Metastasis is the leading cause of cancer-related deaths, and metastatic cancers remain largely incurable due to chemoresistance. Biomarkers of metastatic cells are lacking, and probes that could be used to detect and target metastases would be highly valuable. Here we hypothesize that metastatic cancer cells express cell-surface receptors that can be harnessed for identification of molecules homing to metastases. Screening a combinatorial library in a mouse mammary tumor model of spontaneous metastasis identified cyclic peptides with tropism for cancer cells disseminated to the lungs. Two lead peptides, CLRHSSKIC and CRAGVGRGC, bound murine and human cells derived from breast carcinoma and melanoma in culture and were selective for metastatic cells in vivo. In mice, peptide CRAGVGRGC radiolabeled with 67Ga for biodistribution analysis demonstrated selective probe homing to lung metastases. Moreover, systemic administration of 68Ga-labeled CRAGVGRGC enabled noninvasive imaging of lung metastases in mice by PET. A CRAGVGRGC-derived peptide induced apoptosis upon cell internalization in vitro and suppressed metastatic burden in vivo. Colocalization of CLRHSSKIC and CRAGVGRGC with N-cadherin+/E-cadherin cells indicated that both peptides are selective for cancer cells that have undergone the epithelial-to-mesenchymal transition. We conclude that CRAGVGRGC is useful as a probe to facilitate the development of imaging modalities and therapies targeting metastases.

Significance:

This study identifies new molecules that bind metastatic cells and demonstrates their application as noninvasive imaging probes and vehicles for cytotoxic therapy delivery in preclinical cancer models.

Metastatic progression to therapy-resistant stages remains the overriding cause of cancer-associated mortalities (1). Metastasis from primary carcinomas is mediated by a hierarchical series of events (2). The ability of tumors to survive therapy directed at primary tumors and invade other tissues is rooted in the phenotypic plasticity of cancer cells (3). The latent developmental program of epithelial-to-mesenchymal transition (EMT) underlies the plasticity of carcinomas (4). EMT encompasses the continuum of morphologic and genomic transitions observed during carcinoma progression (5). Besides exacerbating the invasive, migratory, and metastatic potential of cancer cells at the invasive tumor front, the EMT endows malignant cells with therapy-resistance properties (6, 7). Currently, there are no effective approaches for early detection of these cells in tumors. Development of imaging probes for aggressive cancer cells would enable timely detection of metastatic cancers. Such probes could also lead to new therapeutic strategies to target these prometastatic cancer cells, which could improve patient outcome (8).

Our group has established expertise in identifying probes binding to cell-surface receptors expressed on cell populations of interest (9). Through in vivo screening of combinatorial libraries, we have isolated cell-targeting peptides that have been used for targeted delivery of experimental therapeutics in vivo (10). Bimodal ‘hunter-killer’ peptides, composed of a cytotoxic proapoptotic domain conjugated to a cell surface receptor-binding domain, have been used by our group for targeted cell ablation (11). Here, a screen of phage-displayed peptide libraries was conducted in an orthotopic mouse model of spontaneous breast cancer metastasis. We have identified two heptameric cyclic peptides homing to pulmonary metastases in the mouse model. We termed them Breast Lung Metastasis-homing Peptides (BLMP). Our results indicate that BLMPs home to cells undergoing EMT. The use of one of the BLMPs for metastasis detection was demonstrated in an experimental metastasis model using radiolabeled peptide probes. A BLMP conjugated with an apoptosis-inducing moiety was also shown to selectively kill cancer cells and reduce metastatic burden.

Cell lines and primary cell culture

All cell lines were cultured in DMEM media (Hyclone) supplemented with penicillin-streptomycin (Gibco) and 10% FBS (Atlas Biologicals). B16F10, MDA-MB-231, and 3T3L1 cell lines were from ATCC. 4T1.2 cells expressing mCherry and Luciferase were a gift from Sendurai Mani. All cells were tested for mycoplasma before expansion for experiments. Cisplatin-resistant 4T1.2 (4T1.2-CTX) cells were generated after treatment with cisplatin (Tocris Biosciences, 180 ng/mL).

