Optical imaging reveals chemotherapy-induced metabolic reprogramming of residual disease and recurrence

Abstract

Fewer than 20% of triple-negative breast cancer patients experience long-term responses to mainstay chemotherapy. Resistant tumor subpopulations use alternative metabolic pathways to escape therapy, survive, and eventually recur. Here, we show in vivo, longitudinal metabolic reprogramming in residual disease and recurrence of triple-negative breast cancer xenografts with varying sensitivities to the chemotherapeutic drug paclitaxel. Optical imaging coupled with metabolomics reported an increase in non-glucose-driven mitochondrial metabolism and an increase in intratumoral metabolic heterogeneity during regression and residual disease in resistant MDA-MB-231 tumors. Conversely, sensitive HCC-1806 tumors were primarily reliant on glucose uptake and minimal changes in metabolism or heterogeneity were observed over the tumors' therapeutic life cycles. Further, day-matched resistant HCC-1806 tumors revealed a higher reliance on mitochondrial metabolism and elevated metabolic heterogeneity compared to sensitive HCC-1806 tumors. Together, metabolic flexibility, increased reliance on mitochondrial metabolism, and increased metabolic heterogeneity are defining characteristics of persistent residual disease, features that will inform the appropriate type and timing of therapies.

Fig. 1.. Resistant MDA-MB-231 tumors show metabolic plasticity over their therapeutic life cycle.

(A) MDA-MB-231 xenografts were treated with paclitaxel under a maximum dose density regimen. Mitochondrial metabolism (TMRE), glucose uptake (2-NBDG), and fatty acid uptake (Bodipy FL C16) were imaged in vivo. Xenografts showed an initial response to paclitaxel, a period of residual disease, and a resurgence in tumor volume at ~60 days after drug withdrawal. (B) Mean tumor volumes (mm³) where D0 indicates the start of treatment. Error bar, SE (n = 9). (C) Representative images of primary (P) (n = 10), early regressing (ER) (n = 10), late regressing (LR) (n = 10), residual disease (RD) (n = 10), and recurring (Rec) (n = 13) MDA-MB-231 tumors. Ridge plots across all pixels and all treated mice show (D) an increase in mitochondrial metabolism, (E) a decrease in glucose uptake, and (F) a slight increase in fatty acid uptake at key time points during RD compared to the primary tumor (P < 0.05). Representative images of (G) mitochondrial metabolism, (H) glucose uptake, and (I) fatty acid uptake of untreated (n = 5) and treated (n = 10) tumors every 2 days up to the tumor burden limit. Ridge plots of probe uptake across all pixels of all untreated mice at each time point show (J) an increase in mitochondrial metabolism at one time point (P < 0.05: D8 versus D0), (K) no significant temporal changes in glucose uptake (P > 0.05), and (L) an increase in fatty acid uptake at three time points (P < 0.05: D0 versus D10, D14, and D18). Statistical differences in probe uptake were determined using a KS test. Vertical dashed lines superimposed on each curve correspond to the average fluorescence. For bracketed time points, each individual time point is significantly different from the primary tumor (D0). The representative image is consistent with trends seen in the average fluorescence across all mice at each time point.

Fig. 2.. Sensitive HCC-1806 tumors are metabolically inflexible over their therapeutic life cycle.

(A) HCC-1806 xenografts were treated with paclitaxel under a maximum dose density regimen when tumor sizes reached ~150 mm³. Mitochondrial metabolism (TMRE), glucose uptake (2-NBDG), and fatty acid uptake (Bodipy FL C16) were imaged in vivo. Sensitive xenografts showed no evidence of residual disease up to 120 days after treatment withdrawal. (B) Mean tumor volumes (mm³) where D0 indicates the start of paclitaxel treatment. Error bar, SE (n = 19). (C) Representative images of primary (P) (n = 10), early regressing (ER) (n = 10), late regressing (LR) (n = 10), and residual disease (RD) (n = 10) HCC-1806 tumors. Ridge plots of (D) mitochondrial metabolism, (E) glucose uptake, and (F) fatty acid uptake across all pixels and all treated mice at each time point. Representative images of (G) mitochondrial metabolism, (H) glucose uptake, and (I) fatty acid uptake of untreated (n = 5) and treated (n = 10) tumors every 2 days up to the tumor burden limit. Ridge plots of probe uptake across all pixels and all mice at each time point show no significant changes in (J) mitochondrial metabolism, (K) glucose uptake, or (L) fatty acid uptake (P > 0.05) in untreated mice. Statistical differences in probe uptake were determined using a KS test. Vertical dashed lines superimposed on each curve correspond to the average fluorescence. An asterisk, “*,” signifies statistical significance between two time points joined by the corresponding line. The representative image is consistent with trends seen in the average fluorescence across all mice at each time point.

Fig. 3.. Imaging and metabolomics confirm increased mitochondrial metabolism during regression and residual disease in resistant MDA-MB-231 tumors.

