Therapeutically reprogrammed nutrient signaling enhances nanoparticulate albumin bound drug uptake and efficacy in KRAS-mutant cancer

Abstract

Nanoparticulate albumin-bound paclitaxel (nab-paclitaxel, nab-PTX) is among the most widely prescribed nanomedicines in clinical use, yet it remains unclear how nanoformulation impacts nab-PTX behavior in the tumor microenvironment. Here, we quantified the biodistribution of the albumin carrier and its chemotherapeutic payload in optically cleared tumors of genetically engineered mouse models, and compared the behavior of nab-PTX with other clinically-relevant nanoparticles. We found that nab-PTX uptake is profoundly and distinctly affected by cancer-cell autonomous RAS signaling, and RAS/RAF/MEK/ERK inhibition blocked its selective delivery and efficacy. In contrast, a targeted screen revealed that IGF1R kinase inhibitors enhance uptake and efficacy of nab-PTX by mimicking glucose deprivation and promoting macropinocytosis via AMPK, a nutrient sensor in cells. This study thus shows how nanoparticulate albumin bound drug efficacy can be therapeutically improved by reprogramming nutrient signaling and enhancing macropinocytosis in cancer cells.

Albumin fuels macropinocytosis-driven catabolism in cancer cells¹, and past reports have found that tumor macropinocytosis is regulated by signaling through activated RAS², as well as numerous other environmental cues, cell-surface proteins, and signaling molecules including RAC/PAK1^3,4. In part to take advantage of this behavior, as well as the prevalence of albumin-binding glycoproteins gp60 on endothelial cells and SPARC in tumor tissues⁵, albumin formulation has been used to improve tumoral drug delivery⁶, especially in cancers with constitutively-activating RAS mutations. Nab-PTX (Abraxane) is the first FDA-approved chemotherapy incorporating albumin into its formulation⁵, is the highest selling nanomedicine in past five years^7,8, and is used in treating diseases including metastatic non-small cell lung adenocarcinoma (NSCLC)⁹ and metastatic pancreatic ductal adenocarcinoma (PDAC)¹⁰, which are malignancies with high rates of KRAS mutation¹¹. Nonetheless, the impact of albumin formulation on efficacy has been mixed in clinical trials¹², and there has been little direct, mechanistic evidence of how such albumin-nanoformulation strategies affect the distribution, cellular uptake, and activity of drug payloads within the spatially heterogeneous tumor microenvironment (TME) in vivo.

The distribution and uptake of albumin are governed by multiple competing processes throughout the body, and it has been difficult to understand what the true rate-limiting processes are in determining the ultimate fate and action of nab-PTX. For example, SPARC has been proposed as a biomarker¹³, but Ph3 trials¹⁴ and genetic experiments proved disappointing¹⁵. Alternatively, tumor associated macrophages (TAMs) accumulate nab-PTX¹⁶, but their relative importance compared to cancer cell uptake remains unclear. It therefore is still uncertain what role cancer-cell autonomous macropinocytosis plays in nab-PTX delivery.

Single cell quantification of nab-PTX biodistribution

To address these issues, here we applied confocal microscopy and tissue clearing techniques to image the delivery and cellular uptake of nab-PTX in mouse models of KRAS-mutant NSCLC and PDAC at a single-cell resolution. Nab-PTX is based on the high-pressure homogenization of paclitaxel (PTX) with albumin, forming ~130nm nanoparticles. We developed a two-color imaging strategy based on AlexaFluor555-labeled albumin and a fluorescent taxane payload (SiR-taxane; Fig. 1a), yielding a fluorescently-labeled lab-made nab-PTX formulation with size, charge, characteristics of disaggregation, morphology, in vivo drug target, biodistribution, and tumoral accumulation similar to that of commercially available pharmacy-grade nab-PTX (Fig. S1). PTX stabilizes microtubules, and SiR-taxane fluorescence enhances 10-fold upon microtubule binding¹⁷, thus reporting target engagement in vivo. We co-imaged nano-albumin (nab-A555) and its payload (SiR-taxane) in the KP1.9 cell line derived from the Kras^G12D/+p53^−/− mouse model of lung adenocarcinoma¹⁸. In vitro, cellular uptake of nab-A555 correlated with SiR-taxane on a cell-by-cell level, not observed when albumin and SiR-taxane were added separately (Fig. 1c and S2a), therefore suggesting that payload activity depends on nab formulation uptake. We imaged the in vivo distribution of nab-PTX using the orthotopic KP1.9 model in syngeneic C57BL/6 hosts. Optical clearing with CUBIC reagents did not diminish the fluorescent intensity of nab-PTX (Fig. S2b), but enabling quantification through whole tumor-bearing lungs (Fig. 1b)¹⁹. Nab-PTX uptake on a cell-by-cell level spatially correlated between nanoparticulate albumin and its taxane payload (Fig. 1b–c, S2c). Nab-PTX accumulated 5-fold higher in tumors compared to non-tumor lung, and a large portion of uptake was in cancer cells (Fig. 1d–e). A Cremophor-EL formulation of SiR-taxane (as used with clinical solvent-based sb-PTX formulations) injected separately from A555-labeled albumin (Alb-A555) showed minimal tumor-selective uptake and no correlation between albumin and SiR-taxane (Fig. 1c, Fig. S2c–d), likely because Cremophor-EL is known to reduce taxane albumin binding²⁰. Thus, accumulation of nab-PTX directs the selective uptake of its taxane payload in orthotopic lung tumors.

Figure 1: Nab-PTX accumulates in cancer cells distinctly from other nanoformulations.

a. Schematic for imaging the albumin carrier (nab-A555) and the drug payload (SiR-taxane) of nab-PTX (albumin from Protein Data Bank accession 1E7I). b. Representative confocal microscopy of nab-PTX in lungs of C57BL/6 mice bearing KP-GFP tumors, 24 hr post-intravenous administration, at both whole-lung and single-cell levels (left and right scale bars=2 mm and 10 μm, respectively). Blue, yellow, and white lines outline lung, tumor, and cancer cells, respectively. Arrow highlights SiR-taxane microtubule binding. c. Albumin and PTX uptake were quantified by microscopy and correlated across individual cells either in vitro or from tumor-bearing lungs as in b (n=413 cells across 4 conditions, means ± 95% confidence interval (C.I.), two-tailed t-test for Pearson’s correlation). sb-PTX=solvent-based PTX. d-e. As in b, nab-PTX uptake in tumor tissues (d; coefficient of variance, C.V.=47%) or GFP-expressing cancer cells (e; C.V.=74%) was quantified by microscopy (means ± s.e.m. from n=6 lungs). f. Confocal images (scale bar=100 μm) and quantification (n=397 cells from 5 lungs) of co-dosed nab-PTX and PLGA-PEG nanoparticles (NP) in the KP model at 24 hr. g. Confocal microscopy (see Fig. S3) quantification of nab-PTX and model NPs (PLGA-PEG NP, cyclodextrin nanoparticle (CDNP), ferumoxytol magnetic nanoparticle (MNP), liposome (Lip), and PEGylated liposomal Dox (Doxil)) in cancer cells versus macrophages in orthotopic (ortho.), subcutaneous (s.c.), or lung metastasis (lung) mouse tumor models. NSCLC=non-small cell lung cancer, PDAC=pancreatic ductal adenocarcinoma, ATC=anaplastic thyroid cancer. Data are means ± s.e.m. from n=12–300 cells from n=3 tumors from n=3 mice per condition. Two-tailed t-test is used for all comparisons. For b and f, images are representative from 4 independent experiments with similar results.

We next examined the degree to which nab-PTX delivery merely reflected the general process of passive macromolecular accumulation via factors related to the enhanced permeability and retention (EPR) effect. We compared the distribution of nab-PTX with that of a co-administered model PLGA-PEG polymeric nanoparticle (PLGA-PEG NP) with similarities to clinical PTX formulations such as genexol-PM, and modest correlation was observed on a tumor-by-tumor basis (Fig. S2e). This suggests that nab-PTX tumor uptake is still governed somewhat by non-specific effects including EPR and possibly endothelial transcytosis, the latter known to mediate intratumoral extravasation of albumin⁵ and nanoparticles²¹. Yet, surprisingly, the overlap between nab-PTX and PLGA-PEG NP uptake on a single-cell level was minimal (Fig. 1f). TAMs play substantial roles in nanomaterial accumulation, and we observed that nab-PTX can indeed accumulate in TAMs, as found previously¹⁶. Furthermore, cancer cells and macrophages accounted for ~92±4% (mean ± 95% C.I. from n=4 tumors) of nab-PTX uptake in the tumor microenvironment. However, nab-PTX was taken up at highest levels in cancer cells, and in contrast, other model nanoformulations — including polymeric micelles, liposomes, and dextran-based NPs including the FDA-approved ferumoxytol — accumulated at highest levels in TAMs (Fig. 1g and S3a–g). This finding was corroborated in vitro with nab-PTX, which disaggregates to particles of ~10 nm diameter upon dilution (Fig. S1a), showing higher uptake in cancer cells compared to macrophages. By contrast, polymeric nanoparticles (~70 nm diameter), as well as nab-PTX that is covalently cross-linked to prevent disaggregation (~160 nm diameter), were taken up at a higher level in macrophages (Fig. S3h and i). These findings suggest that the size of the therapeutics, among other properties, may play a role in determining their relative uptake by cancer cells and macrophages, consistent with prior observations²².