Peptides

Hunter-killer (HK) peptides HK-BLMP5 and HK-BLMP6 were generated as “all-D-amino acid” cysteine-cyclized peptides (Celtek Bioscience). These proteolysis-resistant peptides were synthesized as acetate salts with an amino-hexanoic acid linker [NH-(CH2)5-CO] conjugating the BLMP and the apoptotic moiety KLAKLAKKLAKLAK (KLAKLAK2). Following chromatographic purification, peptides were dissolved in PBS into 10 mmol/L aliquots and stored frozen until further use. For the homing/internalization assay, peptides were biotin-labelled using the Sulfo-NHS-LC-Biotin Kit (APExBIO). Cells were plated in 24-well plates (1 × 105 cells/well). Following treatment with KLAKLAK2 or HK-peptides in 0.5 mL serum-free culture media for 8 hours, cells were stained with 0.4% Trypan Blue solution (MP Biomedicals; ref. 10). For radioactive studies, 1,4,7,10-tetraazacyclotetradecane-N′,N″,N,N″″–tetra-acetic acid (DOTA) t-Bu ester was purchased from Macrocyclics and peptides were synthesized by Bachem Americas. DOTA-peptide conjugation was performed on solid-phase using previously described protocols (12). Radiolabeling with 67Ga or 68Ga and high performance liquid chromatography (HPLC) determination of purity were performed according to published procedures (13). Radio-HPLC analysis of the radiolabeled peptides showed more than 95% radiochemical purity following C-18 Sep-Pak purification.

Animal experiments

Animal studies were approved by the Institutional Animal Care and Use Committee of UTHealth. Spontaneous cancer progression was modeled by injecting 2 × 104 4T1.2 (mCherry+ Luciferase+) cells into the mammary fat pad of 10-week-old female Balb/c mice (Jackson Laboratory) with a 31-gauge needle. Primary tumor volume was measured weekly with caliper and calculated using the (length × width2)/2 formula. At a tumor volume of approximately 200 mm3, mice were treated with cisplatin (Tocris Biosciences; 2.5 mg/kg) retro-orbitally once a week for 3 weeks to model chemoresistance. Metastatic progression was monitored with bioluminescence imaging. For experimental metastasis modeling in C57BL/6 mice (Jackson Laboratory), 1 × 105 B16F10 cells were intravenously injected into the tail vein. In both mouse models, pulmonary metastases were observed 3 weeks post implantation. The screen for metastasis-homing peptides was performed using a 50:50 mixture of phage cyclic peptide libraries CX7C and CX8C (C: cysteine; X: any amino acid residue) provided by Erkki Koivunen (9). For every biopanning round, a mouse with metastases 3 weeks postgrafting was intravenously injected with 1010 transducing units (TU) of the library. After 3 hours of circulation, the heart was perfused with 10 mL PBS. Flow cytometry was performed on the cell suspension from collagenase/dispase-digested lungs to isolate mCherry-positive metastatic tumor cells by using FACSAria/FacsDiva software (BD Biosciences). Selected phage were amplified through K91 E. coli infection and quantified via TU counting. For each round, clone enrichment was quantified based on recovery frequency among 96 phage peptide-coding inserts sequenced from PCR-amplified phage DNA (14). Homing was validated by injection of 1010 TU of individual phage-peptide clones (week 3 after tumor-cell injection). For radioactive biodistribution analysis, B16F10 tumor-grafted female C57BL/6 mice (n = 4 per cohort) were intravenously injected with 5 nmol/L 67Ga-DOTA (nontargeted control), 67Ga-DOTA-BLMP5, or 67Ga-DOTA-BLMP6 (100 μCi/mL). Mice were euthanized 1 hour after tracer injection, resected tissues were washed with PBS, and radioactive uptake was measured using a Cobra II automated gamma counter (Packard) and expressed as percentage of the injected dose per gram tissue (% ID/g tissue). For PET/CT studies, mice were intravenously injected with 7.4 MBq (200 μCi, 5 nmol) of 68Ga-labeled DOTA, DOTA-BLMP5, or DOTA-BLMP6, and imaged 45 minutes later using Albira small-animal PET/CT scanner (Bruker). Region-of-interest (ROI) analysis was done with the integrated PMOD software (PMOD technologies LLC.) to standardize uptake values and determine tumor-to-background ratios (TBR).

