Introduction

The dynamic interplay between resident and infiltrating cell populations within the tumor microenvironment (TME) is a key driver of cancer initiation, progression, and metastasis1,2,3. The TME comprises malignant cells, cancer-associated fibroblasts, endothelial cells, and a heterogeneous mix of immune cells, collectively forming the tumor-immune microenvironment (TiME). The TiME has emerged as a critical therapeutic target, underscored by the transformative impact of immune checkpoint inhibitors4. In parallel, tumor necrosis, often arising from chronic ischemia due to rapid tumor expansion, is a common feature of many solid tumors and is associated with poor clinical outcomes5,6,7. Given the complex and spatially heterogeneous nature of both immune cell infiltration and necrotic regions, there is a pressing need for robust imaging tools capable of comprehensively characterizing immunologically active tumor regions while also delineating the extent and margins of necrosis. Such tools are essential for advancing both preclinical research and clinical decision-making.

Molecular imaging tools provide longitudinal non-invasive readouts of cellular and sub-cellular events by employing specialized probes paired with powerful imaging modalities such as positron emission tomography (PET) or magnetic resonance imaging (MRI). Molecular MRI (mMRI) provides a promising platform for imaging the TiME due to its impressive spatial resolution, exquisite soft tissue contrast and anatomical detail from conventional 1H MRI, lack of signal-depth dependence, and a plethora of strategies to manipulate contrast. The value of mMRI tools is especially evident when orthogonal contrasts are acquired in the same subject, as in the case of multinuclear imaging with non-competitive imaging probes.

Fluorine-19 (19F) MRI is a well characterized mMRI tool that employs specialized radiofrequency hardware to image the in vivo 19F spin distribution upon perfluorocarbon (PFC) injection, producing images that are directly quantifiable due to a lack of 19F background signal in biological tissue8. Importantly, 19F MRI has been used to perform non-invasive imaging of the TiME in preclinical models of breast cancer9,10,11,12,13,14,15,16,17, brain cancer18, pancreatic cancer19, colorectal cancer9, head and neck squamous cell carcinoma20, and lung cancers21 through in situ labeling of immune populations. Ex vivo tissue analyses have primarily attributed the MRI signal to the uptake of PFC by tumor-associated macrophages (TAMs), a key myeloid cell population within the TiME13,14,15,16,17,18,19,20. Notably, an increased presence of TAMs has been associated with poorer clinical outcomes22.

Reporter genes for mMRI can offer additional value in visualizing the TiME; for instance, by providing spatiotemporal information on the distribution of engineered cancer cells in vivo. As an example, liver-derived rat organic anion transporter polypeptide 1A1 (rOatp1a1) was first described in 2014 as an mMRI reporter gene due to its ability sequester the clinically-approved liver-specific paramagnetic contrast agent gadolinium ethoxybenzyl-diethylenetriaminepentaacetic acid (Gd-EOB-DTPA; Primovist/Eovist) and generate strong positive T1 contrast of engineered cells23,24. We previously employed rOatp1a1 to map the intratumoral distribution of viable cancer cells within individual tumors, enabling improved delineation between necrotic and viable regions. A human-derived isoform of this transporter, OATP1B3, has also been used by others for preclinical cancer applications25,26. However, all these studies have been limited to imaging in immunodeficient mouse models due to immunogenicity concerns with current OATP reporter genes. A mouse-derived Oatp1 reporter gene would be highly valuable across many imaging applications in immunocompetent mouse models of cancer and beyond.

In this study, we developed murine Oatp1a1 (mOatp1a1) as a novel syngeneic mMRI reporter gene capable of imaging in immune-intact mouse models. By pairing mOatp1a1 expression with a mOatp1a1-transported fluorescent dye, we enabled enrichment of engineered cells via fluorescence-activated cell sorting (FACS) without introducing immunogenic transgenes. This system enabled in vivo visualization of dynamic intratumoral changes in regions both containing and lacking viable engineered murine breast cancer cells. We further integrated this approach with complementary 19F mMRI to detect immunologically dense regions within individual tumors, spanning both viable and necrotic compartments, as well as in secondary lymphoid organs. To analyze our imaging results, we performed multiplexed flow cytometry to profile PFC+ immune cells which indicated that 19F MRI signals extend beyond TAMs, reflecting a broader pan-myeloid imaging profile that, in line with previous reports19,27, may be tumor-type dependent. Additionally, we observed substantial 19F uptake by both myeloid and lymphoid cells in secondary lymphoid organs in tumor-bearing mice, highlighting important considerations for interpreting 19F signals in preclinical models.

Results

Engineering, enrichment, and MRI detection of mOatp1a1-expressing cells

Murine breast carcinoma cells (4T1) were transduced with the pEF1α-mOatp1a1 lentiviral vector, as illustrated in Fig. 1A. Consistent with the findings of Tachikawa et al.28, Sulforhodamine-101 (SR-101) was specifically internalized by Oatp1a1-expressing cells, enabling their enrichment by FACS (Fig. 1B). Selective uptake of SR-101 by sorted cells was further validated microscopy (Fig. 1C).

Fig. 1: Engineering and functional validation of mOatp1a1 expression for in vitro and in vivo molecular imaging.
Fig. 1: Engineering and functional validation of mOatp1a1 expression for in vitro and in vivo molecular imaging.The alternative text for this image may have been generated using AI.
Full size image

A Schematic representation of pEF1α-mOatp1a1 lentivirus used to transduce HEK-293T, EO771, and 4T1 cells. B Flow cytometric quantification of SR-101 uptake in mOatp1a1-expressing 4T1 cells, enabling enrichment via FACS. C Fluorescence microscopy evaluating SR-101 accumulation in mOatp1a1-expressing 4T1 cells and naïve controls. D R1 relaxation rates measured in 4T1 cell pellets post-incubation with Gd-EOB-DTPA. E Timeline of orthotopic bilateral 4T1 tumor model implantation and serial imaging. F Post-contrast T1-weighted MRI in mOatp1a1-expressing 4T1 tumors (orange contours) compared to naïve tumors (gray contours). G Tumor volumes as calculated from 1H segmentation at all timepoints. H Contrast-to-noise ratios (∆CNR) in mOatp1a1-expressing tumors following Gd-EOB DTPA injection. I Representative 19F MRI images following VS-1000H injection overlaid on 1H axial MR images. J Quantification of 19F spins contained in inguinal lymph nodes, spleen, and tumors (mOatp1a1-expressing and naïve). K Correlation analysis between tumor volume and 19F content. * p < 0.05, ** p < 0.01.

