvironment, in particular the vascular compartment, promotes tumor maintenance and may have a role in promoting therapeutic resistance. In a variety of tumors, subsets of highly tumorigenic cells have been found to exist in close proximity to tumor endothelium (7, 8). This interaction has been shown to be reciprocal; tumors can remodel tissue microenvironments and specialized niches to their competitive advantages. Here, we investigate the stepwise development of the tumor microenvironment via genetic fate mapping, lineage tracing, and gene expression profiling. We demonstrate that the clonal architecture within developing tumor microenvironments undergoes dynamic changes, with a bias toward mono-to-oligoclonality as tumors progress. Indeed, the loss of founder clones (F₀) associated with tumor disease progression is followed by the acquisition of genetic traits in surviving subclones (F₁) characterized by a loss in expression of functionally significant primary lymphocyte adhesion and chemokine molecules and upregulation of cell-autonomous growth signaling and alternative angiogenic programs. These findings are in contrast to the clonal architecture and gene expression programs of homeostatic endothelium, which exists in a quiescent, polyclonal state. We therefore describe a developmental switch in the stepwise progression of tumor vasculature from a restricted set of progenitors to clonally selected subclones. The consequence of these endothelial changes we propose will have both a quantitative and functional impact on lymphocyte distribution that may well influence tumor immune function and underlies pharmacologic escape mechanisms from current antiangiogenic therapies.

Mice were bred and maintained at Stanford University Research Animal Facility (Stanford, CA) in accordance with Stanford University guidelines. Rosa-26 Rainbow mice were generated as previously described (2). ActinCreERT2 transgenic and VE-cadherin-CreERT2 were obtained from The Jackson Laboratory and crossed to homozygous Rosa-26 Rainbow. Briefly, for Cre-induction, 10-week-old mice were injected with 4 to 6 mg of tamoxifen intraperitoneally every other day for five days (three injections).

The B16-F1 melanoma cell line was obtained from ATCC in July 2012 and maintained in DMEM supplemented with 2 mmol/L l-glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, and 10% FBS (PromoCell). To induce tumor angiogenesis, the mice were injected subcutaneously with B16 cells (5 × 10⁴ cells in 20 μL) five days following tamoxifen administration, and tumors were allowed to grow for 7 to 40 days.

Tumor tissues were mechanically dissociated in Medium 199 containing Liberase TM and TH enzymes (Roche), DNase (Worthington), and Pluronic-F68 (Sigma) at 37°C until single-cell suspension was achieved (2–6 hours). Cells were then washed twice with PBS and filtered through a 70-μm filter. Cell suspensions were stained with phycoerythrin (PE/Cy7)-conjugated anti-CD31 (eBioscience; 25-0311), Biotin-conjugated anti-CD144 (BioLegend; 328120), and anti-CD45 microbeads (Miltenyi Biotec) with simultaneous detection of GFP, CFP, YFP, or PE-Texas Red. Flow cytometry analysis and cell sorting was performed on a BD FACSAria (Becton Dickinson) cell sorting system under 20 psi with a 100-μm nozzle. In analyses, dead cells and cell debris were excluded by gating the population according to LIVE/DEAD Aqua Fixable Dead Cell Stain (Invitrogen) and using forward and side light scatters. The number of positive cells was compared with the number of cells positive in the staining with the IgG isotype controls (BD PharMingen) and determined with FACSAria Flow Cytometer (Becton Dickinson).

Tissues were fixed in 4% paraformaldehyde at 4°C. Samples were prepared for embedding by soaking in 30% sucrose in PBS and then quick-frozen in optimum cutting temperature (OCT) compound. Frozen sections (5–8 μm) were then blocked for 30 minutes with 10% BSA and 2% goat serum followed by incubation with primary antibody for 12 to 16 hours. For immunostaining on sections from multicolored mice, Alexa 647 or 750–conjugated antibody (1:1,000) was used as a secondary stain for 1 hour (Invitrogen) and was visualized in the far-red channel (Cy5 or Cy7). Fluorescent images were taken with a Leica DM400B microscope (Leica Microsystems).

To quantitatively assess clone size, defined by single color segments of blood vessels uninterrupted by any single cell of another color, we counted the number of contiguous colored cells with nuclei in each clone per section. The peak and mean clone sizes from serial sections were quantified.

Endothelial cells were FACS-sorted and total RNA was isolated by TRIzol extraction, labeled, and hybridized to Affymetrix microarrays (Mu430v2) as previously described. Each common reference sample was normalized to publicly available Affymetrix mouse arrays (∼20,000), and for each probeset, the dynamic range was calculated as the difference between the lowest and the highest expression values and normalized as described in ref. 26. Microarray data may be viewed under the Gene Expression Omnibus accession number GSE38461.

All data are presented as mean ± SE. For two group comparisons, two-tailed Student t test was used. χ2 tests were employed to examine the compositional bias of countable clones in independent samples, and a P value less than 0.05 was considered significant.