Immunofluorescence for tissue and cell analysis

Immunofluorescence (IF) on cells and formalin-fixed paraffin-embedded tissue sections was performed as described (15). Primary antibodies used: rabbit anti-Fd bacteriophage (Sigma Aldrich, 1:500), rabbit anti-Asp175-cleaved caspase-3 (Cell Signaling Technology, 1:100), mouse anti–E-cadherin and mouse anti–N-cadherin (BD Biosciences, 1:100). Secondary antibodies used: donkey Alexa 488-conjugated IgG (Invitrogen, 1:300) and Cy3-conjugated IgG (Jackson ImmunoResearch, 1:300). Biotinylated peptide internalization was visualized with Cy3-conjugated streptavidin IgG (Thermo Fisher Scientific, 1:50). Nuclei were stained with Hoechst 33258 (Invitrogen, 1:2000). Tissue and cell images were acquired with a Nikon Eclipse TE2000E Widefield Fluorescence Microscope (Nikon Instruments). Image quantification included 5 random microscopic fields at 20X magnification. Images were processed and analyzed with ImageJ software (NIH). Cadherin colocalization was quantified based on the previously described ImageJ software object-based approach (JACoP) methodology based on the Mander coefficient (16).

Statistical analysis

GraphPad Prism v.9.0.2 was used to graph data as mean ± SEM. Statistical significance was determined with unpaired t test or two-way ANOVA built-in analysis tools. P values < 0.05 were considered significant.

Data availability statement

The data generated in this study are available upon request from the corresponding author.

Screen for peptides homing to lung metastases

An adaptation of the previously described in vivo phage display biopanning methodology (17), was implemented in an orthotopic mouse model of triple-negative breast cancer (TNBC) to isolate peptides homing to pulmonary metastases (Fig. 1A). To simulate the clinical setting in which patients are treated with chemotherapy, we used cisplatin, a drug that has shown efficacy in patients with TNBC. We screened a pool of over 1010 combinatorial peptides displayed as fusions on the pIII protein of filamentous M13 bacteriophage. Mice were grafted with 4T1.2 tumor cells expressing red fluorescent protein (mCherry) and luciferase (Luc) and, once primary tumors became detectable, were treated with cisplatin as described (18). Following intravenous injection of the CX7C/CX8C phage-peptide library into mice, phage bound to metastatic cells was recovered from mCherry+ cells isolated from lung suspension by FACS (Fig. 1A). After three enrichment/selection rounds, 6 phage-displayed peptides were repeatedly recovered with a frequency above 1%. These peptides were named BLMPs, BLMP1 through BLMP6 (Fig. 1B). Sequencing of phage DNA revealed the following amino acid sequences of the peptide inserts: CGVLPYSLC (BLMP1); CSGVGIASC (BLMP2); CEGPMYAKC (BLMP3); CHLSFSTAC (BLMP4); CLRHSSKIC (BLMP5); and CRAGVGRGC (BLMP6). BLMP5 and BLMP6 were selected as candidates for future studies owing to their most significant enrichment (P ≤ 0.0001) in the last round of biopanning (Fig. 1B).

