Analyzing cell migration history in vivo using fluorescent fibrillar collagen trails Authors: Zhongming Chen, Branden S Moriarity, Paolo P Provenzano Published: 2026 ABSTRACT AbstractCell migration is a critical process in development, homeostasis, and disease. While cell migration in vitro is well investigated, much less is known about migration deep within tissues, largely due to limitations associated with deep cell imaging in tissues. In this study, we investigated cell migration history in vivo by developing a strategy based on recording cell trails formed by fusion of fluorescent protein and collagen secreted by migrating cells. By engineering different cell lines to express either fluorescent protein-fused Col1a1 or Col1a2 we identified trails formed in normal mouse tissues as well as primary tumors and metastatic organ sites. Analyses of the trail patterns revealed the paths taken by migrating cells and regions that suggest group trails, including trails along vascular adventitia. These results demonstrate that cell migration history can be traced in mouse models through postmortem analysis of normal and cancerous tissues.Subject terms: Cancer imaging, Imaging Introduction Cell migration plays key roles in development, tissue homeostasis, and disease, including cancer progression, metastasis, and immune cell infiltration1–3. Much of our understanding of cell migration is from studies imaging movement within in vitro environments or ex vivo cultured tissues4–7. Likewise, intravital imaging with multiphoton microscopy in vivo has revealed key aspects of cell migration8–10, but provides information that is largely restricted to regions within ~200–1000 μm from the tissue surface due to current limitations in optical imaging depth10–12. Additionally, intravital imaging most often provides shorter-term imaging (up to a few hours per session) in a single location or region8,13,14. Furthermore, while the use of optical windows can allow for imaging over multiple days within the field of view of the window15, this requires a surgical intervention and the imaging depth remains limited. Hence, as little is known about cell migration deep in organs over long time scales, technologies are needed that provide key movement information. In recent years, an alternative strategy of recording cell migration in fixed tissues rather than live cell imaging was developed based on analyzing cell trails assembled from GFP coupled to the basement membrane16. In transgenic Drosophila, this secreted fusion protein approach allowed for trace mapping of follicle rotation during embryo development15. Overall, this approach demonstrated that cell migration history could be evaluated in fixed tissues from a relatively small and simple model system, i.e., Drosophila embryo. However, this strategy has not yet been applied to more complex mammalian models, particularly with fibrillar proteins that allow for line-like tracks. In this study, we generate mouse and human cell lines that secrete fluorescent protein-tagged extracellular matrix (ECM), specifically, fusion proteins of mCherry and collagen I subunit a1 or a2 (Col1a1 or Col1a2), and examine the secretion of fusion proteins and collagen fiber assembly in cultured cells. We then evaluate trail formation in normal mouse tissues and breast cancer mouse models. The strategy based on the fusion proteins and the engineered cell lines will advance the fields of cell and cancer biology as information on cell migration and position deep within opaque organs can be obtained and examined with fluorescence microscopy from postmortem and archival tissues. Results Tracking migration history of engineered cells in vitro and in vivo Our strategy is based on cell migration trails formed from fusion proteins of fluorescent proteins and extracellular matrix (FP-ECM) secreted by the cells. We first validated the strategy through proof-of-principle experiments by recording cell migration trails in cultured cardiac cells. Cell migration is critical for normal heart development and repair16–18, where new cells can move to the injury sites to heal the wound. A few cardiac progenitor cells (e.g., c-kit + CD45- cells and ABCG2+ cells) may contribute to heart regeneration by migrating from specific niches to cardiomyocytes19,20. However, the migration from the niches to cardiomyocytes is challenging to investigate in the heart. To investigate the migration of these cells, we first tested our strategy using cell membrane-localized tdTomato+ c-kit + CD45- cardiac cells isolated from the hearts of mTmG transgenic mice21, and transfected with a plasmid encoding an enhanced yellow fluorescent protein (EYFP)-tagged fibronectin. Indeed, the cells emitted red tdTomato fluorescence and formed long visible trails via YFP fluorescence in culture over 3 days (Supp Fig. 1a). The trail was longer than 2000 µm, with a bright cell body and a clear trail that was captured over time (Supp Fig. 1b). This result suggests that fluorescent protein-tagged fibronectin forms visible trails and records cell migration history in vitro. As such, we next tested this concept by transplanting the cardiac cells into either the heart or the skeletal muscles of the legs of immunodeficient mice. However, we failed to accurately identify EYFP+ trails in either setting due to a weak