Longitudinal relaxation rate (R1) mapping of MRI images from mOatp1a1-engineered 4T1 (Fig. 1D) and HEK-293T, EO771 (Supplementary Fig. 2) cell pellets demonstrated an increase in R1 values in mOatp1a1-expressing cells incubated with Gd-EOB-DTPA compared to all other conditions.

An initial cohort of three mice received bilateral mammary fat pad injections of either naïve or mOatp1a1-expressing 4T1 cells and were imaged weekly pre- and post-Gd-EOB-DTPA injection once tumors became palpable at day 11 (Fig. 1E). Both tumors appeared isointense in the pre-contrast images, whereas post-contrast images revealed strong positive contrast enhancement in engineered tumors only (orange contours; Fig. 1F). T1 signal enhancement was uniform across the boundaries of the mOatp1a1-expressing tumors on days 12 and 19 post-tumor implantation, and the presence of regional hypointensities were visible on day 26 post-tumor implantation in all three mice. Tumor volumes were not different between naïve and mOatp1a1-expressing tumors at all timepoints (Fig. 1G). Pre-to-post-Gd-EOB-DTPA \(\Delta {CNR}\) values were increased for the mOatp1a1-expressing tumors compared to the naïve tumors (Fig. 1H).

At endpoint, 19F MRI was also performed following intravenous injection of 200 \(\mu L\) of VS-1000H, showing high uptake in both tumors, the spleen and the inguinal lymph nodes (Fig. 1I). 19F signal was observed at both the periphery of tumors and within necrotic regions in mOatp1a1-engineered tumors, and at the periphery of naïve tumors (not enough information was obtained from MR images to discern the location of the intratumoral 19F signal within these tumors). The number of 19F spins contained within each of the three tissues were quantified, revealing higher total VS-1000H uptake in the spleen compared to the inguinal lymph nodes, naïve tumor, and engineered tumor (Fig. 1J). Total 19F uptake in naïve and mOatp1a1-expressing tumors was not different (Fig. 1J). Correlational analysis revealed no relationship between tumor volume and 19F content (Fig. 1K). Histology revealed the presence of distinct viable and necrotic (decreased nuclear content and pale cytoplasmic staining) regions in both naïve and mOatp1a1-expressing tumors (Supplementary Fig. 3A, B). Areas of necrosis matched closely with regions of hypo-intensity in matched MR images for mOatp1a1-expressing tumors. The 4T1 carcinoma cell line is known to metastasize to the lungs29. Although lung metastases were not visible in MR images, lung histology revealed increased alveolar density and the presence of cancer cell infiltration (Supplementary Fig. 3C).

Longitudinal 1H/19F MRI monitoring of mOatp1a1-expressing tumors in mice

Tumor-free mice (n = 3) and mice bearing single Oatp1a1-expressing orthotopic 4T1 tumors (n = 11) were injected with PFC or Gd-EOB-DTPA and imaged with 19F MRI, T1w MRI and/or T2w MRI at various timepoints.

Three tumor-bearing mice were imaged weekly with 1H/19F MRI (Fig. 2A). On day 5, a small region of hyperintensity measuring <2 mm in diameter was seen in one of the three mice, despite this tumor not being palpable (Fig. 2B). By day 12, all three mice had MR-detectable tumors, and tumor volume increased by days 19 and 26 (Fig. 2C). Heterogenous 1H signal hyperintensities within tumors were seen as early as day 12 in one of the three mice, with all three mice showing heterogenous 1H signal by day 19. 19F signal was only visible on day 26, consistent with total 19F tumor content being higher by day 26 compared to day 12 (Fig. 2D). In the spleen, 19F uptake was observable at all timepoints, yet no differences in either 1H volume or 19F spin content compared to day 5 were observed (Fig. 2E, F). For ipsilateral and contralateral inguinal lymph nodes, 19F uptake was first seen on day 5 in one mouse, with all three mice showing 19F uptake in both lymph nodes at day 12. Ipsilateral lymph node volume was elevated compared to contralateral lymph nodes on days 12 and 19 (Fig. 2G) and 19F content was similar at all timepoints in this small cohort and no differences were observed (Fig. 2H).

Fig. 2: Longitudinal 1H/19F MRI of tumor-bearing mice.
Fig. 2: Longitudinal 1H/19F MRI of tumor-bearing mice.The alternative text for this image may have been generated using AI.
Full size image

A Experimental timeline for tumor-bearing BALB/c mice undergoing dual 1H and 19F MRI. B Representative MRI showing tumor progression from day 5 to day 26 with tumors denoted with a green arrowhead and inguinal lymph nodes denoted with a magenta arrowhead. C Quantification of tumor volume over time. D Quantification of tumor 19F content over time. E Quantification of spleen volume over time. F 19F content in the spleen over time. G Ipsilateral and contralateral inguinal lymph node volume changes across timepoints. H 19F content in inguinal lymph nodes. * p < 0.05, ** p < 0.01, *** p < 0.005. R denotes 19F reference sample position in MR images.

Pre- and post-Gd-EOB-DTPA T1- and T2-weighted images were taken at endpoint for the remaining 8 tumor-bearing mice. Both T1- and T2-weighted pre-contrast images appeared fairly isointense (Supplementary Fig. 4A). As expected, post-Gd-EOB-DTPA T1-weighted MR images showed increased signal within the tumor periphery, whereas post-contrast T2-weighted images showed reciprocal information, with the necrotic core of tumors appearing hyperintense relative to the tumor periphery. However, the magnitude of the \(\Delta {CNR}\) differed between the two contrasts. At the tumor’s edge, the mean \(\Delta {CNR}\) in the T1-weighted images of 131.3\(\pm\)15.3 was higher than the mean \(\Delta CNR\) of 45.5\(\pm\)8.2 in the T2-weighted images (Supplementary Fig. 4B). At the tumor’s core, the mean \(\Delta CNR\) of 21 in the T1-weighted images was not different from the mean \(\Delta {CNR}\) of 7.9 observed from the T2-weighted images. (Supplementary Fig. 4B).