MAPK/ERK signaling controls nab-PTX uptake by cancer cells

Prior literature has implicated macropinocytosis in the uptake of albumin by RAS-mutant cancer cells², and we found that the macropinocytosis inhibitor 5-(N-ethyl-N- isopropyl)amiloride (EIPA) eliminated nab-PTX uptake in KP1.9 cells (Fig. 2a). Punctate subcellular co-localization of nab-PTX with fluorescent dextran, dequenched BSA, and endolysosomal markers RAB5A, RAB7A, and LAMP1 further supported uptake via macropinocytosis and subsequent catabolic degradation in cells (Fig. 2b and S4a–d). Further imaging suggested that taxane from nab-PTX dissociated from albumin vehicle before microtubule target engagement (Fig. S4e–f). Nab-PTX uptake was 5 to 25-fold greater in RAS-mutant cancer cells compared to wild-type RAS comparators (Fig. 2c and S5a), and uptake correlated between albumin and its drug payload (Fig. S5b and c). Transient KRAS^G12D over-expression in KRAS-wt BxPC3 PDAC cells, and doxycycline induction of Kras^G12D expression in genetically engineered mouse tetO-Kras^G12D p53^−/− PDAC (iKras) cells, both enhanced nab-PTX uptake by ≥8-fold (Fig. 2c and S5d–e). High uptake was also seen in BRAF-V600E mutant cells, and both BRAF and KRAS mutation are known to stimulate downstream ERK phosphorylation. Hence, we used a transgenic kinase translocation reporter (ERK-KTR) to monitor ERK activation (Fig. 2d)²³, which revealed correlation between nab-PTX uptake and ERK activity on a cell-by-cell level (Fig. 2e and S6a–c). Inhibition of ERK signaling using either the clinical MEK1/2 inhibitor trametinib (Mekinist), or doxycycline withdrawal to silence upstream Kras^G12D, both blocked nab-PTX uptake (Fig. 2e and f). Similarly, inhibiting ERK signaling using MEK1/2 or ERK1/2 inhibitors reduced the cancer cell uptake of 70 kDa dextran (Fig. 2g and S6d–g). Macropinocytosis inhibition with EIPA, and MEK1/2 inhibition with trametinib, also reduced the uptake of covalently cross-linked nab-PTX (Fig. S6h). These data collectively suggest ERK signaling controls macropinocytic uptake of nab-PTX by cancer cells.

Figure 2. RAS signaling drives nab-PTX uptake by cancer cells.

a. Representative fluorescence microscopy (scale bar=10 μm) and quantification (means ± s.e.m., one-way ANOVA/Tukey compared to DMSO) of 4 hr nab-PTX uptake by KP cells treated with EIPA for 6 hr (n=228 cells across all conditions). b. Images (scale bar=5 μm, white line outlines cancer cell, white arrow points to co-localization) and pixel-by-pixel quantification (n=30 cells, Pearson’s correlation R, means ± 95% C.I., two-tailed t-test) of co-localization between 70 kDa dextran signals and nab-PTX in KP cells in vitro. c. Quantification of 4 hr nab-PTX uptake in cancer or epithelial cell lines (n=2305 cells across all conditions; means ± s.e.m., one-way ANOVA/Holm compared to MCF10A or two-tailed t-test as specified). d-f. Schematic of ERK-KTR imaging (d) and representative images (e, iKras; scale bars=10 μm) and corresponding averaged (f, n=923 cells across all conditions, means ± 95% C.I.) quantification of ERK-KTR and 4 hr nab-PTX uptake in cells treated for 24 hr with trametinib (tram) and/or doxycycline (doxy). iKras cells were treated with doxy except for “-doxy”, denoting 5 day doxy withdrawal. g. Quantification of the 70 kDa dextran (Dex) uptake, as a measurement of macropinocytic activity², in iKras cells treated with EIPA, MEK inhibitors (tram=trametinib, cobi=cobimetinib, refa=refametinib), and ERK inhibitors (MK=MK-8353, SCH=SCH772984). Data are means ± s.e.m. from n=909 cells across all conditions (one-way ANOVA/Tukey compared to DMSO).

We next examined if inhibiting RAS/RAF/MEK/ERK signaling could impact the accumulation of nab-PTX in vivo. Subcutaneous and intravenous inoculation of iKras cells into doxycycline-fed syngeneic C57BL/6 hosts led to tumor formation in the flank and lungs, respectively, and the latter is a frequent site of PDAC metastasis. Selective tumor accumulation of nab-PTX was eliminated as the result of RAS signaling inhibition via doxycycline withdrawal (Fig. 3a–c and Fig. S7a–d; imaged by confocal microscopy and/or fluorescence reflectance imaging, FRI), which correlated with decreased ERK signaling (Fig. S7e) and decreased uptake of co-administered albumin (Fig. S8a). In contrast, doxycycline withdrawal neither changed nab-PTX in non-tumor tissues (Fig. 3d), tumor vascularization (Fig. 3e and S8b–c), nor tumoral accumulation of a crosslinked dextran-NP (Macrin NP)²⁴ or PLGA-PEG NP (Fig. 3f–g and S8d–e), suggesting cancer-cell autonomous effects. In vitro doxycycline withdrawal decreased nab-PTX uptake (Fig. 2f) and increased the concentration at which nab-PTX inhibited 50% cell count (IC₅₀) by >30-fold (Fig. S9a). Similar effects were seen with trametinib: simultaneous nab-PTX and trametinib treatment (Co-tx) increased IC₅₀ by >30-fold (Fig. 4a, c, and S9b–f). This increase can largely be attributed to reduced cellular uptake of nab-PTX, rather than impacts on cell proliferation that may affect sustained paclitaxel activity as an anti-mitotic, since trametinib had minimal impact on nab-PTX IC₅₀ when it was dosed sequentially after nab-PTX treatment (Sep-tx, Fig. 4a, c, and S9b–f). Clinically, nab-PTX is frequently used in combination with gemcitabine, and trametinib increased the IC₅₀ of a nab-PTX+gemcitabine combination (Fig. S9e). No difference in IC₅₀ was observed when solvent-based PTX was used or across single-agent controls using the different dose timings (Fig. 4b–c and S9g–j). In vivo, both spontaneously developed autochthonous lung tumors and subcutaneous xenografts accumulated less nab-PTX when pre-treated with trametinib (Fig. 4d–e, S10a, and S11a–b), which translated to reduced efficacy: simultaneous treatment with nab-PTX and trametinib was less effective at blocking tumor growth compared to a regimen in which trametinib was administered after nab-PTX (Table S1, Fig. 4f–g and S12), but no difference was found in the same comparison but using sb-PTX (Table S2 and Fig. S13a–f). Consistent with the selective tumoral uptake of nab-PTX compared to sb-PTX (Fig. 1d and S2d), nab-PTX was more efficacious in blocking tumor growth in vivo compared to sb-PTX (Fig. S13h), while this difference was not observed in vitro (Fig. S13g). Finally, nab-PTX accumulation in non-tumor organs (Fig. 4e, S10b–c, and S11b), as well as perfused tumor vasculature and accumulation of dextran-NP in tumors (Macrin NP, Fig. S10d–g and S11c–e), did not decrease with trametinib, suggesting reduced nab-PTX uptake was due to a tumor-cell intrinsic change, rather than changes in the ability of macromolecular materials to reach the tumor. Overall, these results demonstrate that the signaling state of cancer cells can substantially impact nab-PTX delivery, therefore making drug response sensitive to the order in which nab-PTX and trametinib are dosed, which could have implications to clinical trials combining MEK1/2 inhibitors and nab-PTX (trametinib+nab-PTX: NCT01192165, ibrutinib+nab-PTX+gemcitabine: NCT02436668).

Figure 3. Cancer cell RAS signaling drives in vivo tumor accumulation of nab-PTX.