fluorescent signal relative to the background autofluorescence signal, particularly in thick sections ( ~ 60 μm). These results suggest that the EYFP-tagged ECM can be used for some in vitro settings for long-term tracing but has limited applicability for tracing cell migration history in vivo. To overcome these signal-to-noise issues, we designed overexpression constructs encoding a promoter (CAG or EF-1a), signal peptide, N-peptide, FP (GFP or mCherry), a linker, the Col1 subunit (Col1a1 or Col1a2), C-peptide, as well as a terminator (Fig. 1a), where the fusion protein from the fluorescent protein and the collagen mRNA is translated into polypeptide chains that undergo post-translational modifications and are ultimately hierarchically assembled from microfibrils to fibrils and fibers that form ECM22,23. The new constructs switch fluorescent proteins and replace fibronectin with collagen I, which is more resistant to proteases than fibronectin due to the triple helical conformation and thus has the potential to generate more stable trails in vivo24. To confirm expression, we expressed GFP-Col1a2 in 293 T cells and collected media 48 hours after transfection to evaluate protein secretion (Supp Fig. 2a). The secretion of the FP-Col1a2 fusion protein was confirmed by immunoprecipitation with antibodies of mouse 1a2 and GFP (Supp Fig. 2b, c). However, the GFP fluorescence in the GFP-Col1a2-transfected mouse cells was relatively weak relative to background fluorescence. We thus generated a plasmid with mCherry and expressed mCherry-Col1a2 in GFP+ cardiac cells (ATP-binding cassette transporter, ABCG2, cardiac cells (mC+GFP+ cells; see Supp. Table 1 for a full list of cell line nomenclature) isolated from the hearts of transgenic mice)25. The transfected cells formed trails in vitro with a GFP+ cell body and a bright mCherry+ trail (Fig. 1b). Cells form migration trails in vitro and in vivo from secreted fusion protein of mCherry and Collagen I. a Diagram of a plasmid encoding the fusion protein, collagen I post-translational processes and assembly, and trail formation. The construct contains a promoter, fluorescent protein (FP), a linker (3x GGGGS), mouse Col1 subunit (Col1a1 or Col1a2). b, c mCherry-Col1a2+ GFP + ABCG2+ cells (mC+ GFP+ cells) form trails (b) in vitro or (c) in leg muscles of immunodeficient mice 7 days after transplantation. d A trail assembled from image segments from different z optical sections from the white rectangle region outlined in (c). (e–g) Representative trails formed by mCherry-Col1a2+ cells in vivo. Bar = 50 μm. e Trails formed by mCherry-Col1a2+ c-kit+ CD45- cells (mC+ c-kit cells) in mouse hearts; Red: mCherry, Green: Alexa 488-phalloidin staining, Blue: DAPI, Bar = 50 μm. f, g mCherry-Col1a2+ smooth muscle cells (mC+ SMCs) form trails 7 days after transplantation; Red: mCherry, Blue: DAPI, Bar = 50 μm. The plots of each trail (b, d, e, g) represent the fluorescence intensities at noted positions on the trails, with the numbers 1–5 in the plots corresponding to the positions of cell trails from tail (1) to cell body (5). h–j mCherry immunofluorescence to detect trails in a heart frozen section two weeks after transplantation with mC+ GFP+ cells. The tissue sections had been stored for 18 months in −20 °C before examination. In overlay images, Red: mCherry background, Green: mCherry antibody staining, Blue: DAPI. White arrows indicate the positions of the same trails in different channels (h=mCherry; i=Alexa 488-conjugated secondary antibody; j=merged image). Bar = 50 μm. Migration trail patterns from mCherry-Col1a2 EO771 cells in orthotopic breast tumors. Frozen tissue sections from primary tumors 5 weeks after transplantation of mCherry-Col1a2 + EO771 cells (mC+ EO771 cells) into the fat pad of C57BL/6 mice. a A trail formed by one cell; b Group trails formed by either a group of cells or a cell traveling back and forth along a path; Red: mCherry, Blue: DAPI. c mCherry fluorescence intensities of optical cross sections shown in 3 (top image with arrows indicating the region), marked as 3 in (b). The arrows indicate peaks of mCherry fluorescence intensities. d Quantification of the peaks in optical cross sections 1–5 marked in (b). (e) Group trails (Red) along blood vessels (Green; tissue sections stained with CD31 antibody). f Group trails (Red) along vascular adventitia (Green; vasculature labeled by intravital injection of WGA (Green). Bar = 10 μm (a, b) and 50 μm in (e, f). Regions for inset images are marked with white boxes. To test for the formation of migration trails in solid tissues in vivo, we transplanted the mC+GFP+ cells into the heart (left ventricle) of immunodeficient mice to avoid transplant rejection. Indeed, trails with bright red fluorescence and green cell bodies were identified in the heart in one to two weeks (Fig. 1c, d). We further tested the robustness of trail formation in both mouse heart and leg skeletal muscles (lateral and medial head of gastrocnemius) using mCherry-Col1a2 expression in additional mouse cell lines: either CD45- c-kit+ cardiac cells (mC+c-kit cells; Fig. 1e) or aortic smooth muscle cells (mC+SMCs; Fig. 1f, g). The mCherry-collagen I trails in the heart and skeletal muscles were bright and distinctive from tissue autofluorescence, with intensity varying along the trails. In