The 19F content within tumors, spleens, and inguinal lymph nodes was analyzed for all 11 tumor-bearing mice imaged at endpoint (day 26 for n = 3 and day 23 for n = 8; Fig. 3A). One mouse exhibited little tumor and spleen growth, and its volume and 19F spin estimation data were excluded from subsequent analyses and figures. Spleens contained higher levels of 19F (average 3.9 × 1019 spins) compared to both inguinal lymph nodes (4.2 × 10¹⁷ spins) and mOatp1a1-expressing tumors (1.8 × 1019 spins), with tumors also exhibiting higher 19F content than lymph nodes (Fig. 3B). No correlation was observed between tumor volume and total 19F content (Fig. 3C).

Fig. 3: 19F content and spatial distribution in tumor-bearing mice.
Fig. 3: 19F content and spatial distribution in tumor-bearing mice.The alternative text for this image may have been generated using AI.
Full size image

A Experimental timeline for single timepoint tumor imaging after repeated PFC injection. B Quantification of 19F spins in spleens, inguinal lymph nodes, and mOatp1a1-expressing tumors at endpoint. C Correlation analysis between tumor volume and total intratumoral 19F content. D Frequency of 19F signal observed in distinct tumor regions (viable, periphery, necrotic core, and transition zone). E Representative 1H/19F image pairs illustrating spatial PFC signal distribution across tumors with increasing 19F content (scale bar = 4 mm). F, G Confocal images showing co-localization of VS-DM-Red with F4/80⁺ macrophages in viable and necrotic tumor regions (scale bar = 20 μm) PFC co-localized with F4/80+ cells (yellow arrows), F4/80+ cells lacking PFC (green arrows), PFC+ in F4/80- cells (white arrows). * p < 0.05, ** p < 0.01.

The majority (80%) of mOatp1a1 tumors exhibited visible 19F signal: 40% showed signal at the tumor periphery, 90% within the necrotic core, and 60% in the transition zone between viable and necrotic regions (Fig. 3D). By arranging 1H/19F image pairs by total intratumoral 19F content, spatial trends in PFC distribution became evident (Fig. 3E). Tumors with low 19F levels displayed dense signal within the necrotic core. In tumors with moderate 19F content, peripheral signal became apparent, sometimes extending beyond the tumor margin. In high content tumors, 19F signal was distributed across the necrotic core, periphery, and increasingly within the regions with viable mOatp1a1-expressing cancer cells.

Confocal microscopy revealed co-localization of Texas Red-conjugated PFC (VS-DM-Red) with F4/80, a pan-macrophage marker, particularly within viable tumor regions (Fig. 3F). However, not all F4/80+ cells were PFC-labeled, and not all PFC-labeled cells expressed F4/80. In necrotic regions, identified by a mix of intact and fragmented DAPI-stained nuclei, F4/80/PFC co-localization was less frequent, with PFC signal more often localized in apoptotic/dead cells with fragmented nuclei or acellular spaces (Fig. 3G).

1H/19F MRI of secondary lymphoid organs in naïve and tumor-bearing mice

19F MRI signal was detected in the spleens of both healthy and tumor-bearing mice, with markedly increased signal intensity and notable splenomegaly observed in the latter (Fig. 4A, B). Volumetric MRI analysis revealed that spleens were significantly larger in tumor-bearing mice compared to naïve controls at days 19 and 26, which was corroborated by ex vivo spleen weight measurements (Fig. 4C, D). Total 19F content within the spleens was significantly elevated in tumor-bearing mice relative to naïve mice (Fig. 4E), and a strong positive correlation was observed between 19F content and spleen volume (Fig. 4F). Histological analysis in the spleen revealed the presence of PFC-labeled F4/80+ macrophages; however, a substantial fraction of PFC-labeled cells did not express the F4/80 marker, suggesting the involvement of additional cell populations in PFC uptake (Fig. 4G).

Fig. 4: 19F MRI signal and volumetric analysis of spleens and lymph nodes in tumor-bearing and non-tumor-bearing mice.
Fig. 4: 19F MRI signal and volumetric analysis of spleens and lymph nodes in tumor-bearing and non-tumor-bearing mice.The alternative text for this image may have been generated using AI.
Full size image

A, B 19F MRI and ¹H/19F overlay showing increased spleen signal and splenomegaly in tumor-bearing mice. C Quantification of spleen volume by MRI in non-tumor-bearing and tumor-bearing mice D Ex vivo weight and E 19F content in spleens at day 26. F Correlation between spleen volume and 19F content. G Confocal imaging of spleen sections showing PFC co-localized with F4/80+ cells (yellow arrows), F4/80+ cells lacking PFC (green arrows), PFC+ in F4/80- cells (white arrows) (scale bar = 20 μm). H Representative 19F MRI signal in lymph nodes of non-tumor-bearing mice. I, J Comparison of ipsilateral vs. contralateral inguinal lymph nodes in tumor-bearing mice. K, L Comparison of lymph node volume and 19F content between tumor-bearing and non-tumor-bearing mice. M, N Ipsilateral inguinal lymph nodes in tumor-bearing mice show increased volume and 19F signal relative to contralateral nodes.

Tumor-free mice displayed PFC uptake into the inguinal lymph nodes (iLN), as well as variable uptake into renal, lumbar, sciatic, and cardiac lymph nodes (Fig. 4H). 1H volumes of the ipsilateral and contralateral inguinal lymph nodes of the tumor-bearing mice were compared to their side-matched counterparts in the non-tumor-bearing cohorts and revealed that only the ipsilateral inguinal lymph node displayed significantly increase volume (**p < 0.001; Fig. 4I). Interestingly, when comparing the 19F spin numbers in these same inguinal lymph node ROIs, no significant differences were found (ns; Fig. 4J). However, when comparing each of the ipsilateral and contralateral inguinal lymph nodes within the tumor-bearing mice, both the 1H volume and 19F spin number were statistically greater in the ipsilateral lymph node relative to the contralateral lymph node (**p < 0.001 and *p < 0.05, respectively) (Fig. 4K–N), suggesting increased drainage from the tumor contributing to increased PFC presence. H&E staining of lymph nodes revealed the increased presence of B cell follicles on only the ipsilateral LN, further suggesting the increased presence of lymphocytes in these tissues (Supplementary Fig. 5).

Multiplexed immunophenotyping of PFC+ cells in mOatp1a1-expressing tumors and secondary lymphoid organs

Based on our histological findings, specifically the presence of PFC+ cells lacking F4/80 expression, we performed multiplexed flow cytometry on tissues from mice injected with FITC-conjugated PFC (VS-DM-Green) to better characterize cellular uptake.