To confirm cancer cells as the target of peptides, cell culture was used to analyze individual phage-displayed BLMPs. IF analysis of cultured 4T1.2 and B16F10 murine melanoma cells incubated with individual phage-displayed peptides revealed that both BLMP5-phage and BLMP6-phage were internalized by over 50% of cancer cells (Fig. 1C). No internalization was seen upon incubation with control insertless phage that lacks a peptide insert (Supplementary Fig. S1A). Murine 3T3L1 preadipocytes were used as a negative control and uptake was not observed for BLMP6-phage (Fig. 1C). Compared with phage-displayed BLMP6, BLMP1, BLMP2, BLMP3, BLMP4, and BLMP5 exhibited higher nonspecific uptake by 3T3L1 cells (Fig. 1C; Supplementary Figs. S1B and S1C). Quantification of phage-displayed peptide binding based on phage TU recovery demonstrated either a lack of binding to B16F10 cells or nonspecific binding to 3T3L1 cells for BLMP1, BLMP2, BLMP3, BLMP4, and BLMP5 (Fig. 1D). This experiment confirmed BLMP6 phage selectivity for both B16F10 and 4T1.2 cells (Fig. 1D). Phage-displayed BLMP5 and BLMP6 exhibited binding to MDA-MB-231 human TNBC cells, suggesting interspecies conservation of the peptide-bound receptors (Supplementary Fig. S1D).

Validation of peptide metastasis homing

Homing of individual BLMPs to lung metastases was examined through intravenous administration of 1010 TU individual phage-displayed peptide clones into mice. Tissue biodistribution was analyzed by IF using phage-specific antibodies. Both BLMP5-phage and BLMP6-phage were found localized to lung metastases, identified by histopathology, but not to normal lung tissue (Fig. 2A–B). A negative control phage lacking a peptide insert did not localize to metastases upon injection (Supplementary Fig. S2A). BLMP5-phage exhibited homing to lungs in 4T1.2 spontaneous metastasis (Fig. 2A) but not in melanoma B16F10 experimental metastases (Fig. 2A) mouse model. BLMP6-phage exhibited homing to lungs in both 4T1.2 spontaneous metastases and melanoma B16F10 experimental metastases mouse models (Fig. 2B). In macro-metastases, BLMP5 localized predominantly along the invasive edges (Fig. 2A), whereas BLMP6 was abundant throughout (Fig. 2B). The lack of BLMP-phage localization to lungs in cancer-free mice further confirmed the metastasis tropism for both BLMP5 and BLMP6 (Supplementary Fig. S2B). Nonspecific accumulation of phage in the liver reticulo-endothelial system was, as expected, observed for phage irrespective of the peptide insert (Fig. 2A–B), including insertless control phage (Supplementary Fig. S2A). Besides the expected liver uptake, trace amounts of BLMP6-phage were observed in the kidney for the B16F10 model (Fig. 2B), suggesting renal dissemination of cancer cells. BLMP5-phage and BLMP6-phage was also observed in primary 4T1.2 tumors. While BLMP5 was dispersed throughout the tumor (Fig. 2A), BLMP6 was observed selectively in a subpopulation of cells associated with the vasculature (Fig. 2B). These results indicate that BLMP5 and BLMP6 bind to distinct receptors expressed in a fraction of tumor cells and in metastases.

Noninvasive detection of metastases with peptide probes

Next, we confirmed tumor cell tropism of BLMP peptides outside the phage context. We used the B16F10 experimental metastasis model to test if cancer cells can be detected by tissue distribution studies with radiolabeled peptide probes. BLMP5 and BLMP6 were conjugated with the radiometal chelator DOTA, which was then used for labeling with 67Ga. Uptake in most normal organs was less than 0.5% ID/g for both 67Ga-labeled BLMP5 and BLMP6 peptides (Fig. 2C), indicating low nonspecific uptake that was similar for the nontargeted control agent (67Ga-DOTA; Supplementary Fig. S2C). As expected, the major site of uptake was the kidneys due to high renal excretion of the radiolabeled peptides (Fig. 2C; Supplementary Fig. S2C). Analysis of organs for melanin-positive B16F10 cells revealed the presence of metastases in the kidneys of some B16F10 cell-injected mice (Fig. 2C). This likely accounts for the kidney uptake of 67Ga-BLMP6 being higher than for nontargeted 67Ga-DOTA (Supplementary Fig. S2C) and demonstrates peptide homing to cancer cells. Importantly, the lung signal for 67Ga-BLMP6–injected mice containing lung metastasis was more than 4 times higher than in the lungs of cancer-free mice (Fig. 2C; Supplementary Fig. S2D). No significant lung uptake was observed for 67Ga-BLMP5–injected mice containing lung metastasis (Fig. 2C). In a separate experiment, we injected 67Ga-BLMP5 and 67Ga-BLMP6 peptides into mice grafted with B16F10 tumors subcutaneously. While there was a detectable uptake of 67Ga-BLMP5, no uptake of 67Ga-BLMP6 by primary tumors above the 67Ga-DOTA background was observed (Supplementary Fig. S2E).