general, the mCherry fluorescence intensity on the tail side is weaker than on the head side (positions 1 and 5 in Fig. 1d–g, respectively). Likewise, the mC+ GFP+ cells often form long trails in skeletal muscles. Indeed, we were able to overlay several fluorescence images to reconstruct long trails ( > 500 µm, Supp Fig. 3a, b). Notably, these results suggested that mCherry-collagen I fusion proteins can form bright trails that record cell migration history in the body. Lastly, to assess the stability and robustness of our platform, we examined mCherry signaling after long-term storage. After being stored at -20 °C for 18 months, we found that the mCherry-Col1a2 fluorescent signal was completely lost (Fig. 1h). However, we found that mCherry-Col1a2 fusion protein trails can still be identified by immunofluorescent staining using anti-mCherry antibodies after long-term storage (Fig. 1i, j). Furthermore, we observed that the engineered cells in this study, while selected with antibiotics or FACS sorting after transfection, had heterogeneity in their FP-ECM secretion. Some trails were weak in fluorescence (Supp Fig. 4a). Thus, to more robustly identify weak trails, we tested immunofluorescent staining using mCherry antibody to enhance detection signaling. Staining resulted in bright trails that were easy to identify (Supp Fig. 4b, c). These results suggest that migration history can still be evaluated when the primary signal is diminished in archived tissue samples, and that a weak trail signal can be enhanced through immunofluorescence staining. Human mCherry-Col1a1 + MDA-MB-231 carcinoma cells form trails in orthotopic breast tumors. a Schedule of mCherry-Col1a1+ MDA-MB-231 (mC1a1 + MM231) cell transplantation into mammary fat pads of NSG mice. b Representative images from frozen sections of mammary glands 62 days after engraftment of mC1a1 + MM231 cells into the 4th inguinal mammary gland. c A single trail and d a group of trails entangled together. White box in (b) indicates the region in (c); the blue box identifies the region magnified in (d). Red: mCherry. Bar = 50 μm. Engineered mouse breast carcinoma cells form migration trails in primary and metastatic tumors We next applied the cell migration history platform to study cell migration in solid tumors, where migration plays a crucial role in disease progression to metastasis, which remains the leading cause of cancer deaths26–28. We first employed an orthotopic breast cancer model using C57BL/6 mice and the EO771 luminal B mammary carcinoma cell line derived from the same mouse strain29. EO771 cells were transfected with the mCherry-mCol1a2 plasmid (mC+EO771 cells). First, to determine whether the expression of the FP-collagen fusion protein affects cell migration, we quantified in vitro migration of control vs. mC+ EO771 cells. We observed that the migration paths, direction change rate, mean speed, and total travel distance are not significantly different (Supp Fig. 5a–d), suggesting that cell migration is not significantly impacted by the secreted collagen I fusion protein. To analyze migration trials in vivo, we transplanted mC+EO771 cells into the fourth mammary gland of syngeneic C57BL/6 mice and identified two types of trails in primary breast glands: single trails presumably formed by single cells (Fig. 2a) and groups of trails (Fig. 2b). We quantified mCherry fluorescence intensity in different optical cross sections and resolved 7 peaks representing 7 individual trails in a group trail (Fig. 2c). The intensities of the individual peaks varied in different optical cross-sections, but the average intensities were not significantly different (Fig. 2d). Furthermore, the morphology of some group trails suggest that some trails may also localize along blood vessels. We hypothesized that these trails may be formed during cancer cell migration along blood vessels. Thus, we stained the tissue sections with the CD31 antibody. Indeed, the group trails were along CD31+ endothelial cells (Fig. 2e). To confirm this finding, we further labeled the mouse vasculature lumen by tail vein injection of FITC-conjugated wheat germ agglutinin30. We imaged the whole-mount tumor tissues and confirmed trails along vascular adventitia (Fig. 2f). These results suggest that the group trails observed in primary tumors and distant organs can often be found along blood vessels. We further examined the trail formation in time course by harvesting tumors in the primary mammary glands and distant organs at defined time points (Fig. 3a) to identify early trails and trails present after weeks in vivo. Although primary tumors were palpable in 2 to 4 weeks, the trails formed by mCherry-Col1a2 + EO771 cells were visible as early as day 3 after grafting (Fig. 3b). Furthermore, we observed thick and dense trails formed on day 7 and later in mammary glands (Fig. 3c) or primary tumors (Fig. 3d, e). In order to examine metastatic sites, we evaluated migration trails in remote organs collected five weeks after tumor formation. We identified trails in the pancreas, intestines, kidney, and lung (Fig. 3f–i). The visible trails in these organs also suggest that fluorescent proteins incorporated in collagen I fibers are reasonably stable for some time, at least during cancer cell dissemination in distal organs. We quantified mCherry-Col1a2 