Upon formation of iKras lung metastases or subcutaneous (s.c.) tumors, C57BL/6 mice were withdrawn from doxy for either 2 or 5 days, treated with fluorescent nab-PTX, dissected, and imaged 24 hr post-treatment. a-b. Representative confocal microscopy (a, scale bar=2 mm) and quantification (b) of nab-PTX uptake in the iKras lung tumor. c. Representative fluorescence reflectance images (FRI, left, scale bar=5 mm) and quantification (right) of tumor and liver uptake of nab-PTX in the iKras s.c. tumor model. d. 24 hr nab-PTX uptake was combined across all non-tumor organs. e-f. Lectin (e) and dextran-NP uptake (f) were quantified in lung tumors. g. Co-injected with nab-PTX in c, PLGA-PEG NP was i.v. injected into mice bearing iKras s.c. tumors and quantified 24 hr later by FRI to measure tumor and liver uptake. For c, d and g, data are means ± s.e.m. from n=3 mice per group (one-way ANOVA/Tukey compared to +doxy). For b and e, data are means ± s.e.m. from n=180 lung tumors from n=3 animals per condition (two-tailed t-test). For f, data are means ± s.e.m. from n=804 tumors from n=8 animals across two conditions (two-tailed t-test). For b, e, and f, each data point represents a tumor. For a, c, and g, images are representative from 3 independent experiments with similar results.

Figure 4. MEK1/2 inhibition via trametinib reduces nab-PTX uptake and efficacy.

a-c. Treatment schedule and viability of iKras cells receiving either simultaneous (Co-Tx) or sequential (Sep-Tx) treatment of trametinib (tram) and nab-PTX (a) or sb-PTX (b) in vitro, measured by cell count at 72 hr. IC₅₀ values are compared across treatments (c, IC₅₀ value displayed in each bar). d. 129S mice bearing Kras^G12D/+p53^−/− autochthonous lung tumors were pre-treated with tram or vehicle control, followed by i.v. nab-PTX injection, and 24 hr later analyzed by CUBIC clearing and confocal imaging for nab-PTX uptake in lung tumors (visualized by DAPI staining). Left = representative images, blue and yellow lines outline the lung and tumors, respectively (scale bar=2 mm); right = quantification (data are means ± s.e.m. for n=265 tumors from 7 mice across 2 conditions, two-tailed t-test). e. C57BL/6 mice bearing subcutaneous (s.c.) iKras tumors were treated with tram, nab-PTX, and/or vehicle controls, and 24 hr later were analyzed by fluorescent reflectance imaging for nab-PTX (e; scale bar = 5 mm). f-g. Mice bearing iKras s.c. tumors were treated with vehicle controls, nab-PTX, tram, nab-PTX/tram Co-Tx or Sep-Tx (nab-PTX followed by tram) with equal cumulative doses, and tumor volume was monitored over time by caliper (f). Representative images of tumor size (scale bar=5 mm) and quantification at 6 days post treatment are shown (g, bars denote individual tumors). For a-c, n=3 independent experiments per condition (two-tailed t-test); for e, n=3 animals per condition (one-way ANOVA/Tukey compared to control). For f-g, n=12 tumors from n=6 animals per condition (one-way ANOVA/Tukey). Data are means ± s.e.m for all except g. For d-g, the controls are tumor-bearing animals treated via i.v. injection with PBS (vehicle control for nab-PTX) and/or oral gavage with methylcellulose+tween 20 (vehicle control for tram).

Reprogramming nutrient signaling enhances nab-PTX efficacy

We next tested whether nab-PTX uptake could be therapeutically enhanced to improve efficacy. Guided by a prior genome-wide shRNA screen²⁵, we tested how 13 compounds affected nab-PTX uptake (Fig. 5a and S14–15), and found six hits: CFI-400945 (polo like kinase 4 inhibitor), AXL1717 (IGF1R inhibitor), VX-745 (p38α inhibitor), AZD-7762 (CHK1/2 inhibitor), JNK-IN-8 (JNK1/2/3 inhibitor), and AZD-1390 (ATM kinase inhibitor). Of these, we focused on AXL1717 and IGF1R as a target, since IGF1R-targeted therapies (IGF1Ri) have entered Ph3 in NSCLC²⁶ and IGF1Ri has been shown to synergize with PD-1 blockade²⁷. iKras and KP cells express IGF1R (Fig. S16a) and two IGF1R inhibitors (IGF1Ri) enhanced nab-PTX uptake by >4-fold: AXL1717 as above, and the Ph3-tested drug linsitinib (OSI-906) (Fig. 5b). Furthermore, IGF1Ri enhanced the macropinocytic activity of cancer cells as measured by the uptake of 70 kDa dextran (Fig. S16b), similar to the results obtained for nab-PTX.

Figure 5. IGF1R inhibitor enhances uptake of nab-PTX in an AMPK- and macropinocytosis-dependent manner.

a. A targeted screen tested the ability of drugs to enhance nab-PTX uptake in KP and iKras cancer cells across a 6-point dose response (means ± s.e.m., n=3), from which maximum change and area under the curve (AUC) were calculated and averaged across both cell types. b-d. Representative images (b, scale bar=10 μm) and quantification of nab-PTX uptake (b: n=522 cells, c: n=404 cells, d: n=417 cells across all conditions) in treated iKras cells (control=DMSO). Data are means ± s.e.m. (one-way ANOVA/Tukey compared to control). e-g. nab-PTX cytotoxicity in iKras cells treated either before or after with AXL1717, linsitinib (Lin), or A769662 (A76), measured at 72 hr by resazurin-based assay. Data are means ± s.e.m. for n=3 (two-tailed t-test). Data in g (IC₅₀ value is displayed in each bar) summarize curves in e-f and Fig. S16.

We hypothesized that IGF1Ri enhanced macropinocytic nab-PTX uptake by modulating AMP-activated protein kinase (AMPK), since AMPK is a key regulator of metabolism and nutrient signaling, its activation is linked to catabolism and macropinocytosis^28,29, and IGF1R signaling can downregulate AMPK activity^30,31. To test this hypothesis, we treated iKras cells with the AMPK activator A769661 (A76), which phenocopied AXL1717 by enhancing nab-PTX uptake (Fig. 5b). We also hypothesized that glucose deprivation could enhance nab-PTX uptake, since it is known to stimulate AMPK²⁸. Low glucose culture media also phenocopied AXL1717’s effects on nab-PTX uptake (Fig. 5b). Using AMPK inhibitor dorsomorphin (Compound C) and the macropinocytosis inhibitor EIPA, we found that AXL1717- and low glucose-mediated enhancements in nab-PTX uptake were both dependent on AMPK activation and macropinocytosis (Fig. 5c–d and S16c). Moreover, both AXL1717 treatment and low glucose culturing condition enhanced the uptake of cross-linked nab-PTX by cancer cells (Fig. S17).

Stimulated uptake of nab-PTX enhanced its cytotoxicity. In vitro, pre-treatment with AXL1717, A76, or linsitinib improved (decreased) the IC₅₀ of nab-PTX or nab-PTX+gemcitabine, but not sb-PTX, by >5~10-fold (Fig. 5e–g, S16d–e, and S18a–d), and controls showed no single-agent dose-timing effects (Fig. S18e–i). In vivo, both AXL1717 and 2 days of animal fasting (Fig. S19a and S20a) enhanced the nab-PTX accumulation in autochthonous lung tumors and/or subcutaneous tumors (Fig. 6a–c). However, fasting also increased nab-PTX in the liver, while AXL1717 did not, suggesting the latter may improve nab-PTX therapeutic index (Fig. 6b–c, S19b–c, and S20b–c). Vascular perfusion and accumulation of dextran-NP (Macrin NP) did not change with AXL1717 or fasting (Fig. S19d–g and S20d–e), suggesting enhanced nab-PTX delivery was not due to broadly altered EPR effects.

Figure 6. IGF1R-targeted kinase inhibitor enhances nab-PTX uptake and efficacy in vivo.

a. 129S mice bearing Kras^G12D/+p53^−/− autochthonous lung tumors were pre-treated with AXL1717 or vehicle (re-shown control, Fig. 4d), followed by i.v. nab-PTX injection, and 24 hr later analyzed by CUBIC clearing and confocal imaging for nab-PTX uptake in lung tumors (visualized by DAPI staining). Left=representative images, blue and yellow lines outline the lung and tumors, respectively (scale bar=2 mm); right = quantification (data are means ± s.e.m. for n=329 tumors from 8 animals across 2 conditions; two-tailed t-test). b-c. Animals bearing subcutaneous (s.c.) iKras tumors were treated with AXL1717 (b, control=vehicle control) or fasted (c, control=normal diet), nab-PTX was i.v. injected, and biodistribution was measured 24 hr later by fluorescent reflectance imaging. Representative images (left, scale bar=5 mm) and quantification (right) show the impact of AXL1717 (b) and fasting (c) on nab-PTX accumulation. For b-c, data are means ± s.e.m. for n=3 animals per group (b, one-way ANOVA/Tukey compared to control; c, two-tailed t-test). d-g. Mice bearing subcutaneous iKras tumors were treated with vehicle controls, nab-PTX, AXL1717, simultaneous treatment of both (Co-tx), or sequential treatment of both (Sep-tx) with equal cumulative doses. Tumor volume (d-e) and survival based on humane endpoint (f, n=4 per group, log-rank test) were monitored. Representative images of tumor size (e, scale bar=5 mm), as well as percent change in tumor volume (g), are displayed. For d and g, data are means ± s.e.m from n=12 tumors from 6 mice per condition (one-way ANOVA/Tukey). For a-g, the controls are tumor-bearing animals treated i.v. with PBS (vehicle control for nab-PTX) and/or i.p. with olive oil (vehicle control for AXL1717). h. Proposed mechanisms for effects of MEK1/2 and IGF1R inhibitors on nab-PTX uptake. MAPKi, such as trametinib or mutant RAS silencing, can reduce nab-PTX uptake by inhibiting the activity of RAS/RAF/MEK/ERK or RAS/RAC/PAK1 pathways³. Fasting and IGF1R inhibitor can enhance uptake by promoting AMPK signaling.

