Info
Background:
Blood flow within tumors is highly variable, due to the irregular organization and incomplete functionality of the tumor vasculature. Although this is well recognized, short-term temporal dynamics within these vascular networks remain poorly characterized, as many imaging studies and clinical assessments are performed as single baseline measurements or over intervals spanning days to weeks (1).
Hypothesis/Objective:
We hypothesize that physiologically relevant fluctuations in tumor microvascular perfusion can occur on an hour-scale and can be resolved using super-resolution ultrasound localization microscopy (ULM), thereby providing insights into temporal variability relevant for biomarker interpretation, and therapy delivery.
Methods:
Longitudinal in vivo ULM imaging was performed in two breast cancer xenograft models (n = 5/group) including estrogen receptor-positive MCF-7 and patient-derived triple-negative BCSC1 (Fig. 1A). Under continuous anesthesia, tumors were imaged repeatedly at an identical imaging plane over 9 hours, following the intravenous injection of poly(n-butyl cyanoacrylate)-based microbubbles (PBCA-MB) at 0h, 3h, 6h, and 9h. Ultrasound data was analyzed using a ULM framework (2), in which localized and tracked microbubble signals were used to reconstruct super-resolution vascular parameter maps and assess microvascular dynamics over time. To exclude systemic confounding effects during prolonged imaging, contrast-enhanced ultrasound (CEUS) measurements were acquired in adjacent muscle tissue. Finally, endpoint histological validation of vascular architecture, perfusion, and vessel maturity was performed.
Results:
Pronounced intra-tumoral heterogeneity and temporally evolving microvascular dynamics were consistently observed within individual tumors over the 9-hour imaging period, indicating substantial fluctuations in tumor perfusion at identical tumor locations. Due to prolonged anesthesia, 6 out of 10 animals reached the final 9-hour measurement time point. Despite reduced animal availability at later measurements, observed fluctuations remained consistent across tumors and time points, supporting the reproducibility of the observed perfusion dynamics.
Of the analyzed microvascular parameters analyzed, vessel coverage emerged as the most variable parameter, with deviations from tumor-specific mean values ranging from -67% to +173% in BCSC1 and -47% to +60% in MCF-7 tumors (Fig. 1B,D), indicating substantial fluctuations in perfused vessel area over time. Similarly, distance-to-closest track, a parameter reflecting the spatial organization of perfused vessels, showed strong variability over time, indicating dynamic changes in vascular accessibility and perfusion patterns (Fig. 2A,C). Vessel tortuosity additionally exhibited intra-tumoral heterogeneity over time, suggesting temporally changing microbubble trajectories within the reconstructed vascular networks (Fig. 2B). In contrast, functional parameters such as microvascular flow velocity remained comparatively stable, with deviations of approximately -6% to 8% across tumors and time points (Fig. 1C, 2D). Despite this pronounced intra-tumoral variability, neither consistent temporal trends, nor subtype-specific differences were detected at the group level, indicating heterogeneous and non-uniform microvascular dynamics across individual tumors. Endpoint histological analysis (Fig. 1E) further revealed heterogeneous vascular architectures with variable vessel density, perfusion, and maturity states across tumors and demonstrated that only a subset of vessels (~40-45%) was functionally perfused. This is consistent with the dynamically changing perfusion patterns observed by ULM and validates perfusion estimates derived from vessel coverage measurements.
Finally, to demonstrate that the observed fluctuations were tumor-intrinsic rather than systemically driven, stable muscle perfusion during CEUS measurements was confirmed at the individual time points (Fig. 2E).
Conclusion:
Longitudinal ULM demonstrated that the tumor microvasculature represents a continuously changing rather than a constant state. These findings suggest that, in tumor vasculature that is already characterized by incomplete baseline perfusion, short-term fluctuations in vascular accessibility may further influence perfusion dependent imaging measurements and therapeutic efficacy. This is particularly important for therapies with short circulation times. Active consideration of these temporally variable vascular dynamics may improve the interpretation of perfusion imaging biomarkers, therapeutic timing, and the design of preclinical and clinical intervention strategies.
Image/Figure:
Click to view full size
Click to view full size
Image/Figure Caption:
Longitudinal assessment of tumor perfusion heterogeneity and microvascular dynamics in breast cancer xenografts.
(A) Schematic overview of the longitudinal imaging workflow. Female mice with estrogen receptor-positive (ER+) MCF-7 xenografts or patient-derived triple-negative breast cancer (TNBC) BCSC1 xenografts (n = 5/group) underwent continuous anesthesia over a 9-hour imaging period. Poly(n-butyl cyanoacrylate) (PBCA)-based microbubbles (MB) were administered intravenously every 3 hours at 0h, 3h, 6h, and 9h, followed by repeated contrast-enhanced ultrasound (CEUS) and B-mode acquisitions at an identical imaging plane. The acquired ultrasound videos capturing microbubble movement within the tumor vasculature were analyzed using ultrasound localization microscopy (ULM), in which localized and tracked microbubble trajectories were reconstructed into super-resolution vascular maps.
(B) Characterization of microvascular vessel coverage, representing the relative perfused vessel area, revealed pronounced intra-tumoral and temporally evolving variability across individual tumors. It emerged as the most variable parameter among all analyzed ULM-derived metrics, showing substantial temporal instability in tumor perfusion over only a few hours. Despite this variability, linear mixed-effects modeling did not reveal consistent temporal or tumor subtype-specific trends at a group level.
(C) Analysis of microvascular flow velocity of MBs revealed comparatively lower temporal variability than vessel coverage, while still demonstrating heterogeneous intra-tumoral fluctuations over time. As the least variable ULM-derived parameter, flow velocity indicated a greater stability of blood flow dynamics despite high perfusion-related variability. Despite temporal changes within individual tumors, linear mixed-effects modeling also showed no consistent temporal or tumor subtype-specific trends across the monitored time points.
(D) Representative ULM-derived vessel coverage maps of an MCF-7 tumor acquired at 0, 3, 6, and 9 hours. Maps illustrate temporally changing perfused vascular regions within the same tumor and imaging plane. White dashed boxes indicate representative regions shown at a higher magnification below, highlighting local temporal fluctuations in perfusion. Scale bars: top row: 2mm; magnified regions: 200 µm.
(E) Endpoint histological validation of vascular architecture and functionality. (i) Vessel area fraction (%) (ii) fraction of perfused vessels relative to total vessels (%) (iii) fraction of mature vessels (%), reflecting vascular stabilization and perivascular coverage, together with (iv) representative fluorescence microscopy images of the tumor vasculature. Histological analysis revealed heterogeneous vascular architectures with variable vessel density, maturity, and perfusion states across tumors. MCF-7 tumors showed trends toward slightly higher vessel area and perfused vessel fractions compared to BCSC1 tumors, whereas the relative fraction of mature vessels remained comparable between the two tumor models. Notably, only a subset of vessels (~40-45%) was functionally perfused, consistent with ULM-derived vessel coverage measurements. Scale bar: 200 µm.
Author
Alexandra Raab
RWTH Aachen University