expression at different time points after cell transplantation by measuring the mean mCherry fluorescence intensity in each fluorescence image (Fig. 3j). We found that the mCherry-Col1a2 protein signal in the transplanted tissues increases over time. We further quantified trails formed in the primary tumors at different time points after cancer transplantation by counting the number of single (Fig. 3k) and group trails in each optical field (Fig. 3l). We found the number of single trails significantly increased one or two weeks after transplantation. After three weeks, the single-trial numbers diminished (Fig. 3k). Group trial numbers significantly increased one week after transplantation and remained elevated at 3–5 weeks (Fig. 3l). Overall, these data demonstrate that migration path history can be obtained from both primary and metastatic disease across a broad range of time points. Cell migration trails in primary tumors and metastatic organs. a Schedule of mCherry-Col1a2 + EO771 cell transplantation into mammary fat pads of C57BL/6 mice and tissue harvest. b–e Representative fluorescence images taken from frozen sections of mammary glands (3 days to 5 weeks). f–i Representative trails formed in metastatic organs 5 weeks after transplantation. j Mean mCherry-Col1a2 fluorescence intensities at different days after cell transplantation. k, l Numbers of k single or l group trails per optical field. Each point represents a biological sample (sections from a mouse) of each time point. Regions for inset images are marked with white boxes. Error bars are presented as mean ± s.d. *p < 0.05 compared to the day 0 condition in (j); #p < 0.05 compared to the <1 condition in (k, l) Bars = 50 μm. Engineered human breast cancer cells form migration trails in mice To test whether our platform can map migration history for human carcinoma cells, we examined paths using the mCherry-Col1a2 fusion proteins secreted from the MDA-MB-231 triple-negative breast carcinoma line31. mCherry-Col1a2 MDA-MB-231 cells (mC+MM231 cells) were grafted into the fourth mammary fat pad of immunodeficient mice (Supp Fig. 6a). Using this approach, we observed punctate structures that can link to form trail-like paths (Supp Fig. 6b). However, they did not produce clean lines for tracking. Therefore, to determine if we could enhance this system, we tested an alternative plasmid encoding mCherry-Col1a1 fusion protein, since Col1a1 can also form Col1a1 homotrimers that are more stable than the Col1a1 and Col1a2 heterotrimer32–35. MDA-MB-231 cells stably expressing the mCherry-Col1a1 fusion protein (mC1a1 + MM231 cells) produce trails in vitro and in vivo. To examine trail formation from live cells, we embedded mC1a1 + MM231 in 3D matrices co-cultured with fibroblasts and supplemented with ascorbic acid, as modifying enzymes and ascorbic acid, which are present in vivo, are required for collagen fiber formation in vitro (Supp Fig 7a, d). In the presence of fibroblasts and ascorbic acid supplementation, we observed mCherry+ trails (Supp Fig 7b, c, e). The mC1a1 + MM231 cells were then grafted into the mammary glands of immunodeficient mice (Fig. 4a). Fluorescence signal was noticeably stronger in vivo with the mCherry-Col1a1 fusion protein, and we were again able to identify trails in the primary tumors as far out as 9 weeks after transplantation (Fig. 4b–d). These trail patterns were similar to those of mCherry-Col1a2 expressing EO771 cells (Fig. 2b, c). These results suggested that mCherry-Col1a1 fusion protein forms trails in tissues and can be used to trace the migration of cells with low to moderate expression of endogenous collagen I. Trail unmixing by 3D reconstruction or cell dilution After transplantation, engineered cells proliferate, express FP-Col1 and form dense migration trails in tissues or tumors, which can sometimes make it difficult to identify and unmix single trails from maximal projection fluorescence images (Fig. 5a). As such, we have explored two approaches to further unmix trails: 3D reconstruction from different Z-stack images where each imaging plane is evaluated, similar to what we reported for cardiac tissues, and dilution of engineered cells with unlabeled cells. We examined trails in Z-stack images and found that whole trails can be reconstructed from segments in different Z positions (Fig. 5b) as the intersected trails in maximal projection images are separated from each other in different Z positions (Fig. 5c). These results suggest that 3D reconstruction can support observation of trails in tumors. Alternatively, we diluted mCherry-Col1a2+ (mC + ) EO771 cells with unlabeled EO771 cells with different ratios before transplantation (Fig. 5d–f). After dilution, the trails are highly separated from each other in the mammary glands, particularly with a 1:16 dilution of the mC+ EO771 cells in two weeks. These results suggest that cell dilution is an effective alternative approach to more cleanly track single-cell behavior, especially in the long term. Trail unmixing by 3D reconstruction and the impact of cell dilution. a Trails formed by mC+ EO771 cells in primary tumors 5 weeks after transplantation. z-stack images. z: 1-28 with regions of interest marked as 1–14. b Two trails reconstructed from imaging across different z positions. c Trail segments across different z