Drug-enhanced nab-PTX uptake translated into improved efficacy in iKras allografts. Alternating treatments with AXL1717 and nab-PTX (Table S3), such that the two drugs were separately and sequentially dosed (Sep-tx), only modestly slowed tumor growth compared to monotherapy. However, tumor growth was blocked and survival (time to humane endpoint) was significantly extended (despite the limited sample size per group) when simultaneous dosing allowed AXL1717 to enhance nab-PTX uptake (Co-tx, Fig. 6d–g and S21). In contrast, this enhancement in efficacy was not observed using sb-PTX (Table S4 and Fig. S22). These results show that IGF1R inhibitors can enhance macropinocytosis in an AMPK-dependent manner, similar to glucose deprivation, and improve the accumulation and tumor-killing action of nab-PTX (Fig. 6h).

Conclusion

In summary, we report that the in vivo delivery and action of nab-PTX are affected by RAS signaling state and macropinocytosis in cancer cells. We further demonstrate that therapeutic MAPK inhibition can block the selective delivery and efficacy of nab-PTX, while IGF1Ri enhances its uptake in an AMPK-dependent manner (Fig. 6h). Indeed, a signature of gene expression controlled by MAPK and AMPK signaling output^32,33 correlates with nab-PTX uptake across multiple cancer cell lines, further suggesting that nab-PTX uptake is controlled by MAPK and AMPK activities (Table S5 and Fig. S23). Moreover, since both IGF1R and AMPK are key regulators of cell metabolism²⁸, our results thus indicate that nab-PTX uptake and efficacy can be enhanced by reprogramming nutrient signaling. We also note that while these macropinocytosis- and nutrient signaling-dependent effects apply to cross-linked nab-PTX, they do not extend to sb-PTX (formulated with Cremophor EL as used clinically), possibly because Cremophor-EL can reduce taxane binding to albumin²⁰. Hence, we anticipate generalizability of our findings to other albumin-bound drugs including nanotherapies with albumin components³⁴, albumin-binding prodrugs³⁵, albumin-fusion biologics³⁶, and potentially numerous drugs that have high intrinsic affinity for albumin and plasma protein binding, such as the BRAF inhibitors dabrafenib or vemurafenib³⁷.

Our results have implications to clinical trials using MAPK inhibitors+nab-PTX therapy, and provide rationale for IGF1R inhibitor+nab-PTX combination therapy. We have also demonstrated, through in vitro experiments, that MEK1/2 and IGF1R inhibitors affect the efficacy of nab-PTX+gemcitabine treatment. Future in vivo experiments should test these results in animal models, and provide further evidence on how MAPK/AMPK activities affect nab-PTX+gemcitabine or nab-PTX+oxaliplatin combination therapies. Although here we focus solely on downstream RAS/RAF/MEK/ERK signaling, it is known that nutrient uptake can be stimulated by upstream receptor tyrosine kinases including EGFR and HER2³⁸. Hence, it is possible that EGFR- and HER2-targeted kinase inhibitors such as erlotinib and lapatinib may similarly impact nab-PTX, pertinent to clinical trials testing their combination (T-DM1+lapatinib+nab-PTX: NCT02073916; erlotinib+nab-PTX+gemcitabine: NCT01010945). Overall, our results demonstrate the potential impacts of MAPK and nutrient signaling state on nab-PTX action, and we anticipate that the image-based experimental approach used here will be useful in understanding the extent to which this mechanism applies in other cancer-types and drug combinations.

Methods

Cell Culture and Animal Models

Cell Lines and Cell Culture

KP1.9 mouse cancer cells (KP, a gift from Dr. Annette Zippelius, University Hospital Basel), derived from Cre-activated Kras^LSL-G12D/+Trp53^flox/flox genetically engineered mouse model (GEMM) of lung adenocarcinoma¹⁸, were cultured in Iscove’s Modified Dulbecco’s Medium (IMDM, Invitrogen). AC37 mouse cancer cells with doxycycline-inducible Kras^G12D expression (iKras, a gift from Dr. Haoqiang Ying, MD Anderson Cancer Center, by way of Dr. Nabeel Bardeesy, MGH), derived from triple transgenic p48-Cre;ROSA26-LSL-rtTa-IRES-GFP;TetO-LSL-Kras^G12D genetically engineered mouse model of pancreatic ductal carcinoma³⁹, were maintained in Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 media (DMEM/F12, Invitrogen) supplemented with 2 μg/mL doxycycline (doxy, Sigma). NCI-H1838 human lung adenocarcinoma cells, BxPC-3 human pancreatic adenocarcinoma cells, CT26 mouse colon carcinoma cells, and AsPC-1 human pancreatic adenocarcinoma cells were cultured in RPMI1640 media (Corning). Mia PaCa-2 human pancreatic carcinoma cells (Mia), LS 180 human colon adenocarcinoma cells, HT-1080 human fibrosarcoma cells (HT), Raw 264.7 macrophages (Raw), MDA-MB-231 human breast adenocarcinoma cells (MDA), and A-375 human melanoma cells were all cultured in Dulbecco’s Modified Eagle Medium (DMEM, Corning). The murine 4306 cancer cell line (Kras^G12D;Pten^−/− endometrioid ovarian adenocarcinoma, OVA, from Dr. Daniela Dinulescu, Harvard Medical School), isolated from the Cre-activated LSL-Kras^G12D/+Pten^loxP/loxP ovarian cancer genetically engineered mouse model^40,41, and 3743 mouse thyroid cancer cells (ATC, from Dr. Sareh Parangi, MGH), isolated from Braf^V600E/WTp53^−/− genetically engineered mouse model of anaplastic thyroid tumors^42,43, were also cultured in DMEM. MCF10A human breast epithelial cells were cultured in DMEM/F12 media supplemented with 20 ng/mL EGF (Peprotech), 0.5 μg/mL hydrocortisone (Sigma), 100 ng/mL cholera toxin (Sigma), and 10 μg/mL insulin (Sigma). Bone marrow-derived macrophages (BMDMs) were isolated from the femurs of 10 wks old C57BL/6 mice, and maintained in IMDM supplemented with 10 ng/mL M-CSF (PeproTech). After 6-days of culturing, the BMDMs were subsequently treated with 100 ng/mL LPS (Sigma)+50 ng/mL IFNγ (PeproTech) or 10 ng/mL IL4 (PeproTech) for polarization toward a pro-inflammatory M1-like or anti-inflammatory M2-like phenotype, respectively⁴⁴. All growth media were supplemented with 10% fetal-bovine serum (FBS, Invitrogen) or 5% horse serum (for MCF10A growth media, Invitrogen), 100 IU penicillin, and 100 μg/mL streptomycin (Invitrogen). All cell lines, except for KP, iKras, OVA, and ATC, were obtained from American Type Culture Collection (ATCC). In some experiments, eGFP-expressing versions of KP¹⁹ and ATC⁴⁵ cells were used. All cells were cultured at 5% CO₂ and 37°C in a humidified incubator, and were routinely tested for mycoplasma contamination (PlasmoTest Mycoplasma Detection Kit, InvivoGen).

Animal Models

Animal research was performed in compliance with guidelines from the Institutional Animal Care and Use Committee (IACUC) at Massachusetts General Hospital (MGH). Unless otherwise noted, experiments were performed using female mice that were 6–12 weeks old. 129S1 mice harboring Cre-activated conditional Kras^LSL-G12D/+Trp53^flox/flox (129KP mice, a generous gift from Dr. Mikael J. Pittet, MGH) were used as an autochthonous mouse model of non-small cell lung cancer¹⁸. C57BL/6J (JAX), B6129SF1/J (JAX), and 129KP mice were fed with autoclaved food and water and maintained in ventilated cages in a light-dark cycle, temperature (18–23 °C), and humidity (40–60%) controlled pathogen-free vivarium at MGH. Litter-mates were randomly assigned to experimental groups, and the mice underwent procedure and monitoring on the same day in a randomized order.