We also used PET/CT to determine if peptide homing to metastases can be imaged. Mice developing B16F10 lung metastases were intravenously injected with 68Ga-BLMP5 (n = 3), 68Ga-BLMP6 (n = 4), and the control agent, 68Ga-DOTA (n = 2). As shown in Fig. 2D, no signal was seen in the lung of the mice injected with 68Ga-BLMP5 (TBR of 1.61 ± 0.35) and 68Ga-DOTA (TBR of 1.51 ± 0.11) despite the presence of metastatic lesions in the lungs. Importantly, the presence of lung signal in mice injected with 68Ga-BLMP6 was associated with lung metastatic burden (TBR of 2.26 ± 0.1 in metastasis-positive and 1.39 ± 0.12 in metastasis-negative mice). Combined, our data demonstrate BLMP6 homing to metastases in and out of phage context and identify this peptide as a superior lead for further investigation.

Cancer cell ablation with HK peptides

To test if BLMPs can be used for targeted therapy delivery, we converted them into HK peptides based on the previously published strategy (19). The lead proapoptotic peptide termed HK-BLMP6 was synthesized as a fusion with the KLAKLAK2 apoptosis-inducing moiety via an amino-hexanoic acid linker (Fig. 3A). HK-BLMP5 of the same design was synthesized to be tested in parallel. First, we labeled HK-BLMPs with biotin to generate Bio–HK-BLMP6 and Bio–HK-BLMP5, in which KLAKLAK2 domain activity was neutralized. These biotinylated peptides were used to confirm that cancer cell uptake, essential for proapoptotic activity of HK peptides, was retained. Fluorescence analysis with Cy3-streptavidin probe revealed comparable uptake of Bio–HK-BLMP6 by murine 4T1.2 and B16F10 tumor cells, as well as by human MDA-MB-231 adenocarcinoma cells (Fig. 3B–C). Confirming specificity for cancer cells, uptake of Bio–HK-BLMP6 by control 3T3L1 murine fibroblasts was marginal (Fig. 3B–C). In contrast, Bio–HK-BLMP5 displayed nonspecific internalization by both cancer and control 3T3L1 cells (Supplementary Fig. S3A). This observation is consistent with the lack of radiolabeled BLMP5 homing to lung metastasis. Homing to B16F10 lung metastases was also confirmed for intravenously injected DOTA-BLMP6 labeled with Cy3 fluorophore. In contrast, DOTA-BLMP5 and control DOTA Cy3 conjugates did not display comparable metastasis localization (Supplementary Fig. S3B).

We then tested the activity of biotin-free HK peptides. Following treatment with 0.1 mmol/L HK-BLMP6 in vitro, the trypan blue exclusion assay revealed significant cell-death induction in the 3 tumor cell lines (Fig. 3D). IF analysis corroborated HK-BLMP6–induced apoptosis mediated by caspase-3 cleavage (Supplementary Fig. S3C). Importantly, HK-BLMP6 did not induce cell death in control 3T3L1 cells (Fig. 3D; Supplementary Fig. S3C). No cancer cell death was observed upon treatment with untargeted KLAKLAK2 (Supplementary Fig. S3D and S3E). In contrast to HK-BLMP6, HK-BLMP5 did not induce cell death at 0.1 mmol/L, as measured by trypan blue staining. While 0.3 mmol/L HK-BLMP5 concentration did kill a fraction of cancer cells, elevated trypan blue staining was also observed in 3T3L1 preadipocytes (Supplementary Fig. S3E), again pointing to relatively inferior selectivity of BLMP5. We also assessed the effect of HK-BLMP6 on cancer metastases in vivo. Mice grafted with Luc-positive 4T1.2 cells were metronomically injected with HK-BLMP6 and imaged by In Vivo Imaging System (Supplementary Fig. S4A). After 5 weeks, mice treated with HK-BLMP6 peptide had lower metastatic burden than vehicle-treated mice, which was confirmed by histopathology (Supplementary Fig. S4B and S4C). These findings demonstrate the potential utility of BLMP6 as a vehicle for selective agent delivery to cancer cells.