(A) Mean tumor volumes (mm³) where D0 indicates the start of paclitaxel treatment. MDA-MB-231 xenografts showed an initial response to paclitaxel, followed by residual disease, and tumor recurrence ~60 days after paclitaxel treatment, while sensitive HCC-1806 xenografts showed sustained responses to paclitaxel with no palpable tumor at 120 days. Error bar, SE (n = 9 MDA-MB-231; n = 19 HCC-1806). Representative images show (B) mitochondrial metabolism (TMRE₆₀), (C) glucose uptake (2-NBDG₆₀), and (D) fatty acid uptake (Bodipy₆₀) of primary (P) (n = 10), early-regressing (ER) (n = 10), late-regressing (LR) (n = 10), residual disease (RD) (n = 10) HCC-1806 and MDA-MB-231 tumors, and recurring (Rec) MDA-MB-231 tumors (n = 13). Ridge plots show the probability density distribution of (E) mitochondrial metabolism, (F) glucose uptake, and (G) fatty acid uptake across all pixels and all mice at each time point. Statistical differences in probe uptakes were determined using a KS test. Vertical dashed lines superimposed on each curve correspond to the average fluorescence. (H) Volcano plot of fold change in metabolite levels between regressing and primary MDA-MB-231 tumors (n = 5 per group). (I) Volcano plot of fold change in metabolite levels between regressing and primary HCC-1806 tumors (n = 5 per group). Dashed lines represent P < 0.05 via a two-sided t test and | Log₂FoldChange | > 0.05. Significantly different metabolites are labeled within the graph. (J) Bar plot of change in pyruvate levels between regressing and primary time points across MDA-MB-231 and HCC-1806 tumors (n = 5). Metabolite levels were normalized to account for batch effects across cell lines. Statistical differences in pyruvate levels were determined using a one-way ANOVA followed by Tukey’s post hoc test. The representative image is consistent with trends seen in the average fluorescence across all mice at each time point.

Fig. 4.. Resistant MDA-MB-231 tumors show increased intratumoral metabolic heterogeneity compared to sensitive HCC-1806 tumors.

(A) Representative images of primary (P) (n = 10), early regressing (ER) (n = 10), late regressing (LR) (n = 10), residual disease (RD) (n = 10), and recurrence (Rec) (n = 13) for mitochondrial metabolism (TMRE₆₀), glucose uptake (2-NBDG₆₀), and the corresponding mitochondrial metabolism/glucose uptake (TMRE₆₀/2-NBDG₆₀) cluster mask in MDA-MB-231 tumors. (B) Representative images of primary (P) (n = 10), early regressing (ER) (n = 10), late regressing (LR) (n = 10), and residual disease (RD) (n = 10) for mitochondrial metabolism (TMRE₆₀), glucose uptake (2-NBDG₆₀), and the corresponding mitochondrial metabolism/glucose uptake (TMRE₆₀/2-NBDG₆₀) cluster mask in sensitive HCC-1806 tumors. (C) Bar graphs showing mean changes in area fraction [cluster percent (%)] of cluster distributions corresponding to [2-NBDG_High/TMRE_Low] clusters, [2-NBDG_High/TMRE_High] clusters, or [2-NBDG_Low/TMRE_High] clusters across primary, regression, residual disease, and recurrence (if applicable) for each tumor line. Statistical differences in cluster percent were determined using a one-way ANOVA followed by Tukey’s post hoc test.

Fig. 5.. Resistant HCC-1806 tumors show elevated intratumoral metabolic heterogeneity compared to sensitive HCC-1806 tumors.

HCC-1806 xenografts were treated with paclitaxel under a maximum dose density regimen when tumor sizes reached ~150 mm³. Xenografts showed a heterogeneous response to treatment, where some mice recurred and some did not. TMRE, 2-NBDG, and Bodipy FL C16 uptake were measured in vivo 60 min after injection in day-matched tumors that had high (recurred) or low (sensitive) tumor burdens. (A) Mean tumor volumes (mm³) where D0 indicates the start of paclitaxel treatment. Error bar, SE (sensitive HCC-1806 tumors, n = 19; resistant HCC-1806 tumors, n = 11). (B) Representative images of mitochondrial metabolism (TMRE₆₀), glucose uptake (2-NBDG₆₀), and fatty acid uptake (Bodipy₆₀) for day-matched sensitive and resistant HCC-1806 tumors. Ridge plots show probability density distribution of (C) mitochondrial metabolism, (D) glucose uptake, and (E) fatty acid uptake across all pixels and all mice at each time point (n = 3, day-matched sensitive HCC-1806 tumors; n = 3, day-matched resistant HCC-1806 tumors). Vertical dashed lines superimposed on each curve correspond to the average fluorescence. (F) Scatterplots showing pixel-by-pixel distribution of glucose uptake versus mitochondrial metabolism (2-NBDG₆₀ versus TMRE₆₀ intensity) across sensitive HCC-1806 tumors and resistant HCC-1806 tumors. (G) Representative images of the representative TMRE₆₀/2-NBDG₆₀ cluster mask for day-matched sensitive and resistant HCC-1806 tumors. (H) Bar graphs show changes in area fraction [cluster percent (%)] across sensitive HCC-1806 tumors and resistant HCC-1806 tumors. Statistical differences in probe uptakes were determined using a KS test. Statistical differences in cluster percent were determined using a one-way ANOVA followed by Tukey’s post hoc test. The representative image is consistent with trends seen in the average fluorescence across all mice at each time point.

Fig. 6.. Approach for manual clustering for analyzing intratumoral metabolic heterogeneity.

Schematic of clustering analysis using TMRE₆₀ and 2-NBDG₆₀ endpoints to visualize spatial clusters within individual images. Manual clustering was performed using mean probe uptakes as cutoff values for assignment into four major clusters.