positions (z18, 22, and 28) and in a maximum intensity projection image (z11-18). d A group of merging or separating trails. Each trail (indicated with white arrows) separated into different z positions images (z3-z28) where some trails rapidly change directions (z28). e Two groups of trails interconnected by three trails (indicated with yellow arrows). f Measurement of trail width is based on fluorescence intensity profile along a line perpendicular to a trail. The length of the arrow line at 2-fold of the background signal indicated the trail width. g Comparison of cell nucleus diameter and trail width, data are presented as mean ± s.d., n = 22 and 23, respectively; *: p < 0.05 (h–j) mC+ EO771 cells were diluted with unlabeled EO771 cells at 1:1, 1:4, or 1:16 ratios before transplantation into fat pads of C57BL/6 mice. Red: mCherry; Blue: DAPI counter staining, Bar = 50 μm. Tracking migration paths history of adoptively transferred therapeutic T cells To determine if we could track T cells in the therapeutic setting, we examined adoptively transferred CAR-T cells in pancreatic duct adenocarcinoma (PDA). PDA is a poor-prognosis cancer with an increasing incidence that is highly metastatic and largely resistant to impactful therapeutic interventions, including immune therapies36. While chimeric antigen receptor (CAR) T cell therapies can be effective for the treatment of some liquid tumors, their impact on solid tumors has, to date, been much more limited. This is due in part to our heterogeneous distribution of T cells in PDA37,38 and limited understanding of T cell movement in complex tumor microenvironments39,40,44. Hence, unlocking new knowledge about CAR T cell distribution and movement over time can help provide design criteria for cellular therapeutics and/or immune therapy regimes. We generated mesothelin-CAR (MSLN-CAR) T cells secreting mCherry-Col1a1 fusion proteins (mC1a1+CAR-T cells) since mesothelin is a tumor-associated antigen common to human PDA41,42. In 3D matrices, analysis of live CAR T cells demonstrates that they are capable of producing trails in the presence of fibroblast co-culture (Supp Fig. 7d, e). The engineered MSLN-CAR T cells were then adoptively transferred into NSG mice harboring AsPC-1-derived pancreatic tumors (Fig. 6a). To monitor T cell distribution in organs and infiltration into tumors, we labeled the engineered T cells with CellTracker Green dye and transplanted the cells on day 14. We harvested pancreatic tumor tissues on day 6 after T cell transplantation and were able to identify T cells expressing fluorescent protein with little movement, and also some cells generating distinct path history in tumors (following mCherry immunofluorescence staining to enhance signal; Fig. 6b–f). Along these lines, we note that, compared to the trails formed by breast carcinoma cells, path signals from T cells are less intense and often discontinuous, consistent with T cells expressing very low collagen levels and likely having a more limited capacity for fibrillar collagen assembly. The much faster migration speeds for T cells in PDA (vs. carcinoma cells43,44) are likely another factor for less continuous paths as rapidly sprinting T cells would have less ECM processing and deposition time in a given location43–45. However, while further optimization will be beneficial, we are able to observe path history for adoptively transferred therapeutic T cells in solid tumor environments. Human mCherry-Col1a1 + MSLN-CAR-T cells form trails in a grafted pancreatic tumor. a Schedule of cell transplantation and tissue harvest for AsPC-1 pancreatic adenocarcinoma cells subcutaneously xenografted into NSG mice, followed by subcutaneous infusion of (CellTracker Green dye-labeled for cell tracking) mCherry-Col1a1+ MSLN-CAR-T cells (mC1a1+ CAR-T cells) near the tumor. b–g Representative fluorescence images taken from frozen sections of AsPC-1 tumor 7 days after T cell transplantation. b, d, f Maximum intensity projection of z-stacks from sections stained with mCherry antibody. c, e, g Tracking single trails across z planes. White boxes in (b, d, f) indicate the regions of z plane analysis. The numbers 1–2 and 1–4 represent sample position in a trail. Bar=20 μm. Discussion While live cell and tissue imaging are extremely powerful and provide unique insights, information on cell migration deep within the body over long time periods is harder to obtain due to the technical challenges of imaging cells in opaque tissues within a moving body. Here, we propose a complementary strategy by examining cell migration trails formed from secreted fusion proteins of fluorescent protein and ECM. The trails record cell migration history in the body, allowing for information on cell movement from postmortem tissues. The applications of this approach are wide-reaching. It can be used to learn whether or not, or where cells moved in response to genetic perturbations or therapies. Likewise, as we demonstrate, it can be used to track cells following adoptive cell transfer therapies, which has the potential to add insight into regions where therapeutic cells do or do not infiltrate and move well. Furthermore, we studied the effects of cell type, trail density and fusion protein on trail identification. We demonstrated that the fusion protein-based strategy is versatile