To investigate the clonal origin and growth patterns of tumor blood vessels, we utilized a multicolor Cre-dependent marker system (red, yellow, blue, and green) similar to Brainbow (9, 10). In this model, Cre-mediated recombination, induced by tamoxifen, causes cells (and all of their progeny) to express one of four randomly assigned fluorescent proteins. We first analyzed clonal patterns under steady-state conditions. Quantitative lineage tracing analysis 180 and 270 days after tamoxifen administration revealed minimal cellular proliferation in the steady state throughout small and medium arterial conduits, with median endothelial clone sizes of 1.2 and 1.4 (Supplementary Fig. S1), respectively. These findings are consistent with previous clonal tracing studies demonstrating mature vasculature to be relatively quiescent (11–14). Areas of clonal proliferations were observed mostly at sites of vascular bifurcations in medium-sized arteries, with clone sizes ranging from 8 to 15 endothelial cell divisions (Supplementary Fig. S1). These sites were characterized by single-color segments uninterrupted by any single cell of another color, ruling out migration of other cells into the clonal region. The anatomic distribution of these sites suggests these expansions likely represent vascular remodeling induced by local hemodynamic changes.

We next used a lineage tracing strategy to investigate the spectrum of blood vessel clonality in response to pathologic stimuli. The tumor microenvironment is composed of numerous signaling molecules that induce changes in endothelial behavior to promote tumor growth and metastasis. To more completely understand the consequences of these angiogenic programs on the formation and maintenance of tumor vasculature in vivo, we engrafted tamoxifen-pulsed VE-cadherin-CreERT2 Rainbow mice with syngeneic B16 melanoma. Melanomas are known to acquire a rich vascular network through the release of diffusible activators of angiogenesis (15). Four weeks after engraftment of tumor, blood vessels were abundant within the tumor microenvironment. Clonal analysis of infiltrating tumor stromal elements (hematopoietic, fibroblast, and nerve) revealed a heterogeneous patterning with multiple colored clones equally represented (Fig. 1A, D, and E). In marked contrast, the clonal composition of infiltrating tumor vasculature was mono-to-oligoclonal, with large fields of single-colored capillaries observed on cross-sectional imaging (Fig. 1B and C). Quantitative analysis demonstrated a mean clone size of 15.1 (3.6) compared with 1.2 (0.99) in adjacent structures (Table 1).

To further characterize and quantify the contribution of individual clones, we next utilized a Lin⁻, FSC intermediate, CD144⁺ CD31⁺ gating scheme on FACS to isolate single cells from suspensions of enzymatically digested tumor (Fig. 2). Using this approach, we could detect and quantify Cre-induced fluorescent labeling specifically within the tumor vasculature across an entire engrafted tumor sample. Quantitative analysis by FACS demonstrated the distribution of clonality in the early period following engraftment was polyclonal, with relatively equal frequencies of fluorescent proteins (Fig. 1D and E, Fig. 2A and D, Table 1). Over extended periods, however, a subset of endothelial cells increased in size over time, giving rise to large clones that contribute substantially to the overall tumor vasculature (Fig. 2B and C). A 40-day chase revealed a mean clone size of 40.2 (53.9), with peak clone sizes varying widely up to 157.3, indicating large fields of single-colored capillary networks (Table 1). At these later time points, 79.7% (±16) of all CD45⁻CD31⁺CD144⁺ endothelial cells were a single color, consistent with ever fewer but larger clones dominating the microenvironmental landscape (Fig. 2D).

The above experiments suggested the clonal composition within evolving tumor microenvironments undergoes dynamic changes in response to angiogenic cues. These findings support a model where F₀ diminish in frequency and are replaced by F₁ as tumors evolve. To explore this possibility, we isolated endothelial subsets, distinguished functionally by the magnitude of their contribution to the microenvironment, for gene expression profiling. Microarray analysis of FACS-isolated Lin− FSC low, CD144⁺ CD31⁺ F₀ and F₁ endothelial cells from adult mice demonstrated that F₀ and F₁ endothelial cells shared a high degree of transcriptome-wide similarity (R² = 0.92) and similar expression of endothelial-related genes, such as CD144, Flt1, KDR, VWF, ITGB3, and the ETS transcription factor ETV6 (Fig. 3A and B). However, key differences in transcript expression were present, including differential expression of Met, Tie1, VCAM1, VEGFB, and SELE among many others (Fig. 3C).

One of the biologic themes to emerge from this list is a loss in expression of functionally significant primary lymphocyte adhesion molecules. As tumor microenvironments mature, we find strong evidence for downregulation of SELE, MAdCAM-1, VCAM-1, ICAM-1, and ICAM-4—endothelial adhesion molecules that promote lymphocyte homing (Fig. 3C). We note the interaction between endothelial cell adhesion molecules and lymphocyte homing receptors is a key determinant in the infiltration of tumors by different subsets of lymphocytes, and downregulation of these pathways has been associated with tumor outgrowth. On the basis of these findings, we hypothesized the consequences of clonal evolution within the tumor microenvironment in the expressed genetic program may well influence tumor immune function. In addition to the downregulation of key lymphocyte adhesion molecules, we also observe strong evidence for decreases in expression of lymphocyte chemokines CXCL16, CXCL13, CXCL12, CXCL9, CXCL10, CCL19, and inflammatory cytokines IL15 (natural killer cell and T cell), IL11 (monocyte/macrophage), and G-CSF (granulocyte/dendritic Cell; Fig. 3C). The loss of proinflammatory cytokine and chemokine expression was associated with decreases in class I MHC expression and diminished expression of costimulatory molecules CD40 and ICOSL (Fig. 3C).