Peptide selectivity for cancer cells undergoing EMT

Nonuniform localization of BLMP peptides within pulmonary metastases (Fig. 2A–B) prompted an investigation into the potential characteristics of cancer cells favoring such a distinctive localization pattern. Because EMT is a feature of cells predisposed to metastasize (4), we tested if peptides home to cancer cells that lose their epithelial phenotype. For this, we used antibodies against N-Cadherin, marking mesenchymal cells, and E-Cadherin, marking epithelial cells (20) for IF analysis on lung tissue sections derived from murine TNBC and melanoma models. BLMP6-phage demonstrated a definitive selectivity for metastatic tumor cells that primarily expressed N-Cadherin (Fig. 4A). Concomitantly, limited localization to E-cadherin+ epithelial cells was seen, an observation consistent across both cancer models assessed (Fig. 4A). Significance of EMT colocalization for BLMP6-phage in both metastasis models was quantified by ImageJ (Fig. 4B). Selectivity for N-cadherin+ E-cadherin cells was also observed for BLMP5-phage (Supplementary Fig. S5A).

Peptide selectivity for cells undergoing EMT was verified by analysis of cell lines, which display heterogeneity in culture. Both BLMP6-phage and Bio–HK-BLMP6 preferentially bound to 4T1.2, B16F10, and MDA-MB-231 cancer cells that lacked E-cadherin (Fig. 4C). A similar preference for E-cadherin cells was exhibited by BLMP5 in culture (Supplementary Fig. S5B). Because chemotherapy has been shown to preferentially kill epithelial cancer cells and select for those that undergo EMT (21), we treated cells with cisplatin. Indeed, cisplatin-resistant 4T1.2 displayed reduced E-Cadherin expression while maintaining BLMP6 binding (Fig. 4C). These results were confirmed by N-cadherin/BLMP6-phage costaining on cultured 4T1.2 cells treated with cisplatin and B16F10 cells (Supplementary Fig. S5C). N-cadherin IF on primary 4T1.2 tumors from mice injected with BLMP6-phage indicated its homing to cells undergoing EMT (Supplementary Fig. S5C). Selectivity for cells with the EMT phenotype was also confirmed for BLMP5 (Supplementary Fig. S5D and S5E). Combined, our observations suggest that BLMP5 and BLMP6 bind to cancer cell receptors expressed on metastatic cancer cells losing epithelial properties. Protein BLAST analysis identified similarity of BLMP5 to a segment of mucin 5 (MUC5) and similarity of BLMP6 to a segment of latent TGFβ-binding protein 4 (LTBP4), with 5 amino acids completely matching for each peptide (Supplementary Fig. S5F). This suggests that these proteins, with a known function in EMT, could be mimicked by the peptides.

Metastatic carcinoma progression despite the initial response of primary tumors to chemotherapy often occurs due to survival of cells undergoing EMT. Identification of compounds selective for cancer cells post-EMT could help target metastases. Here, we for the first time characterize peptides homing to metastatic cells. There are several preceding reports of peptides binding to cancer cells and suppressing metastases identified by display technologies (22, 23). However, the screens to identify them were performed on cell lines ex vivo and the peptides were shown to home to primary tumors rather than metastases. An in vivo screen has been reported for the 4T1 graft model; however, the isolated peptides targeted macrophages rather than cancer cells (24). There are no publications reporting peptides selectively homing to metastatic cells in mouse models that were validated by biodistribution or experimental therapy in vivo.