Tumors contained a higher proportion of myeloid cells compared to lymphoid cells with 10.4 ± 5.3% of myeloid cells and 1.6 ± 0.3% of lymphoid cells labeled with PFC Supplementary Fig. 6A). Among all VS-DM-Green⁺ CD45⁺ immune cells in the tumor, myeloid cells accounted for the vast majority (89.1 ± 0.1%), while lymphoid cells represented only 10.1 ± 0.1%\(\,(\)Supplementary Fig. 6B). Further phenotyping revealed that the predominant VS-DM-Green⁺ immune cell populations in tumors, in descending order, were neutrophils (26.2 \(\pm\) 7.8% of all labeled immune cells), Ly6C⁻ TAMs (TAMs; 18.4 \(\pm\) 8.9%), G-MDSCs (16.3 \(\pm\) 4.3%), Ly6C⁺ TAMs (12.6 \(\pm\) 6.4%), M-MDSCs (8.6 \(\pm\) 6.4%), other myeloid cells (6.9 \(\pm\) 1.9%), T cells (6.6 \(\pm\) 5.8%), and B cells (4.3 \(\pm\) 3.9%), with no differences in the abundance of neutrophils and Ly6C- TAMS (Fig. 5A and Supplementary Table 3).

Fig. 5: Flow cytometric analysis of immune cell populations and PFC uptake in tumors, spleens, and lymph nodes.
Fig. 5: Flow cytometric analysis of immune cell populations and PFC uptake in tumors, spleens, and lymph nodes.The alternative text for this image may have been generated using AI.
Full size image

A Composition of PFC⁺ CD45⁺ immune cells in tumors by subtype. B ΔMFI of PFC signal in dominant myeloid subsets in tumors. C Weighted signal contribution of myeloid subsets to tumor 19F MRI signal. D Composition of PFC⁺ immune cells in spleens. E ΔMFI of PFC signal in spleen immune subsets. F Weighted contribution to 19F MRI signal by spleen cell types. G Composition of PFC⁺ immune cells in lymph nodes. H ΔMFI of PFC signal in myeloid vs. lymphoid subsets in lymph nodes. I Weighted contribution to 19F MRI signal by lymph node cell types. To denote statistical differences, groups that share the same letter are not significantly different (p > 0.05), while groups with no shared letters are significantly different from one another (p < 0.05). \(\bigotimes\) indicates that the number of cells measured in the ΔMFI was below the desired threshold, and subsequent statistical tests were not able to be completed with these groups.

To assess VS-DM-Green uptake across immune cell types, we analyzed changes in mean fluorescence intensity (ΔMFI). Among the three most abundant VS-DM-Green⁺ populations in tumors, Ly6C⁻ TAMs exhibited higher ΔMFI compared to G-MDSCs (Fig. 5B), and no differences in ΔMFI compared to neutrophils. Due to low cell numbers, T cells (present above the 100-cell threshold in only two tumors) and B cells (below the threshold in all tumors) were excluded from ΔMFI statistical analysis.

While TAMs showed greater per-cell uptake of V-Sense than other myeloid populations, we next calculated a “weighted signal contribution” by combining ΔMFI with the relative abundance of each labeled population. This analysis revealed that neutrophils, Ly6C⁺ TAMs, Ly6C⁻ TAMs, M-MDSCs, and G-MDSCs contributed comparably to the overall 19F MRI signal, with no differences between any of the myeloid subpopulations, other than neutrophils and other myeloid cells (p < 0.05; Fig. 5C).

The spleen contained a higher proportion of lymphoid cells compared to myeloid cells, with an equivalent percentage of both cell types labeled with PFC (Supplementary Fig. 6C). Thus, unlike the tumor, where labeling was predominantly myeloid, the composition of PFC⁺ immune cells in the spleen was more balanced, with 55.9 ± 12.3% of labeled cells being myeloid and 44.1 ± 12.3% lymphoid in origin (Supplementary Fig. 6D). Among these, B cells (35 \(\pm\) 9.9% of all labeled immune cells), neutrophils (23.9 \(\pm\) 5.6%), G-MDSCs (20 \(\pm\) 9.3%), and T cells (9.1 \(\pm\) 4.9%) comprised over 85% of all PFC-labeled immune cells, with no differences in the abundance of labeled neutrophils and labeled B cells or labeled neutrophils and labeled G-MDSCs, however the abundance of labeled B cells was higher than that of labeled G-MDSCs (Fig. 5D & Supplementary Table 3).

Although neutrophils and G-MDSCs exhibited higher PFC uptake per cell, as reflected by greater ΔMFI, compared to B cells (Fig. 5E), the relative abundance of labeled B cells resulted in comparable overall contributions to the 19F MRI signal among all three populations (Fig. 5F).

Finally, in the ipsilateral lymph nodes of tumor-bearing mice, lymphoid cells represented the predominant immune cell population, with comparable percentages of labeled lymphoid and myeloid cells within their respective compartments (Supplementary Fig. 6E). In contrast to tumors and spleens, the majority of PFC⁺ immune cells in lymph nodes were of lymphoid origin (89.7 ± 8.1% lymphoid vs. 10.3 ± 8.1% myeloid; \(*\) p < 0.01), with T cells (51.2 ± 22.0%) and B cells (38.5 ± 18.2%) accounting for nearly 90% of all labeled immune cells (Supplementary Fig. 6F). The abundance of both labeled T cells and labelled B cells were greater than that of each labeled myeloid subpopulation (Fig. 5G and Supplementary Table 3). Some myeloid populations were below the 100-cell threshold for MFI calculation. All myeloid subsets above the 100-cell threshold exhibited higher PFC uptake per cell (ΔMFI) compared to lymphoid cells (Fig. 5H), but the high abundance of T and B cells in lymph nodes resulted in their dominant contribution to the overall 19F MRI signal in this tissue, though no differences were found (Figure 9I).

Discussion

Understanding the dynamic interactions within the TME is central to deciphering the mechanisms driving cancer progression and therapeutic response. Non-invasive imaging modalities capable of longitudinally capturing both cellular and acellular components of the TME offer transformative potential for studying these interactions in vivo. This is particularly important in immunocompetent models, where the interplay between tumor cells and immune populations critically shapes both disease trajectory and treatment efficacy. As immunotherapies become increasingly central in clinical oncology, there is a growing need for tools that can reliably visualize immune activity within the TME. In this study, we established a dual 1H/19F MRI platform for comprehensive, non-invasive profiling of the TME and tumor-immune microenvironment. Applied to a syngeneic breast cancer model, this platform enabled high-resolution, spatiotemporal mapping of viable tumor regions, necrosis, and diverse immune infiltrates, providing valuable insights into the structural and immunological heterogeneity of solid tumors and associated lymphoid organs.