METHOD DETAILS

Materials and Pharmaceutical Agents

Reagents and solvents were purchased from Sigma-Aldrich (St. Louis, MO) and used without further purification, unless otherwise indicated. Water was purified by MilliQ filtration systems (Millipore) prior to experimental use. Trametinib (Tram), gemcitabine (Gem), and paclitaxel (PTX) were purchased from LC Laboratories (Woburn, MA). STEALTH liposomal doxorubicin HCl (Doxil) and DiI-labelled PEGylated DOPC/CHOL liposomes (Lip) were purchased from FormuMax (Sunnyvale, CA). Pharmacy-grade nanoparticulate albumin-bound paclitaxel (nab-PTX, trade name: Abraxane®; Celgene; Summit, NJ) was purchased form McKesson (Irving, Taxes). CFI400945 fumarate (CFI400945), AXL1717⁴⁶, VX-745, AZD-7762, JNK-IN-8, AZD-1390, Bindarit, MK-8033, AS602801, cobimetinib, refametinib, MK-8353, SCH772984, A-769662, Compound C dihydrochloride (Compound C), and Linsitinib (Lin)⁴⁷ were all purchased from MedChemExpress (Monmouth Junction, NJ). RepSox, 7rh, Y-27632 dihydrochloride (Y-27632), and C 021 dihydrochloride (C021) were purchased from Tocris (Bristol, UK), Sigma, Abcam (Cambridge, MA), and R&D Systems (Minneapolis, MN), respectively. 5-(N-Ethyl-N-isopropyl)amiloride (EIPA) was purchased from Sigma-Aldrich.

Fluorescent dye conjugation

Pharmacy-grade nab-PTX (Abraxane^®, Celgene) and human serum albumin (Alb, Sigma) were conjugated to AlexaFluor 488 (A488, λ_ex/λ_em = 490/525 nm), 555 (A555, λ_ex/λ_em = 555/580 nm), or 647 (A647, λ_ex/λ_em = 650/665 nm) via NHS-ester chemistry. Briefly, 6 mg nab-PTX or 5.4 mg albumin was dissolved in 100 μL sodium bicarbonate buffer (0.1 M, pH=8.2) and mixed with 1 mg AlexaFluor NHS ester dyes (Invitrogen). The mixture was continuously stirred for 5 hr at room temperature at 300 rpm. Unreacted AlexaFluor dye was separated from the reaction mixture via centrifugation using 50 kDa (for nab-PTX) or 30 kDa (for Alb) molecular-weight cut-off (MWCO) Amicon centrifugal filter units (Millipore).

BODIPY-TMR-NH₂ (Lumiprobe, λ_ex/λ_em = 545/570 nm) was conjugated to poly(D,L-lactide-co-glycolide) (PLGA, 50:50 lactide:glycolide, mol wt 30 kDa-60 kDa, Sigma) as described previously⁴⁸. Briefly, 200 mg (0.004 mmol) PLGA, 16 mg (0.08 mmol) 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDCI, Sigma), and 4 mg (0.04 mmol) N-Hydroxysuccinimide (NHS, Sigma) dissolved in dimethylformamide (DMF, Sigma) were mixed with 4 mg (0.009 mmol) BODIPY-TMR-NH₂ dissolved in N,N-Diisopropylethylamine (DIPEA, 0.04 mmol, Sigma). The mixture was continuously stirred in room temperature in the dark for 24 hr. Polymers conjugated to the fluorescent dye (PLGA-BODIPY-TMR) were then purified from unreacted dyes with repeated cycle of participation with diethyl ether, followed by centrifugation (2000×g, 10 min) and washing in minimal amount of dichloromethane. The red precipitate resulting from the purification step was vacuum dried over night before dissolving in DMF.

Taxane was conjugated to silicon rhodamine carboxylate (SiR-COOH, λ_ex/λ_em = 650/670 nm) as we previously synthesized and characterized⁴⁹. Cross-linked iron oxide nanoparticles (CLIO) and polyglucose (dextran) nanoparticles (Macrin NP) were previously validated and used as in vivo imaging probes for macrophages, as they preferentially accumulate in these cells^24,50. CLIO and Macrin NP were conjugated to VivoTag 680-NHS (VT680-NHS, Perkin Elmer, λ_ex/λ_em = 665/688 nm) according to previously published protocols^24,50. Cyclodextrin nanoparticle (CDNP), a newly designed nanoparticle targeting macrophages, was conjugated to A555-NHS ester using NHS-ester chemistry following published protocols⁴⁴.

Nanoparticle synthesis and characterization

CLIO was synthesized as described previously^50,51. Briefly, monodispersed superparamagnetic iron oxide colloid (MION, synthesized in MGH-CSB, Boston, MA; U.S. Patent 5,262,176) was reacted with epichlorohydrin (Sigma) to crosslink the dextran coating of the MION. The resulting product was then aminated to produce CLIO.

Macrin NP was synthesized according to previously published protocols²⁴. Carboxymethylated polyglucose (TdB Consultancy; Uppsala, Sweden) activated with EDCI and NHS were mixed with L-Lysine (Sigma) dissolved in 2-(N-morpholino)ethanesulfonic acid (MES, Sigma) buffer. The resulting mixture was stirred at room temperature for at least 5 hr. Macrin NP was recovered from the reactants by precipitation with ethanol and centrifugation (2500xg, 3 min) to produce a white paste. This white paste was dissolved in water and filtered with 0.22 μm nylon syringe filter, followed by 3-day dialysis against water with 8–10 kDa MWCO dialysis cassette at room temperature. The resulting product was again filtered with 0.22 μm syringe filter and lyophilized.

β-cyclodextrin nanoparticle (CDNP) was produced following a previously published protocol⁴⁴. Briefly, succinyl-β-cyclodextrin (CD, Sigma) was mixed with EDCI and NHS in MES buffer and stirred for 30 mins. L-lysine was then added, and the mixture was stirred overnight at room temperature to crosslink cyclodextrin. The resulting product was recovered by precipitation from ice-cold ethanol and centrifugation at 2500 × g for 3 min. The precipitant was re-dissolved in water, filtered with 0.22 μm nylon syringe filter, and concentrated with centrifugal filtration using a 10 kDa MWCO Amicon centrifugal filter unit. The final product was lyophilized before future use.

Poly(D,L-lactic-co-glycolic acid)-b-polyethylene glycol (PLGA-PEG) nanoparticles were synthesized via nano-precipitation and fluorescently labelled with co-encapsulation of PLGA-BODIPY-TMR⁴⁸. Briefly, 5 mg PLGA(75:25 lactide:glycolide)_8.3kDa-PEG_5.5kDa (Advanced Polymer Materials, Inc, Lot# 01-13-146-1) and 1 mg PLGA-BODIPY-TMR was dissolved in 400 μL DMF/acetonitrile mixture (1 DMF:1 acetonitrile). The resulting mixture was added drop-wise into 10 mL water and stirred at room temperature overnight, followed by syringe filtration through a 0.45 μm cellulose acetate filter (Whatman, GE Healthcare). The filtrate was concentrated with a 100 kDa MWCO Amicon centrifugal filter unit (Millipore) operating at 3000 × g for 30 min to obtain PLGA-PEG nanoparticle mixture.

Lab-made AlexaFluor555-labeled nanoparticulate albumin-bound SiR-labeled paclitaxel (nab-A555-SiR-taxane) was prepared by coupling SiR-taxane to A555-labeled albumin (Alb-A555) via ultrasonic homogenization. 1 mg SiR-taxane was dissolved in 64.8 μL chloroform (Sigma) and 7.2 μL ethanol. This mixture was combined with 18 mg Alb-A555 dissolved in 3.6 mL water, and stirred vigorously for 5 mins. The resulting mixture was sonicated at 40% amplitude for 4 cycles (1 min sonication followed by 30 sec of pause) using a probe-tip sonicator (QSonica) to produce albumin nanoparticles. Chloroform and ethanol were extracted from the mixture using a Rotavapor (Buchi) operating at 40°C¹⁶. The aqueous mixture was lyophilized to obtain nab-A555-SiR-taxane. In certain experiments, synthesized fluorescent conjugates of nab-PTX were covalently cross-linked by 8% glutaraldehyde and purified via 100 kDa MWCO Amicon centrifugal filter unit following previous protocols^52,53.

The sizes of nanoparticles were characterized by dynamic light scattering (DLS) using a Zetasizer APS (Malvern): Macrin NP=17 nm (PDI=0.22)²⁴; PLGA-PEG nanoparticle=70 nm (PDI=0.14); CDNP=25 nm (PDI=0.24)⁴⁴; MNP=34 nm (PDI=0.13); liposome=107 nm (PDI=0.10), Doxil=84 nm (0.02); nab-A555-SiR-taxane=140 nm (PDI=0.09); pharmacy-grade nab-PTX=132 nm (PDI=0.11); cross-linked nab-A555-SiR-taxane=167 nm (PDI=0.34). The zeta-potentials of the particles were determined in PBS using a Zetasizer ZS (Malvern): Macrin NP=−4 mV²⁴; PLGA-PEG nanoparticle=−17.2 mV; CDNP=−10 mV⁴⁴; MNP=−15.6 mV; liposome=−1 mV, Doxil=−2 mV; nab-A555-SiR-taxane=−13.2 mV; pharmacy-grade nab-PTX=−11.4 mV, cross-linked nab-A555-SiR-taxane=−9.8 mV. When appropriate, pharmacy-grade nab-PTX, nab-A555-SiR-taxane, or cross-linked nab-A555-SiR-taxane was diluted to 0.75 mg/mL nab-PTX (0.075 mg PTX payload/mL) to simulate the dissociation of the nanoparticles in blood⁵.