We show that peptides BLMP5 and BLMP6, which we validated in vivo and in cell culture models, are selective for cancer cells undergoing EMT. In the future, isolation of cell-surface receptors for BLMP peptides may open new avenues to interrogating aggressive cancer cells. Mucin and TGFβ signaling pathways are well known to mediate EMT. Similarity of BLMP5 and BLMP6 to proteins engaged in these pathways is likely to streamline receptor identification. Selective apoptosis induction and reduction of metastatic burden with BLMP6-derived HK peptide is a proof-of-principle to be subsequently tested in other cancer models. Noninvasive metastasis detection with radiolabeled BLMP6 peptide tracers also outline their promise in potential new imaging applications that could be a step beyond conventional approaches.

In summary, our study shows that BLMP6, or other molecules targeting carcinoma cells post-EMT, could be developed for guiding chemotherapy and gene therapy to metastases, for which systemic therapy delivery has remained poorly effective. Based on our results, we envision new approaches to targeting metastatic cells, which could be transformative in cancer medicine.

S. Subramanian reports a patent for UTSH.P0376US.P1 pending. A.C. Daquinag reports a patent for UTSH.P0376US.P1 pending to UTHealth. M.G. Kolonin reports a patent for UTSH.P0376US.P1 pending. No disclosures were reported by the other authors.

S. Subramanian: Investigation. A.C. Daquinag: Investigation. S. AghaAmiri: Investigation. S.C. Ghosh: Resources, investigation, writing–original draft, writing–review and editing. A. Azhdarinia: Conceptualization, resources, data curation, formal analysis. M.G. Kolonin: Conceptualization, resources, data curation, supervision, funding acquisition, validation, visualization, project administration, writing–review and editing.

This study was supported by the Harry E. Bovay, Jr., John S. Dunn, and Welch Foundations. The authors thank Gheath Alatrash, Sendurai Mani, Joseph Robiya, and Naoto Ueno for helpful discussions. They thank Zhanguo Gao and Zhengmei Mao for technical help.

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.