Reporter gene imaging, particularly bioluminescence imaging (BLI), remains a widely used strategy to monitor tumor burden and treatment response, especially in immunodeficient models. Although some studies have explored BLI in immunocompetent hosts30, the use of non-self, immunogenic optical reporters such as green fluorescent protein (GFP) or firefly luciferase is known to elicit immune-mediated rejection of engineered cells and can alter immune composition within the TME31,32,33,34,35,36. Despite optimizations of transduction of mouse-derived cells with luciferase30, challenges remain in achieving consistent tumor engraftment and reproducible signal intensity37. To overcome the demonstrated immune rejection of optical reporter genes, transgenic reporter-tolerized genetically engineered mice (GEMs) have been used. For example, the “Glowing Head” (GH) GEM expresses a low amount of luciferase-GFP fusion protein in the anterior pituitary gland, so luciferase-GFP-expressing Lewis Lung Carcinoma (LLC) cells were no longer immunogenic in the GH mouse38; however, another study still detected rejection of luciferase-eGFP tumors in a 4T1 mammary tumor model33. The uncertainty surrounding different foreign reporter genes’ immunogenicity when applied to different tumor models is a significant challenge, even in reporter-tolerized GEMs; therefore, there is a need for immunotolerant reporter genes to track cells in numerous species, particularly mice, the most used laboratory animal. In contrast, our use of a “self” reporter through the mOatp1a1 transporter provides a non-immunogenic alternative that supports longitudinal imaging in immunocompetent settings. Unlike existing foreign reporters, this tool preserves the disease context by maintaining the tumor microenvironment, enabling more accurate testing of therapeutic efficacy in immunocompetent models and supporting improved clinical translation. While MRI requires greater technical expertise and infrastructure compared to optical modalities, and is more expensive, it delivers unparalleled spatial resolution and the ability to resolve intratumoral features such as necrosis and spatial heterogeneity in three dimensions, details that are difficult to discern with 2D-BLI. Moreover, the utility of such endogenous reporters likely extends beyond engineered tumor models to include applications in tracking gene- and cell-based therapies for a wide array of diseases.

19F MRI has emerged as a robust platform for immune cell imaging, with successful implementation following both in vitro39,40,41,42,43 and in situ13,14,15,16,17,18,19,20 PFC labeling. Clinically, 19F MRI has already demonstrated feasibility for visualizing ex vivo labeled dendritic cells and tumor-infiltrating lymphocytes (TILs) in humans44,45. In preclinical models, 19F signal within tumors is commonly attributed to tumor-associated macrophages (TAMs), typically identified via immunofluorescent staining of macrophage markers such as F4/80 or CD6813,14,15,16,17,20. For example, a study investigating KRAS-targeted immunotherapy in pancreatic cancer revealed organ-specific patterns of PFC accumulation, with CD11b⁺ monocytes and macrophages contributing to tumor signal and hepatocytes and endothelial cells responsible for liver signal19, highlighting the importance of local microenvironments in shaping PFC biodistribution. Our previous data in the 4T1 breast cancer model also pointed to TAM-dominated uptake13,14,15. However, immunostaining and fluorescence microscopy, while valuable, is inherently qualitative and can be confounded by signal bleed-through, non-specific binding, and challenges in identifying moderately PFC-loaded cells. Thus, relying solely on co-localization microscopy studies using TAM marker staining may lead to an incomplete or misleading interpretation of 19F MRI signal origin. Indeed, our microscopy studies here continued to support F4/80⁺ TAM enrichment for PFCs but also revealed substantial PFC localization in F4/80⁻ cells and acellular necrotic regions. While macrophages are strongly phagocytic, other immune populations, including neutrophils, dendritic cells, and peripheral blood mononuclear cells also demonstrate significant PFC uptake in vitro41,46,47.

To rigorously define the cellular contributors to 19F signal, we employed multiplexed flow cytometry, which provides quantitative, multivariate resolution of immune cell populations across tissue compartments. Our data in this 4T1 breast cancer model revealed broad PFC uptake across multiple myeloid subsets, including Ly6C+ and Ly6C- TAMs, neutrophils, and MDSCs, as well as detectable but low-level uptake in CD45⁺/CD11b⁻ lymphoid cells. When factoring in both labeling intensity (\(\Delta\)MFI) and population abundance, we estimate that tumoral 19F signal arises from a relatively balanced contribution of these myeloid subsets. In secondary lymphoid organs, signal heterogeneity was even greater, with neutrophils, G-MDSCs, and B cells comprising the majority of splenic signal, and both T and B cells contributing to tumor-draining lymph nodes. Only one other recent study has conducted such in-depth flow cytometry analyses following systemic PFC administration18. In a brain tumor model, Croci et al. showed that PFC uptake was largely confined to microglia-derived TAMs, with minimal uptake in blood-derived TAMs, lymphoid cells (CD45⁺/CD11b⁻), or neutrophils. Our findings of more broad myeloid labeling and imaging align with a smaller subset of 19F MRI literature in other disease contexts48,49 and suggest that systemic PFC delivery could act as a pan-myeloid imaging strategy in tumors capable of broadly delineating areas of dense innate immune cell infiltration.

The inclusion of the mOatp1a1 reporter gene further enhanced our ability to interpret tumor architecture by offering readouts of cancer cell viability and distribution. Unlike conventional T1- or T2-weighted images, which offer limited specificity, mOatp1a1 imaging provided complementary information that helped contextualize 19F signal sources. For example, our MRI data indicate substantial 19F signal colocalization with tumor necrotic cores, raising the possibility of PFC presence in acellular spaces. Fluorescent nuclear staining supported this, revealing PFC presence in both fragmented DAPI⁺ cells and DAPI⁻ regions. These findings underscore the need for more rigorous assessment of PFC localization in necrotic tissue and its implications for interpreting 19F signal, including a separate measurement of 19F in viable and necrotic compartments.