Transmission electron microscope (JEOL 1011) was used to assess nanoparticle morphology. Prior to imaging, nanoparticles were prepared as previously described⁴⁸. Briefly, 10 uL of 15 mg/mL pharmacy-grade nab-PTX (Abraxane®, McKesson) or lab-made nab-PTX was deposited on a carbon-coated copper grid. The sample was then stained with a solution of uranyl acetate to provide contrast, and imaged immediately. Finally, encapsulation efficiency and loading capacity of lab-made nab-PTX were calculated by measuring the amount of PTX in the nanoparticle with high pressure liquid chromatography-mass spectrometry (HPLC-Mass Spec), and using the following formula:

• $$\text{Encapsulation}\text{efficiency}=\left(\text{weight}\text{of}\text{drug}\text{in}\text{the}\text{nanoparticle}\text{/}\text{weight}\text{of}\text{drug}\text{added}\text{when}\text{making}\text{nanoparticle}\right)\times 100\%.$$
• $$\text{Loading}\text{efficiency}=\left(\text{weight}\text{of}\text{drug}\text{in}\text{the}\text{nanoparticle}\text{/}\text{total}\text{weight}\text{of}\text{nanoparticle}\right)\times 100\%.$$

Based on these formulas, we calculated the encapsulation efficiency and loading capacity of lab-made nab-PTX to be 88.5 ± 2.6% and 5.11 ± 0.15% (means ± s.e.m., n=3), respectively.

Cell transfection and transduction

KP and iKras cells expressing Erk kinase translocation mClover (related to GFP) reporter (ERK-KTR) were generated by lentiviral transduction using pLentiCMV Puro DEST ERKKTRClover (Addgene plasmid #59150, a gift from Dr. Markus Covert)²³. Vectors were packaged using the Lenti-X HTX Packaging System (Clontech) according to manufacturer’s protocols. KP or iKras cells stably expressing ERK-KTR, referred to as KP ERK-KTR or iKras ERK-KTR, respectively, were selected with growth media containing 10 μg/mL puromycin (Invitrogen) for 7 days and subsequently sorted for cells with high expression of fluorescent proteins via fluorescence activated cell sorting (FACS, MGH Flow Cytometry Core).

BxPC-3 cells transiently expressing Kras^G12D mutation were generated by transfecting parental BxPC-3 cells with pBabe-Kras^G12D plasmid (Addgene plasmid #58902, a gift from Dr. Channing Der) using Lipofectamine 3000 system (Invitrogen) according to manufacturer’s protocols. As a control, parental BxPC-3 were transfected, again using Lipofectamine 3000, with pBabe-Kras Wt plasmid (Addgene, plasmid #75282, a gift from Dr. Channing Der).

In vitro drug uptake and fluorescent microscopy

To quantify nab-PTX, cross-linked nab-PTX, PLGA-PEG NP, or 70 kDa dextran cellular uptake, 7000 MCF10A, H1838, BxPC-3, CT26, Mia, AsPC-1, OVA, LS 180, HT, MDA, A-375, ATC, Raw, BMDM, KP, or iKras cells were seeded in each well of a 96-well tissue-culture treated image plates (ibidi). After overnight incubation, cells were treated with 5 μg/mL fluorescently-labelled nab-PTX (nab-PTX-A555 or nab-A555-SiR-taxane; both containing ~0.5 μg/mL PTX), 5 μg/mL (~0.5 μg/mL PTX) cross-linked nab-A555-SiR-taxane, 6 μg/mL PLGA-PEG NP-BODIPY-TMR, or 100 μg/mL fluorescently-labelled 70 kDa dextran (70 kDa dextran-A647) for 4 hr. The cells were then washed 5x with ice-cold PBS and fixed with 4% para-formaldehyde (PFA, EMS) for subsequent fluorescent microscopy imaging or fluorescent detection by plate reader (Tecan). When appropriate, KP or iKras cells were treated with various pharmaceutical inhibitors for 6 hr (for EIPA) to 24 hr prior to treatment with fluorescent nab-PTX, cross-linked nab-PTX, or 70 kDa dextran. These inhibitors include: 50–100 μM EIPA, 5 μM Compound C, 100 nM Cobimetinib, 100 nM Refametinib, 100 nM MK-8353, 100 nM SCH772984, 1 nM-10 μM Trametinab, CFI-40095, AXL1717, VX-745, AZD-7762, JNK-IN-8, ACD-1390, Bindarit, MK-8033, RepSox, 7rh, Y-27632, C021, AS602801, Linsitinib, and/or A-769662. To assess the effects of glucose deprivation on nab-PTX or cross-linked nab-PTX uptake, iKras cells were washed with PBS and treated with DMEM low-glucose medium (containing 1 g/L D-glucose, Invitrogen) or DMEM high-glucose medium (containing 4.5 g/L D-glucose, Invitrogen), both supplemented with 10% FBS, 100 IU penicillin, 100 μg/mL streptomycin, and 2 μg/mL doxy, for 24 hr prior to the addition of fluorescently-labelled nab-PTX or cross-linked nab-PTX. To evaluate the effects of Kras mutation on nab-PTX uptake, iKras cells were washed with PBS and cultured in DMEM/F12 growth media containing no doxy for 48 hr or 5 days prior to fluorescent nab-PTX treatment. In some experiments, KP cells were treated with 10 μg/mL DQ-Red-BSA or 70 kDa FITC-dextran for 6 hr prior to treatment with fluorescent nab-PTX. In some experiments, to visualize the effects of pharmaceutical inhibition on MAPK pathway activity, KP ERK-KTR or iKras ERK-KTR cells were used. To label endosomes and lysosomes, KP cells were transduced overnight with RAB5A-GFP, RAB7A-GFP, or LAMP1-GFP fusion proteins using BacMam 2.0 reagents (Invitrogen) according to manufacturer’s protocols, prior to drug treatment.

To visualize the nucleus, the fixed samples were counter-stained with 5 μg/mL diamidino-2-phenylindole (DAPI, Invitrogen). When appropriate, the fixed KP and iKras cells were stained with 1 μg/mL anti-IGF1 receptor antibody (Clone #EPR19322, Abcam) followed by appropriate secondary antibody conjugated to A488 (Invitrogen). The fluorescent images of drugs, subcellular compartments, and the ERK-KTR signals inside the cells were captured with a DeltaVision (Applied Precision) modified Olympus BX63 microscope fitted with a Neo sCMOS monochrome camera (Andor).

In vitro drug efficacy study

3000 KP or iKras cells were seeded in each well of 96-well plates (Costar). After overnight incubation, cells were treated with various regimens of drugs or DMSO vehicle control. To assess the effects of KRAS mutation on nab-PTX cytotoxicity, iKras cells were cultured without the addition of doxy in the growth media for 24 hr, followed by the treatment of various concentrations of nab-PTX for 48 hr. For evaluating the effects of Tram on the efficacy of nab-PTX or sb-PTX, KP or iKras cells were treated with 1–100 nM of Tram and various concentrations of nab-PTX or sb-PTX together for 48 hr (Co-treatment). This co-treatment regimen was compared to the Sep-treatment regimen, in which cells were first treated with various concentration of nab-PTX or sb-PTX for 24 hr followed by PBS wash-out and further treatment of 1–100 nM Tram for an additional 24 hr. To assess the effects of AXL1717, Linsitinib, or A-769662 on nab-PTX or sb-PTX cytotoxicity, iKras cells were first pre-treated with 1 μM AXL1717, 1 μM Linsitinib or 100 nM A-769662 for 24 hr, followed by PBS wash-out and further treatment with various concentrations of nab-PTX or sb-PTX for an additional 24 hr. This treatment regimen was compared to the regimen in which cells were first treated with nab-PTX or sb-PTX for 24 hr, followed by PBS wash-out and further treatment of 1 μM AXL1717, 1 μM Linsitinib, or 100 nM A-769662 for an additional 24 hr. In some experiments, nab-PTX was added with equal-molar of gemcitabine. Appropriate single-agent (i.e. nab-PTX, Tram, or AXL1717, etc.) control experiments using matched dosing timings were performed for all drugs evaluated. For all cytotoxicity experiments, all drugs were washed out 48 hr after the initiation of treatment, and the cells were cultured for a further 24 hr prior to the assessment of viability using PrestoBlue (Invitrogen) and a plate reader (Tecan) according to the manufacturer’s protocol. IC₅₀ was determined from the cytotoxicity curve, obtained from non-linear fit of the data using Prism 7.0 (GraphPad Software). In some in vitro experiments, PTX dissolved in DMSO or Cremophor EL:ethanol mixture (sb-PTX) was used instead of nab-PTX.