1.
Welch
DR
,
Hurst
DR
. 
Defining the hallmarks of metastasis
.
Cancer Res
2019
;
79
:
3011
27
.
2.
Valastyan
S
,
Weinberg
RA
. 
Tumor metastasis: molecular insights and evolving paradigms
.
Cell
2011
;
147
:
275
92
.
3.
Ganesh
K
,
Massague
J
. 
Targeting metastatic cancer
.
Nat Med
2021
;
27
:
34
44
.
4.
Kalluri
R
,
Weinberg
RA
. 
The basics of epithelial-mesenchymal transition
.
J Clin Invest
2009
;
119
:
1420
8
.
5.
Yang
J
,
Antin
P
,
Berx
G
,
Blanpain
C
,
Brabletz
T
,
Bronner
M
, et al
Guidelines and definitions for research on epithelial–mesenchymal transition
.
Nat Rev Mol Cell Biol
2020
;
21
:
341
52
.
6.
Aware
V
,
Gaikwad
N
,
Chavan
S
,
Manohar
S
,
Bose
J
,
Khanna
S
, et al
Cyclopentyl-pyrimidine based analogues as novel and potent IGF-1R inhibitor
.
Eur J Med Chem
2015
;
92
:
246
56
.
7.
Thomas
SM
,
Grandis
JR
. 
Targeting mesenchymal exaptation to mitigate tumor growth
.
Cell Cycle
2011
;
10
:
2626
7
.
8.
Pattabiraman
DR
,
Weinberg
RA
. 
Targeting the epithelial-to-mesenchymal transition: the case for differentiation-based therapy
.
Cold Spring Harb Symp Quant Biol
2016
;
81
:
11
9
.
9.
Arap
W
,
Kolonin
MG
,
Trepel
M
,
Lahdenranta
J
,
Cardo-Vila
M
,
Giordano
RJ
, et al
Steps toward mapping the human vasculature by phage display
.
Nat Med
2002
;
8
:
121
7
.
10.
Daquinag
AC
,
Tseng
C
,
Zhang
Y
,
Amaya-Manzanares
F
,
Florez
F
,
Dadbin
A
, et al
Targeted proapoptotic peptides depleting adipose stromal cells inhibit tumor growth
.
Mol Ther
2016
;
24
:
34
40
.
11.
Daquinag
AC
,
Tseng
C
,
Salameh
A
,
Zhang
Y
,
Amaya-Manzanares
F
,
Dadbin
A
, et al
Depletion of white adipocyte progenitors induces beige adipocyte differentiation and suppresses obesity development
.
Cell Death Differ
2015
;
22
:
351
63
.
12.
Azhdarinia
A
,
Wilganowski
N
,
Robinson
H
,
Ghosh
P
,
Kwon
S
,
Lazard
ZW
, et al
Characterization of chemical, radiochemical and optical properties of a dual-labeled MMP-9 targeting peptide
.
Bioorg Med Chem
2011
;
19
:
3769
76
.
13.
Hernandez Vargas
S
,
Kossatz
S
,
Voss
J
,
Ghosh
SC
,
Tran Cao
HS
,
Simien
J
, et al
Specific targeting of somatostatin receptor subtype-2 for fluorescence-guided surgery
.
Clin Cancer Res
2019
;
25
:
4332
42
.
14.
Kolonin
MG
,
Saha
PK
,
Chan
L
,
Pasqualini
R
,
Arap
W
. 
Reversal of obesity by targeted ablation of adipose tissue
.
Nat Med
2004
;
10
:
625
32
.
15.
Zhang
Y
,
Daquinag
AC
,
Amaya-Manzanares
F
,
Sirin
O
,
Tseng
C
,
Kolonin
MG
. 
Stromal progenitor cells from endogenous adipose tissue contribute to pericytes and adipocytes that populate the tumor microenvironment
.
Cancer Res
2012
;
72
:
5198
208
.
16.
Bolte
S
,
Cordelieres
FP
. 
A guided tour into subcellular colocalization analysis in light microscopy
.
J Microsc
2006
;
224
:
213
32
.
17.
Kolonin
MG
,
Sun
J
,
Do
KA
,
Vidal
CI
,
Ji
Y
,
Baggerly
KA
, et al
Synchronous selection of homing peptides for multiple tissues by in vivo phage display
.
FASEB J
2006
;
20
:
979
81
.
18.
Su
F
,
Daquinag
AC
,
Ahn
S
,
Saha
A
,
Dai
Y
,
Zhao
Z
, et al
Progression of prostate carcinoma is promoted by adipose stromal cell-secreted CXCL12 signaling in prostate epithelium
.
NPJ Precis Oncol
2021
;
5
:
26
.
19.
Ellerby
HM
,
Bredesen
DE
,
Fujimura
S
,
John
V
. 
Hunter-killer peptide (HKP) for targeted therapy
.
J Med Chem
2008
;
51
:
5887
92
.
20.
Dongre
A
,
Rashidian
M
,
Reinhardt
F
,
Bagnato
A
,
Keckesova
Z
,
Ploegh
HL
, et al
Epithelial-to-mesenchymal transition contributes to immunosuppression in breast carcinomas
.
Cancer Res
2017
;
77
:
3982
9
.
21.
Su
F
,
Ahn
S
,
Saha
A
,
DiGiovanni
J
,
Kolonin
MG
. 
Adipose stromal cell targeting suppresses prostate cancer epithelial-mesenchymal transition and chemoresistance
.
Oncogene
2019
;
38
:
1979
88
.
22.
Zhou
C
,
Kang
J
,
Wang
X
,
Wei
W
,
Jiang
W
. 
Phage display screening identifies a novel peptide to suppress ovarian cancer cells in vitro and in vivo in mouse models
.
BMC Cancer
2015
;
15
:
889
.
23.
Yang
W
,
Luo
D
,
Wang
S
,
Wang
R
,
Chen
R
,
Liu
Y
, et al
TMTP1, a novel tumor-homing peptide specifically targeting metastasis
.
Clin Cancer Res
2008
;
14
:
5494
502
.
24.
Scodeller
P
,
Simon-Gracia
L
,
Kopanchuk
S
,
Tobi
A
,
Kilk
K
,
Saalik
P
, et al
Precision targeting of tumor macrophages with a CD206 binding peptide
.
Sci Rep
2017
;
7
:
14655
.

Supplementary data