across mouse and human cells, including those that may not secrete large amounts of endogenous collagen I. Indeed, to our knowledge, this is the first migration history recording study in adult mouse models. Previous in vivo migration studies were based on live imaging during embryo development, e.g. transparent Zebrafish, chicken and Drosophila embryos in hours or days46–48. We directly record the migration history of transplanted cells in the opaque mouse organs for up to 9 weeks. As demonstrated, we can study cell migration patterns in the heart, muscles, primary breast tumors and metastatic organs in mouse models, suggesting the potential to track both tumor and therapeutic cell movements during disease progression and during experimental or therapeutic interventions. Additionally, the current studies do not rely on sophisticated transgenic animal models or intravital microscopy, and thus are straightforward and feasible approaches for in vivo cell migration studies. Perivascular trafficking routes of blood cells in the mouse spleen and invasion of primary human glioblastoma cells in human brain slices were studied previously using intravital microscopy or ex vivo tissue culture49,50. Perivascular cancer cell migration was also suggested in an angiotropism melanoma model and renal cell carcinoma via immunostaining51,52. In this study, we directly recorded complex trails near vessels where trails were often in parallel or multiple-helix shape (e.g., Fig. 3). In cancer progression, cancer cell migration along blood vessels is advantageous to promoting metastatic dissemination53,54, therefore the ability to track carcinoma cell – vascular interactions could be impactful. More importantly, the helix trails along vascular adventitia suggest physical mechanisms for cancer cell intravasation at vascular points. We note that our findings of cancer cell migration along vascular adventitia are based on an in vivo mouse model over the long term, suggesting that group cell migration along blood vessels may be an important route for breast cancer spread. Furthermore, perivascular trafficking is crucial for therapeutic T cells delivery to the target tissue. Previous intravital microscopy studies suggested complicated interactions between T cells and vascular wall cells, and T cells adopt an elongated shape for better migration55,56. Our trail studies of perivascular cells and T cells, thus, may be applied to evaluate the migration and infiltration of therapeutic cells in cancer tissues during immune cell therapies. The trails are not uniform, with different thicknesses and fluorescence intensities in different locations. There are several possible explanations for this behavior. This may be due to changes in cell migration speed and direction, especially if the cell secretes the fusion protein continuously and shows no fluorescence loss during in vitro migration. A slower migration may result in more collagen deposits and thus higher mCherry fluorescence intensity at the respective position of a trail. Cells also sometimes move back and forth along the same or very similar) path and groups of cells can move in a coordinated fashion via leader-follower dynamics9,43,57,58, all of which may influence FP-ECM presence. Likewise, tumor environments are notably harsh environments compared to normal tissues, with many proteases at play that can destabilize proteins, and heterogeneity of these factors in TMEs may influence local signaling. However, we note that we observe robust trails in both primary and metastatic tumors, but also note that specifically engineering the ECMs to be more protease resistance, such as has been demonstrated by mutating the collagenase cleavage site of collagen I59,60 may be beneficial for future interactions of the platform. Likewise, approaches to enhance collagen secretion and/or fiber formation may be beneficial for tracking T cells, which show markedly less signal than carcinoma cells, but remain detectable, albeit likely nearing the lower limit of detection. Alternatively, exploration of alternative ECMs to create fluorescent fusion proteins may be warranted to determine if a more homogeneous path can be formed for easier detection. We also note, however, that T cells are among the fastest-moving cells both in vitro and in vivo, with low collagen expression. As such, while we are likely reaching the limits of detection for our current systems, this also suggests that the system is robust enough to capture behavior from a very fast-moving cell type with low base collagen expression, suggesting that applicability to many cells that move more slowly and naturally produce more ECM is likely. We also observed groups of trails in addition to single trails. While our platform does not currently allow us to distinguish trails that are generated by different cells, the lack of interweaving and u-shaped features from cell turning around leads us to speculate that group trails are likely formed by a group of cells or cells migrating along adjacent paths over time, which would be consistent with cells following established ECM fiber alignment architectures in tumors43,61,62. However, it is also possible that turning etc. events happened in regions we did not image, suggesting that imaging very large 3D volumes could provide additional insights. Certainly, future