Animal studies and tumor growth evaluation.

Unless otherwise noted, experiments were performed using mice that were 6–12 weeks old. To generate KP, iKras, or ATC lung tumor models, 2.5 × 10⁵ KP, 1.0 × 10⁵ iKras, or 1.0 × 10⁵ ATC cancer cells in 100 μL PBS were injected intravenously (i.v.) via tail-vein catheter into female C57BL/6 mice (for KP and iKras tumor models) or B6129SF1/J mice (for ATC tumor model). The animals were euthanized and lungs harvested for imaging 15 days, for iKras lung tumor model, or 30 days, for KP and ATC lung tumor models, following tumor inoculation. For subcutaneous (s.c.) tumor models, 1.0 × 10⁶ KP, iKras, or ATC cells in 100 μL PBS were subcutaneously injected in the flanks of C57BL/6 mice (for KP and iKras tumor models) or B6129SF1/J mice (for ATC tumor model). Imaging or tumor growth evaluation was performed approximately 1–2 weeks post-inoculation when the s.c. tumors reached a palpable size (~5 mm in diameter). When indicated, KP cells or ATC cells expressing eGFP or ERK-KTR, or iKras cells expressing H2B-mApple, were used. The mice bearing iKras tumors were fed with water containing 2 g/L doxy and 20 g/L sucrose to maintain the expression of Kras^G12D mutation in the cancer cells. To generate autochthonous KP lung adenocarcinoma, 129KP mice were infected with adenovirus expressing Cre recombinase (AdCre, University of Iowa Viral Vector Core) via intranasal instillation¹⁸. Lung tumors developed spontaneously and are detectable via magnetic resonance imaging (MRI) 8–12 weeks after viral infection of 129KP mice, at which points these mice were used for imaging experiments. When necessary, all mice were anesthetized with 2% isoflurane supplied with 2 L/min O₂ on a heated stage. In all experiments, subjects were monitored daily and any animal reaching humane endpoint, as defined by IACUC, was euthanized. No treatment-related toxicities were observed that triggered >10% loss in body weight or any of the above criteria.

1 million iKras cells (for Fig. 4f and g, and Fig S12 and 13) or 0.5 million iKras cells (for Fig. 6d–g, and Fig. S21 and S22) in 100 μL of PBS were implanted subcutaneously into the flanks of C57BL/6 mice. iKras subcutaneous tumors were allowed to grow to ~50 mm³ prior to the initiation of the treatment. Before treatment, the mice were assigned to each cohort to ensure average tumor volume and body weight were matched across all cohorts. When necessary, tumor growth experiments were performed in cohort batches with control group representation in each batch. For trametinib treatment, the mice were given 0.2–0.5 mg/kg trametinib (see Tables S1–S2), dissolved in 1% methylcellulose and 0.2% tween 80, by oral gavage (o.g.) as previously published⁵⁴. For AXL1717 treatment, 15–37.5 mg/kg AXL1717 (see Table S3–S4), dissolved in 100 μL mixture of 10% DMSO and 90% olive oil (Sigma), was administered via intraperitoneal (i.p.) injection as described previously⁵⁵. 15 mg/kg nab-PTX in 100 μL PBS was administered via i.v. injection. For sb-PTX treatment, 15 mg/kg paclitaxel was dissolved first in a mixture of Cremophor EL (Sigma) and ethanol (1:1 by volume), and further diluted in 100 μL PBS for i.v. administration. The tumor-bearing mice received treatments of nab-PTX, sb-PTX, AXL1717, tram, or appropriate combinations according to various treatment schemes (Table S1–S4). Tumor volumes (V= 0.5 × width² × length) and mice weight were monitored daily for up to 2 weeks following the initiation of the treatment using a digital caliper and scale, respectively. Animals were sacrificed when reaching the humane end-point as defined by IACUC regulations, and the values for tumor volumes at the day of euthanasia were carried over for subsequent days when plotting the tumor growth curves (Fig. 6d and S22a). The tumor growth curves for each group were terminated when half of mice in the group reached humane end-points. Finally, the mice were also monitored for signs of anaphylactic reactions (i.e. itching, dyspnea, paresis, convulsions) based on clinical score system described previously⁵⁶, and no sign was noted. Furthermore, cohorts of immunocompetent C57BL/6 mice were treated with either PBS or 9 doses of 15 mg/kg nab-PTX. Following the treatments, the blood from these mice was collected via terminal cardiac puncture, and incubated undisturbed at 4 °C for 30 min to allow for clotting. The serum was separated from the blood clot by centrifuge at 3000 rpm for 10 min at 4°C. Total IgE levels in the mouse serum was then analyzed with Mouse IgE ELISA kit (Thermo) according to manufacturer’s protocol. For nab-PTX used in the tumor growth measurement, dosing was calculated based on the amount of PTX payload.

Biodistribution analysis

15 mg/kg nab-PTX-A647 (either pharmacy-grade or lab-made), 15 mg/kg Alb-A647, 18 mg/kg PLGA-PEG NP-BODIPY-TMR, or 15 mg/kg Macrin NP-VT680 in 100 μL of PBS was injected i.v. via tail vein into C57BL/6 mice bearing subcutaneous KP or iKras tumors, or 129KP mice bearing lung tumors. 24 hr post-injection, the animals were euthanized, and the tissues were surgically resected, washed in PBS, and weighed. The fluorescence intensities of nab-PTX-A647, Alb-A647, PLGA-PEG NP-BODIPY-TMR, Macrin NP-VT680 in the tissue were measured with fluorescence reflectance imaging (A647: λ_ex=650 nm, λ_em=665 nm; VT680: λ_ex=665 nm, λ_em=688 nm; BODIPY-TMR: λ_ex=545 nm, λ_em=570 nm) using an OV110 system (Olympus). Average fluorescence intensity for each tissue was determined using Fiji (NIH), and was background-subtracted with auto-fluorescence from corresponding tissue of animals treated with vehicle control (PBS). Percentage of injected dose per gram tissue (% ID / g) was calculated using a standards prepared with varying doses of nab-PTX-A647, Macrin NP-VT680, or PLGA-PEG NP-BODIPY-TMR in 1.0% intralipid (McKesson). For nab-PTX-A647, standards were digested in pepsin (pH=4, 37 °C, pepsin:nab = 1:100 by weight) to control for fluorescence de-quenching upon cellular catabolism. When appropriate, 100 μg of Rhodamine-labelled Lectin I (Rho-Lectin, Vector Labs) in 100 μL of PBS was injected i.v. via tail vein into the animals 30 mins prior to euthanasia. Rho-Lectin intensity was assessed with fluorescent reflectance imaging (Rho-Lec: λ_ex=550 nm, λ_em=575 nm). In some experiments, the fluorescent intensities of nab-PTX, Alb, PLGA-PEG NP, Macrin NP, and Lectin in tumors were assessed with confocal microscopy instead of fluorescence reflectance imaging (see next section).

To assess the effects of Kras^G12D expression on the biodistribution of nab-PTX-A647 and PLGA-PEG NP-BODIPY-TMR, doxycycline was withdrawn from the feed water of the animals 1 or 4 days prior to nanoparticle injection (2 or 5 days prior to tissue harvesting). For the assessment of the effects of trametinib, animals were treated with 0.2 mg/kg trametinib in 1% methylcellulose and 0.2% tween 80 via oral gavage for 4 hr, 1 day, or 4 days prior to nanoparticle injection (1, 2, or 5 days prior to tissue harvesting). For assessing the effects of fasting on nab-PTX uptake, animals were fed a water-only diet for 1 days prior to nanoparticle injection (2 days prior to tissue harvesting). Finally, for the assessment of the effects of AXL1717, animals were treated with 15 mg/kg AXL1717, dissolved in 100 μL mixture of 10% DMSO and 90% olive oil, via i.p. injection for 8 hr, 1 day, or 4 days prior to nanoparticle injection (1, 2 , or 5 days prior to tissue harvesting).

In vivo drug competition experiment

To ascertain whether lab-made fluorescent conjugated of nab-PTX (nab-A555-SiR-taxane) and the pharmacy-grade nab-PTX (Abraxane®, McKesson) share the same drug target in vivo, a drug competition experiment was performed. Briefly, 1 million KP GFP cancer cells were implanted subcutaneously into female C57BL/6 mice. Three weeks following tumor implantation, the mice were administered i.v. with unlabeled pharmacy-grade nab-PTX (3 mg/kg or 15 mg/kg) or vehicle control in PBS. Two hours later, lab-made fluorescent version of nab-PTX (nab-A555-SiR-taxane) in PBS was injected intravenously. 24 hr following the fluorescent nab-PTX injection, the mice were perfused with 20 mL of PBS by cardiac puncture, and tumors were excised and immediately imaged for SiR-taxane and nab-A555 by confocal microscopy.