The flow cytometry analysis presented in this work suffered from some limitations that warrant comment. Firstly, the fluorinated perfluorocarbon agent synthesized by CelSense is no longer commercially available, making future studies limited by available agent. However, the synthesis of other perfluorocarbons is being done in-house by other research groups9,10,11,18,50 and similar methodologies will allow for the continuation of these types of imaging investigations. Secondly, the ex vivo 19F analysis was restricted to CD45⁺ cells only. While this allowed for comprehensive immune profiling, it leaves open the question of whether non-immune stromal components, such as cancer cells, fibroblasts, or endothelial cells, also accumulate PFCs following intravenous injection. Future studies will be important to clarify these contributions and validate the full range of cellular sources of in vivo 19F signal. Furthermore, given the short lifespans of immune subtypes such as neutrophils, the single timepoint flow cytometry presented here may not have captured all the crucial information regarding the dynamics of PFC uptake and retention in the TME. Further studies will be needed to observe how these relative labeled cell distributions appear under changing conditions.

Although limited to preclinical use, the dual-MRI system described here addresses a key need in immunotherapy development by enabling longitudinal monitoring of immune characteristics. Clinically, the classification of tumors as “hot” or “cold” traditionally hinges on T cell infiltration and response to immune checkpoint inhibitors (ICIs), but growing evidence points to a critical role for myeloid cells in modulating immunotherapy outcomes51. PET tracers that target T cell activity are being actively developed to stratify immune-hot tumors52; however, few imaging agents are currently capable of robustly assessing myeloid cell content within the TME, let alone with MRI. Our findings suggest that 19F MRI, which broadly labels multiple myeloid subsets, could serve as a powerful tool to identify “myeloid-hot” versus “myeloid-cold” tumors, providing complementary information to lymphoid-focused imaging approaches. This capability could be particularly impactful in the context of integrated PET/MRI systems, which offer the unique opportunity to combine PET-based localization of lymphoid cells with MRI-based anatomical and myeloid immune profiling. The integration of whole-body PET/MRI represents a scalable strategy for comprehensive immune imaging that overcomes the spatial limitations of ex vivo methods such as spatial transcriptomics and proteomics. As myeloid-targeted therapies advance and the interplay between myeloid and lymphoid compartments becomes increasingly recognized as a determinant of immunotherapy response51, the ability to non-invasively distinguish myeloid-rich from myeloid-poor tumors may significantly enhance patient stratification and therapeutic decision-making. Collectively, these imaging innovations hold promise for unraveling the cellular and structural complexity of the TME in both preclinical models and patients, enabling the rational design, monitoring, and personalization of next-generation immunotherapies.

With myeloid-targeted therapies now being actively pursued and the role of myeloid cells in modulating lymphoid-targeted therapies well established51, the ability to distinguish myeloid-rich from myeloid-poor tumors, beyond simply identifying lymphoid-rich (immune-hot) tumors, may enhance our ability to predict which immunotherapies are most likely to succeed. The potential combination of the non-invasive whole-body mMRI tools described here with other TME imaging tools, such as lymphoid-targeted PET52, should begin to complement massively multiplexed analysis of limited ex vivo samples (e.g., spatial transcriptomics/proteomics), to truly unravel the complexity of TME dynamics both in preclinical models and patients.

Methods

Cell lines

Human embryonic kidney (HEK-293T, ATCC Cat# CRL-3216, RRID:CVCL_0063), and murine mammary carcinoma cells (EO771, ATCC Cat# CRL-3461, RRID:CVCL_GR23 and 4T1, ATCC Cat# CRL-2539, RRID:CVCL_0125) were obtained from the American Type Culture Collection (ATCC) and maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% (v/v) fetal bovine serum (FBS) and 1% (v/v) antibiotic-antimycotic. All cell lines were routinely tested for mycoplasma contamination using the MycoAlert™ Mycoplasma Detection Kit (Lonza, Basel, Switzerland) and confirmed to be negative.

Cloning and lentiviral production

Using the In-Fusion HD Cloning (Takara Bio USA, Inc.), the mOatp1a1 transgene (Horizon Discovery Biosciences, plasmid #3061794 G) was cloned into a third-generation lentiviral transfer vector driven by the human elongation factor 1 alpha (pEF1α) promoter, generating the LV-pEF1α-mOatp1a1 construct. The integrity of the final plasmid was confirmed by whole-plasmid sequencing.

pEF1α-mOatp1a1 LVs were produced in-house by co-transfecting HEK-293T cells with the LV-pEF1α-mOatp1a1 plasmid, the packaging plasmids pMDLg/pRRE (Addgene #12251) and pRSV-Rev (Addgene #12253), and the envelope plasmid pMD2.G (Addgene plasmid #12259) using Lipofectamine 3000, following the manufacturer’s instructions (Thermo Fisher Scientific, MA, USA). Viral supernatants were collected at 24- and 48-h post-transfection and stored at −80 °C until use.

Lentiviral cell engineering and enrichment

HEK293T, EO771, and 4T1 cells were transduced with pEF1α-mOatp1a1 LV in the presence of polybrene (8 μg/mL; Sigma Aldrich, MI, USA) for 6 h at 37 °C. After replacing the media, cells were cultured for an additional 48 h. Since mOatp1a1-expressing cells take up SR-10128 we developed a cell sorting protocol based on SR-101 dye uptake. mOatp1a1 engineered cells were sorted using a FACSAria III cell sorter (BD Biosciences, CA, USA) following incubation (1 μM for 20 min at 37 °C) with the fluorescent dye Sulforhodamine-101 (SR-101; Cayman Chemical, Item # 16953; 9-(2,4-disulfophenyl)-2,3,6,7,12,13,16,17-octahydro-1H,5H,11H,15H-xantheno [2,3,4-ij:5,6,7-i’j’]diquinolizin-18-ium). SR-101 uptake was also evaluated via fluorescence microscopy with a Revolve 4 ECHO microscope (Discover ECHO Inc., San Diego, CA, USA) and flow cytometry using a FACSCanto flow cytometer (BD Biosciences, CA, USA).

MRI agents

A biologically compatible 19F PFC tracer composed of a perfluoropolyether nanoemulsion called VS-1000H (also called V-Sense) was used (Celsense Inc.; Pittsburgh, PA, USA)47. VS-1000H contains an average droplet size of 145 nm, containing 20–30% volume fraction of perfluorocarbon in a buffered solution. For fluorescence microscopy and flow cytometry, two additional formulations of PFC conjugated to Texas-Red (VS-DM-Red; Excitation/Emission 590 nm/620 nm) or Fluorescein isothiocyanate (VS-DM-Green; Excitation/Emission 488 nm/506 nm) fluorophores were also used (Celsense Inc.; Pittsburgh, PA, USA). Gd-EOB-DTPA (0.25 mmol/mL) was obtained from Bayer HealthCare.