approaches with mixed fluorophores and very large imaging volumes will likely provide new insights into the questions. Thus, while limitations exist, the work presented here provides a foundation for future studies to further parse out key cell behaviors. However, it appears clear that the current platform has great utility for studying cancer progression and cellular therapy dynamics. Additional future work to optimize fluorescent proteins and engineer both fluorophores and ECM to be maximally stable should further enhance the platform. Yet, our trail-based migration tracing method is a simple, readily applicable technology for studying cell migration in live animals and has the potential to bridge the gap between in vitro and in vivo cell migration. Methods Constructs A plasmid encoding EYFP-Fibronectin was a gift from Michael Davidson (Addgene plasmid # 56600; http://n2t.net/addgene:56600; RRID:Addgene_56600). Plasmids encoding GFP-mCol1a2(Genbank accession: MT525969), mCherry-mCol1a2 (Genbank accession: MT276428), mCherry-mCol1a1 (Genbank accession: PZ193668): were designed and synthesized in VectorBuilder Inc (Chicago, IL). To transduce T cells, lentiviruses encoding mCherry-mCol1a1 or mesothelin chimeric antigen receptor (MSLN-CAR, with scFv antibody SS sequence AAC04760.1) were designed and synthesized in VectorBuilder Inc (Chicago, IL)63. Cell culture Cell membrane associated tdTomato expressing cardiac cells were isolated from the heart of mTmG transgenic mice ordered from Jackson Laboratory (Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J, strain # 007576) by collagenase dissociation. Aortic smooth muscle cell line (Cat# PCS-100-012), MDA-MB-231 cells (Cat# CRM-HTB-26) were purchased from American Type Culture Collection (Manasses, VA). EO771 cells were ordered from CH3 BioSystems LLC (Cat# 94A001, Amherst, NY). To generate cells that stably express fusion proteins, the respective plasmid was transfected into adherent cells by magnetofection with CombiMag per manufacturer’s instructions (OZ Biosciences, San Diego CA), followed by selection with puromycin (1 µg/ml). The cells were cultured in full media following the manufacturer’s instructions, typically DMEM or RPMI 1640 with 10% fetal bovine serum. Mouse models All protocols were approved by the Institutional Animal Care and Use Committee in the University of Minnesota. We have complied with all relevant ethical regulations for animal use. Immunodeficient NSG mice (NOD.Cg-Prkdcscid/J, Strain #001303) and C57BL/6 (C57BL/6 J, Strain #000664), were purchased from Jackson Laboratory (Bar Harbor, ME). Cancer cell lines expressing the fusion proteins were embedded in Matrigel (Corning, NY) and directly injected into the fourth fat pad. Animals and tumors were monitored weekly until the endpoint (20 mm in any dimension or endpoint euthanization criteria). Primary and metastatic tumors were harvested at different time points to assess early trials and longest duration trials. Animals of both genders aged 8–14 weeks were randomly assigned to different time points. Four to twenty samples were harvested at different time points. All samples from live animals were included for data analysis For subcutaneous pancreatic cancer, AsPC-1 pancreatic carcinoma cells were embedded in Matrigel and subcutaneously injected into NSG mice (one million cells per mouse). To evaluate T cell migration in tumor tissues, human T cells were engineered to overexpress mesothelin chimeric antigen receptor (MSLN-CAR)64 and mCherry-mCol1a1. Briefly, T cells from human donors were co-transduced with lentiviruses of MSLN-CAR and mCherry-mCol1a1 after activation and expansion in vitro. Engineered T cells were embedded in Matrigel and transplanted into the subcutaneous pancreatic cancer mouse model through intraperitoneal injection at 5 million per mouse when tumor sizes are 5 mm in any dimension. Histologic samples Whole animals were perfusion fixed with 1% paraformaldehyde through the aorta for one hour, followed by isolation of tumors and nearby tissues and remote organs. Frozen tissues were sectioned at 60 μm in thickness for imaging and immunostaining. Briefly, fixed tissues were incubated in 30% sucrose at 4 °C overnight, then embedded in optical cutting temperature (OCT) compound and frozen in pre-cooled liquid nitrogen-isopentane. Frozen tissues were continuously sectioned at 60 μm in thickness and stored in −80 °C. At least four tissue sections from different positions (e.g., top, middle and bottom) were examined for each sample. Immunoprecipitation and Western blot To examine GFP-Col1a2 fusion proteins secreted in the cell media, 1.5 ml cell supernatant was mixed with rabbit anti-GFP antibody (1:1000, Cat# AB3080P, Millipore) or rabbit anti-mouse collagen Ia2 (1:1000, Cat# A21059, Abclonal Bio) and incubated in 4 °C overnight, followed by incubation with 20 μl protein A/G beads (cat#88802, Thermofisher Scientific, Waltham, MA) to precipitate the fusion protein. The pellet was washed with cold PBS and resuspended in 50 μl sample buffer for Western blot assays with the same antibodies at 1:1000 dilution. Fluorescent secondary antibodies, IR680-conjugated goat anti-rabbit and IR800-conjugated goat anti-mouse were ordered from LI-COR Biosciences (Lincoln, NE). Blots were scanned with an Odyssey CLX Imaging