An additional biologic theme to emerge from our clonal analysis is a shift from paracrine to autocrine signaling as tissue microenvironments evolve. We see strong evidence for upregulation of self-sustaining growth signal networks; F₁ endothelial cells upregulate angiogenic ligands VEGF and VEGFB and their counter-receptor FLT1, as well as, the WNT receptor FZD7 and its ligand WNT5A (Supplementary Table S1). These pathways have been associated with tumor angiogenesis and reflect a shift from bidirectional supportive interactions between tumor and blood vessel to autonomous blood vessel growth. Furthermore, these changes are accompanied by the activation of alternative angiogenic programs such as upregulation of the hepatocyte growth factor receptor (HGF) and c-Met on F₁ endothelial cells (Supplementary Table S1). We note HGF levels contribute to the angiogenesis associated with several human cancers.

Finally, at no time point did we find evidence for contributions for either a hematopoietically derived cell or a circulating cell of any type to formation or maintenance of tumor vasculature in B16 engrafted tumors. Wild-type mice were surgically conjoined to mice expressing GFP under a ubiquitous (chicken β-actin) promoter. These mice develop full hematopoietic chimerism within a month. B16 tumors were engrafted into non-GFP animals and removed after 1 month for analysis. GFP⁺ cells were observed throughout the tumor microenvironment (Supplementary Fig. S2) but did not incorporate into any vascular structures (Supplementary Fig. S2). All GFP⁺ cells expressed the pan-hematopoietic marker CD45. We found no evidence of contributions from circulating cells, of any type, to cell turnover within tumor blood vessels.

This study adapts a unique technology, a multicolor Cre-dependent “Rainbow” reporter mouse, to show the formation and maintenance of vasculature in response to pathologic stimuli, undergoes dynamic changes in clonal architecture. By using lineage tracing that allows for the visualization and categorization of all infiltrating cells, we offer a new approach to shed light on the cellular framework underlying formation of aberrant tumor niches. Although it is generally accepted clonal progression occurs in cancer, with each heritable change increasing competitive competence of the cancer clone versus normal cell, the clonal dynamics of the surrounding tumor microenvironment are unknown. We provide evidence for large clonal expansions within an advanced tumor microenvironment. These clonal patterns are similar in character to the diversification, expansion, and ultimate selection within competitive microenvironments described in cancer–clone evolution. Furthermore, these findings parallel a number of recent reports that show distinct clonal patterns during homeostatic compared with pathologic settings. Two distinct progenitor pools have recently been reported to contribute to intestinal maintenance and repair. Following radiation injury, a normally quiescent population of Bmi-1–expressing intestinal stem cells outcompetes Lgr5⁺-expressing cells to clonally repopulate intestinal crypt (16). Although previous clonal tracing of homeostatic endothelium using X chromosome inactivation analysis indicates subpopulations of endothelial cells undergoes clonal proliferation (17, 18), it is still not determined whether clonogenic “winners” are mature endothelial cells, or whether they are distinct stem or progenitor-less mature cells that can give rise to clonal colonies of endothelium.

Our experiments have several potential implications. First, the evolution of subclones within the microenvironment is one pathway to allow cancer cells to remodel tissue microenvironments to their competitive advantages. In the case of tumor vasculature, alterations in lymphocyte-recognition tip the intratumor balance toward tumor immune escape and tolerance. Downregulation of endothelial adhesion molecules SELE, MAdCAM-1, VCAM-1, ICAM-1, and ICAM-4 may well influence tumor immune function. Moreover, this cellular framework may underlie a current barrier to immunotherapy. Adoptive transfer of engineered T cells for the treatment of solid cancer relies on the intratumoral accumulation of T cells. Licensing microenvironmental homing in favor of lymphocyte trafficking may be one approach to optimize these therapies. Finally, the developmental shift in transcriptomes in evolved subclones to alternative angiogenic programs may underlie escape mechanisms from current antiangiogenic pharmacotherapies and suggests the vascular tumor microenvironment is a moving therapeutic target.

No potential conflicts of interest were disclosed.

Conception and design: D.M. Corey, I.L. Weissman

Development of methodology: D.M. Corey, Y. Rinkevich

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.M. Corey

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.M. Corey, Y. Rinkevich, I.L. Weissman

Writing, review, and/or revision of the manuscript: D.M. Corey, Y. Rinkevich, I.L. Weissman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.M. Corey

Study supervision: I.L. Weissman

This study was supported in part by an NIH fellowship K12 HL087746 from the National Heart, Lung, and Blood Institute (D.M. Corey), 2T32AR050942-06A1 (D.M. Corey), and by grants U01 HL099999 and R01 CA86065 (I.L.Weissman).

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