MRI of cell pellets

Naïve or sorted mOatp1a1-expressing 4T1 breast cancer cells (1 × 106) were seeded in 15-cm cell culture dishes and expanded until >90% confluency was reached. Cells were then incubated with or without Gd-EOB-DTPA (1.6 mmol/L) in 20 mL of media for 60 min. Cells were washed 3 times with PBS, and pelleted cells of known number were placed in a sample holder filled with 1% agarose gel and arranged into a single row within a birdcage imaging coil tuned to 127.728 MHz53. Cell samples were heated to 37 °C, and imaged at 3 Tesla (General Electric Healthcare Discovery MR750 Scanner) using a fast-spin echo inversion recovery (FSE-IR) pulse sequence with the following imaging parameters: matrix size = 400\(\times\)80, repetition time (TR) = 5000 ms, echo time (TE) = 19.1 ms, echo train length (ETL) = 4, number of excitations (NEX) = 2, receiver bandwidth (rBW) = 125.00 kHz, inversion times (TI) = 50, 68, 94, 128, 175, 239, 327, 447, 612, 836, 1144, 1564, 2139, 2925, 4000, in-plane resolution = 0.3 × 0.3 mm2, slice thickness = 2.0 mm, acquisition time = 3 min and 45 s per inversion time.

Longitudinal relaxation rates (R1) were estimated by performing non-linear least-squares fitting of signal intensity at the each of the inversion times listed above, on a pixel-by-pixel basis, using a custom MATLAB application (MATLAB, MathWorks, Natick, Massachusetts, United States), and the following Eq. 1., where \({M}_{{ss}}\) is the final steady state magnetization, and \({M}_{i}\) is the magnetization at the first value of the inversion recovery curve, which is independent of the final steady state magnetization \({M}_{{ss}}\).

Equation 1. Estimating longitudinal relaxation times using inversion recovery

$$S=\left|{M}_{{ss}}-\,\left({M}_{{ss}}-{M}_{i}\right){\cdot e}^{-{TI}/{T}_{1}}\right|$$
(1)

Mouse models

All in vivo experiments were conducted in accordance with the guidelines of the Canadian Council on Animal Care and approved under an animal use protocol (AUP#2022-190) by the Animal Use Subcommittee of Western University’s Council on Animal Care.

To assess the function of the mOatp1a1 reporter in vivo, bilateral orthotopic 4T1 mammary fat pad tumors were established in BALB/c mice (Charles River, Canada). Subjects were female mice, aged 6–8 weeks, weighing approximately 20–23 g by endpoint. Mice were injected with 250,000 naïve 4T1 cells in the left mammary fat pad and 250,000 mOatp1a1-engineered 4T1 cells in the right (n = 3). Tumor growth was monitored weekly using Vernier calipers. Once tumors became palpable (~day 10), pre- and post-contrast 1H MRI were performed weekly, with imaging conducted before and 5 h after tail-vein injection of Gd-EOB-DTPA (1.5 mmol/kg; Primovist, Bayer HealthCare). MRI was performed on days 11/12 (pre- and post-Gd-EOB-DTPA), 18/19, and 25/26 post-cell injection. On day 25, mice were injected with 200 µL of VS-1000H (V-Sense, Celsense Inc., Pittsburgh, PA) via tail-vein injection and imaged 24 h later using 1H/19F MRI for both Gd-EOB-DTPA and 19F PFC uptake.

Next, tumor-free BALB/c mice (n = 3) and BALB/c mice bearing a single mOatp1a1-engineered 4T1 orthotopic tumor (n = 3) were imaged longitudinally on days 5, 12, 19, and 26 post-cell implantation. Mice were imaged with dual 1H/19F MRI 24 h after intravenous injection of 100 μL VS-1000H and 5 h after administration of Gd-EOB-DTPA (2–2.5 mmol/kg) at each timepoint, with the final imaging timepoint occurring after a 200 μL VS-1000H injection. Finally, another cohort of mice bearing a single mOatp1a1-engineered 4T1 orthotopic tumor (n = 8) were injected with 100 μL VS-1000H on days 4, 11, 18 post-cell delivery. On day 22, these mice were imaged with T1- and T2-weighted 1H and 19F MRI prior to being injected with 200 μL of either VS-1000H (n = 2), VS-DM-Red for fluorescence microscopy (n = 1), or VS-DM-Green for flow cytometry (n = 5). Mice were then imaged again 24 h post-PFC and 5 h post Gd-EOB-DTPA injections with T1- and T2-weighted 1H and 19F MRI on day 23.

At experimental endpoint, animals were humanely euthanized using isoflurane overdose with cervical dislocation as a secondary method.

In vivo MRI

Animals were anaesthetized using isoflurane gas (3% induction, 2% maintenance in oxygen). All in vivo images were acquired on a 3 Tesla General Electric Healthcare Discovery MR750 Scanner. Purpose-built radiofrequency hardware for highly sensitive full body 19F detection in deep tissue was used for the entirety of the imaging53.

Anatomical T1-weighted images were acquired using a 3D-Spoiled Gradient Recalled Acquisition in the Steady State (3D-SPGR) with the following image parameters: FOV = 12.0 × 3.0 × 3.0 cm, 94 slices, TR/TE = 14.7 ms/2.456 ms, rBW = 62.5 kHz, matrix size = 400 \(\times\) 100 \(\times\) 100, Flip Angle = 60o, NEX = 4, voxel size = 0.3 mm isotropic, scan time ~9 min. A subset of mice was scanned with a T2-weighted 3D-fast spin echo (General Electric proprietary name CUBE) pulse sequence with FOV = 12.0 × 6.0 × 6.0 cm, 84 slices, TR/TE = 2341 ms/122 ms, rBW = 62.5 kHz, matrix size = 400 × 200 × 200, Flip Angle = 90o, NEX = 3, voxel size = 0.3 mm isotropic, echo train length = 120, scan time ~13 min. Fluorine-19 images were obtained using a balanced steady state free precession (bSSFP) sequence, FOV 12.0 \(\times\) 3.0 \(\times\) 3.0 cm, 60 slices, TR/TE = 7.3 ms/2.732 ms, flip angle = 72 degrees, 63 signal averages, matrix size of 100 × 100, voxel size 1 mm, scan time ~ 15 min.