System. Immunofluorescence Frozen tissue sections were permeabilized with 0.3% Trition X-100 for 10 min and blocked with 1% BSA, 0.5% FBS in cold PBS for 60 min, followed by primary antibody, FITC-conjugated CD31 (Cat# RM5201, Thermofisher Scientific). Tissue sections were examined with a fluorescence microscope or confocal microscope specified in the figure legend, typically with a 20x or 40x objective. For secondary harmonic generation (SHG) imaging, tissue sections were examined with a multiphoton laser scanning microscope with a 20x water immersion objection65. In vitro cell migration For cell migration in vitro, mCherry-Col1A2+ and WT EO771 cells were cultured in DMEM medium with 10% FBS. Cells were cultured on Matrigel and imaged in channels for mCherry and Hoechst4322 dye (1 µg/ml Hoechst4322 dye) every 2 min. Time-lapse videos with 20 or more frames were used for identifying and tracking cells. Time-lapse videos were exported for image analysis after imaging and analyzed for migration paths, direction change rate, mean speed, and total travel distance. 3D cell culture 5000 mCherry-Col1a1+ MDA-MB-231 cells (mC+ MM231 cells) or 5000 human mCherry-Col1a1+ mesothelin CAR-T cells (mC1a1+ CAR-T cells) and 5000 NIH-3T3 cells were mixed with 50 µl collagen I solution composed of 3 mg/ml rat tail collagen I (cat#354236, Corning), 50 mM HEPES (final concentration) in 1x PBS at 100 cells/μl. Mixed cells were transferred to the center of a 24-well glass-bottom plate at 50 μl/well, and incubated in room temperature for 20 min and in 37 °C for 40 min for collagen polymerization. 1 ml DMEM medium with 10% fetal bovine serum, 100 μg/ml L-ascorbic acid 2-phosphate (cat# AST-H10792 25 G, AstaTech Inc) was added to each well thereafter. The cultured cells were examined daily with a fluorescence microscope, and imaged on day 10 with a 40x objective of a Zeiss fluorescence microscope and an Aptome3 to acquire 3D image stacks. Image stacks were imported into NIH ImageJ/Fiji package to measure trail fluorescence intensities and track trail directions of each cell. Image and statistical analysis Fluorescence images were analyzed manually using ImageJ (W. Rasband and contributors, National Institutes of Health). 3D Z-stack images were presented as maximal Z projections. For 3D reconstruction, trail segments in each Z-stack images were manually examined and copied to the trail reconstruction images. For cell migration analysis, time-lapse stacks were analyzed with ImageJ package and TrackMate plugin4, using LAP algorithm and the default settings. Fluorescence intensity along a cell trail was continuously measured with a circular section (with a diameter covering the trail width) moving along the trail. The measured fluorescence intensity was used for statistical analysis. Statistics and reproducibility Mice with mix genders and ages were randomly assigned to different groups. Sample numbers (4 to 20) were harvested and recorded for different groups. All samples from live animals were included for data analysis. T-tests or ANOVA were performed to compare values between two or among more groups, respectively. Tukey’s Honest Significant Difference (HSD) post-hoc tests were performed for multiple comparisons to find the significance between groups. p-value < 0.05 is considered significant. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Transparent Peer Review file Supplemental Information Description of Additional Supplementary files Supplementary data Reporting Summary Acknowledgements The authors acknowledge funding from the American Heart Association Innovative Project Award 18IPA34110189 to ZC, NIH National Cancer Institute grants U54CA268069 and P01CA254849 to PPP and BM, as well as useful discussions with Drs. Liang Liu and Robert Clarke. The authors acknowledge the use of the cell sorting and imaging core facilities at the University of Minnesota and the Hormel Institute, University of Minnesota. Author contributions Z.C.: Concept development, experimental design, methodology, data collection and analysis, investigation, writing original draft and funding acquisition. B.S.M.: Supervision, experimental design, validation, writing-review & editing, funding acquisition. P.P.P.: supervision, experimental design, methodology, data analysis, writing, reviewing and funding acquisition. Peer review Peer review information . Communications Biology thanks Oscar Maiques, Louisiane Perrin and Amanda Haage for their contribution to the peer review of this work. Primary Handling Editors: Eve Rogers and Mengtan Xing. A peer review file is available. Data availability Data is provided within the manuscript and in the Supplementary Data files. Additional datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Branden S. Moriarity, Paolo P. Provenzano. Contributor Information Branden S. Moriarity, Email: mori0164@umn.edu Paolo P. Provenzano, Email: pprovenz@umn.edu Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10488-2. References Associated Data Supplementary Materials Transparent Peer Review file Supplemental Information Description of Additional Supplementary files Supplementary data Reporting Summary Data Availability Statement Data is provided within the manuscript and in the Supplementary Data files. Additional datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.