\(\Delta {CNR}\) of both naïve and mOatp1a1-expressing tumors was calculated using Eq. 2, where the SDNoise refers to the standard deviation of the noise measured in the complex image, in a region with no NMR signal present. In the single-tumor-bearing mice that were scanned with both a T1 and T2 weighted pulse sequence, \(\Delta {CNR}\) was calculated for two regions of tumors: the tumor core and the tumor edge, as defined on the post-contrast T1-weighted MR images.

Equation 2. Identifying change in contrast-to-noise between tumors and muscle

$$\Delta {CNR}={\left(\frac{{{Signal}}_{{Tumour}}-{{Signal}}_{{muscle}}}{{{SD}}_{{Noise}}}\right)}_{{Post}}-{\left(\frac{{{Signal}}_{{Tumour}}-{{Signal}}_{{muscle}}}{{{SD}}_{{Noise}}}\right)}_{{Pre}}$$
(2)

Organs of interest for 19F spin analysis (the tumor, spleen, and inguinal lymph nodes) were segmented manually on post-contrast T1-weighted in Horos and regions-of-interest (ROIs) were exported as .csv files. For analysis of tumors at endpoint, three independent observers generated their own ROIs based on the post-contrast 1H images to help reduce bias in reporting 19F quantification in noisy images where tumor margins are more difficult to distinguish. A custom MATLAB script was developed in-house to process the 1H/19F image sets for analysis by employing a Rician-biased noise correction54,55, image registration, and total 19F spin estimation using Eq. 3.

Equation 3. 19F Spin number estimation within an ROI

$${19\atop}F\,{Spins}\,{in}\,{ROI}=\dfrac{{{Total}\,{Signal}}_{{ROI}}}{{{Total}\,{Signal}}_{{Reference}}}\times \dfrac{{known}\,{19\atop}F\,{spins}}{\mu L\,{of}\,{Reference}}\times {Segmented}\,{Volume}\,{of}\,{Reference}$$
(3)

Tissue collection

Immediately after endpoint MRI, mice were sacrificed and tumors, spleens, lungs, and inguinal lymph nodes (iLN) were harvested. The primary tumor and spleens were weighed before being sliced into two sections, half for staining (immunofluorescence or H&E) and half for flow cytometry. Tumors were sliced parallel to the skin surface to facilitate matching to MR images. Lymph nodes were used for either staining or flow cytometry given their small size.

Histology and immunofluorescence staining

Tissues were fixed overnight in 4% paraformaldehyde (PFA) in PBS at 4 °C before being processed by the Molecular Pathology Core Facility at Robarts Research Institute, Western University. Tissues underwent paraffin-embedding and H&E staining or OCT embedding and cryo-sectioning following a sucrose gradient (overnight incubation in 10%, 20%, and 30% sucrose solutions in 1X PBS) for immunofluorescence staining. Frozen tissue sections were thawed in PBS at room temperature, permeabilized with 0.25% (v/v) Triton X-100 in PBS for 10 min, and subsequently blocked in a solution containing 5% goat serum, 1% BSA, and 0.05% Tween-20 (v/v) in PBS for 1 h at room temperature. Sections were then incubated overnight at 4 °C with primary antibodies (1:150 dilution in blocking buffer). Following three PBS washes, samples were incubated with an Alexa Fluor™ 647-conjugated goat anti-rat secondary antibody (Abcam, ab150159) for 1 h at room temperature. Finally, slides were counterstained with DAPI-containing mounting medium (Abcam, ab104139).

Flow cytometry

Harvested tissues were placed in cold PBS. Single cell suspensions were generated by first dicing tissues using a scalpel blade into pieces <1 mm and pushed through a 70-μm cell strainer, before being centrifuged at 500 × g for 5 min. Cell suspensions from spleens were resuspended in 5 mL of warm RBC lysis buffer for 5–7 min at room temperature. Cell lysis was terminated by adding 10 mL of ice-cold PBS. Samples were incubated with FcX blocker for 10 min (Biolegend, product # 422301), and a True-Stain Monocyte Blocker was added (Biolegend, product # 426101). The anti-mouse antibodies used for phenotyping are in Supplementary Table 1, with single positive and fluorescence-minus-one samples for each antibody prepared for compensation and gating, respectively. Antibodies were added for 30 min at room temperature.

Data was analyzed in FlowJo software (Tree Star) to identify cell phenotypes in Supplementary Table 2 using the gating strategy in Supplementary Fig. 1. Briefly, to assess VS-DM-Green uptake in immune subpopulations, tumors, spleens, and inguinal lymph nodes from mice treated with the non-fluorescent PFC (VS-1000H) were first analyzed as controls. Fluorescence intensity thresholds were set at two standard deviations above the mean fluorescence of this control group, and this gate was used to define the VS-DM-Green–positive population within each immune phenotype. All stained cell samples were analyzed on the BD FACSymphony™ A5 Cell Analyzer (BD Biosciences, CA, USA).

The change in mean fluorescence intensity (MFI) \(\Delta {\rm{MFI}}\) was calculated using Eq. 1, where \({({{MFI}}_{{VS}-{DM}-{Green}+})}_{{cell\; type}}\) was measured using the gating strategy described in Supplementary Fig. 1, and \({({{MFI}}_{{VS}-1000H})}_{{cell\; type}}\) was measured on the individual cell population without a V-Sense fluorescent gate. Labeled cell populations with <100 cells were disregarded from the MFI calculation to avoid skewing the calculation due to small cell populations.

Equation 1. Changes in mean fluorescence intensity within labeled immune subpopulations

$$\Delta {\rm{MFI}}={({{MFI}}_{{VS}-{DM}-{Green}+})}_{{cell}\,{type}}-{({{MFI}}_{{VS}-1000H})}_{{cell}\,{type}}$$
(4)

Statistics

Comparisons between two groups were performed using unpaired two-tailed t-tests. For comparisons involving more than two groups or multiple factors, one-way or two-way ANOVA was used as appropriate, followed by Tukey’s or Šídák multiple comparisons tests. Repeated measures ANOVA was used for longitudinal analyses. Relationships between continuous variables were evaluated using simple linear regression and Pearson correlation. Statistical analyses were performed using GraphPad Prism version 10.4.2 and significance was set at p < 0.05.