Whole-organ optical clearing

Confocal imaging of tumors was performed with optically cleared samples. 30 mg/kg of nab-A555-SiR-taxane, 15–30 mg/kg of pharmacy-grade nab-PTX-A647, nab-PTX-A488, or nab-PTX-A555, 18 mg/kg of BODIPY-TMR labeled PLGA-PEG nanoparticle (PLGA-PEG NP-BODIPY-TMR), 10 mg/kg of Doxil, 41.25 mg/kg of CDNP-A555, 29 mg/kg of Lip-DiI, or 15 mg/kg of Alb-A488 in 100 μL PBS was injected into the animal intravenously via tail vein 24 hr prior to euthanasia and tissue extraction. For Cremophor EL formulation of SiR-taxane (sb-PTX), 30 mg/kg of SiR-taxane was first dissolved in a mixture of Cremophor EL (Sigma) and ethanol (1:1 by volume), and further diluted in 100 μL PBS for i.v. administration. When appropriate, 37 mg/kg Macrin NP-VT680 or 10 mg Fe/kg CLIO-VT680 in 100 μL PBS was injected to label macrophages 18 hr prior to tissue extraction, while 100 μg Rho-Lec in 100 μL PBS was injected to label blood vessels 1 hr prior to tissue extraction. Mice were then anesthetized and slowly perfused with 20 mL of PBS followed by 10 mL of 4% PFA via cardiac puncture. 1 unit of Heparin was injected immediately before perfusion to prevent the coagulation of the blood. Subcutaneous tumors or lungs were then excised and further fixed in 4% PFA at 4°C overnight. The air bubbles in lungs were removed with a brief application of vacuum. For optical clearing, the fixed tissues were placed into a modified CUBIC solution¹⁹ and rocked for 2–3 days at 37°C. Nuclei were labeled with 5 μg/mL of DAPI dissolved directly into the CUBIC solution. The tissues were then placed into a microscopy-grade glass chamber filled with fresh CUBIC solution before confocal imaging. No noticeable decreases in fluorescent intensities of all fluorescent moieties, especially SiR-taxane, were observed after the tissue clearing.

Modified CUBIC solution is composed of 27 wt% Urea (Sigma), 27 wt% N,N,N’,N’-Tetrakis(2-Hydroxypropyl)ethylenediamine (quadrol, Sigma), 16 wt% Triton X-100 (Sigma), 31 wt% water, and was prepared according to published protocol¹⁹. Briefly, 125 g quadrol and 125 g urea were mixed in 144 g of water. The mixture was stirred at 50 °C until the solid dissolved. The homogenous mixture was then allowed to cool prior to the addition of 75 g Triton X-100.

Confocal microscopy

Confocal microscopy was performed with a FluoView FV1000 multi-photon confocal imaging system (Olympus). For confocal microscopy imaging of excised and cleared tissues, an XLFluor 2x air objective (NA=0.14, Olympus) was used for low-magnification imaging, while an XLUMPLFLN 20x water immersion objective (NA=1.0, Olympus) was used for high-magnification imaging. Four fluorescent channels were imaged. DAPI, GFP/A488, A555/Rhodamine/BODIPY-TMR, A647/VT680/SiR signals were excited sequentially with 405, 473, 559, and 635 nm diode lasers, respectively, in combination with DM405, 473, 559, and 635 nm dichroic beam splitters. The emitted light was separated with SDM473/560/640 nm beam splitters. DAPI, GFP/A488, A555/Rhodamine/BODIPY-TMR, A647/VT680/SiR signals were detected with BA430-455, BA490-540, BA575-620, and BA655-755 nm emission filters, respectively. Each fluorescence channel was imaged sequentially using distinct excitation and emission filter sets to ensure minimal bleed-through between channels. All lasers, beam splitters, and emission filters were purchased from Olympus. Images were collected at 150 μm or 10 μm interval in z-direction for low or high-magnification imaging, respectively.

Computation of MAPK and AMPK activity scores

To evaluate the relationships between the activities of MAPK or AMPK pathway and nab-PTX uptake, we computed MAPK or AMPK activity output score for cancer cell lines harboring different RAS and RAF mutation states (H1838, BxPC3, HT1080, Ls180, MiaPaCa2, AsPC1, MDA-MB-231, and A375). Sets of genes whose expressions are controlled and strongly correlated with MAPK and AMPK activities (MAPK or AMPK signature) were identified from previous studies (Table S5)^32,33. All genes in the MAPK signature are involved in cell proliferation, migration, or negative feedback regulation of ERK, while all genes in AMPK signatures are involved in cell metabolism. The mRNA expression level of each of these genes was obtained from publicly available RNAseq data from Cancer Cell Line Encyclopedia (CCLE)⁵⁷. For each cell line analyzed, an average MAPK or AMPK activity score was computed as averaged expression level of all the genes in MAPK or AMPK signature set, respectively. The averaged MAPK or AMPK activity score was plotted against nab-PTX uptake for each cell line for comparison.

QUANTIFICATION AND STATISTICAL ANALYSIS

Image analysis

Imaging data were acquired via Olympus FluoView FV1000 multi-photon confocal microscope or BX63 epifluorescence microscope operating on FV10-SW (Olympus) or MetaMorph softwares 7.8.6.0 (Molecular Devices), respectively. Image analysis was performed with Fiji 2.0.0 (NIH) or Matlab R2019 (Mathworks). When appropriate, maximum intensity z-projections of the images were generated. Wide-field images of cleared whole lung or tumor were generated by stitching a 5 × 5 grid of low-magnification images obtained using 2x objective. Automatic thresholding, using RenyEntropy method, and watershed segmentation were used to generated region of interest (ROI) to define the boundaries of cancer cells (from GFP signals), macrophages (from CLIO-VT680 or Macrin NP-VT680 signals), or tissues (from DAPI signals). The average fluorescence intensity of drugs (i.e. fluorescent nab-PTX) was quantified for each ROI and either background subtracted (displayed as arbitrary unit or a.u.) or normalized to background fluorescence intensity (fold increase over background, displayed as normalized intensity or norm. int.). When appropriate, these intensities were further normalized to fluorescent intensities of control groups in each experiment (displayed as fraction control/frac. ctrl., or fold increase over control/fold inc. over ctrl.). Intensity values in images are directly cross-comparable when corresponding data are displayed on the same graph (for instance as shown in Figs. 2–6). However not all intensities were normalized to the same base values across different experiments (i.e. SiR-taxane in Fig. S2a), and cross-comparing absolute intensities is inappropriate in such cases. Similarly, representative fluorescent images that are not in the same figure panel may be displayed with different exposure settings, and cross-comparing between data from different figure panels is inappropriate in such cases. In certain experiments, the uptake of fluorescent nab-PTX is assessed via measuring the intensity of fluorescently-labelled nab vehicle (i.e. nab-A555). When possible, the images were randomized prior to analysis, and the investigators were blinded to the identity of the sample from which the images were obtained. In particular, relative cellular uptakes of nab-PTX, PLGA-PEG NP, CDNP, liposome, MNP, and Doxil (Fig. 1g) were quantified in this study by measuring the fluorescent intensities of nanoparticles (NP) in cancer cells and macrophages in the cleared tissues, according to the following formula:

$${\text{R}}_{\text{cancer}}={\text{I}}_{\text{cancer}}/({\text{I}}_{\text{cancer}}+{\text{I}}_{\text{macrophages}})$$

$${\text{R}}_{\text{macrophage}}=1-{\text{R}}_{\text{cancer}}$$

where

$${\text{I}}_{\text{cancer}}=\text{fluorescent}\text{intensity}\text{of}\text{NP}\text{in}\text{cancer}\text{cells}$$

$${\text{I}}_{\text{macrophage}}=\text{fluorescent}\text{intensity}\text{of}\text{NP}\text{in}\text{macrophages}$$

$${\text{R}}_{\text{cancer}}=\text{relative}\text{cellular}\text{uptake}\text{of}\text{cancer}\text{cells}$$

$${\text{R}}_{\text{macrophage}}=\text{relative}\text{cellular}\text{uptake}\text{of}\text{macrophages}$$

The relative uptake of ferumoxytol magnetic nanoparticle (MNP) in cancer cells vs. macrophages in KP s.c. tumor model, as well as the relative uptake of PLGA-PEG NP in ATC lung tumor model, was derived from data collected in previous studies^58,59.

Statistical Analysis

All measurements are taken from biological replicate samples, and results are displayed as mean ± standard error of mean (s.e.m.) unless otherwise noted. All data analyses were performed in Excel 16.45 (Microsoft) or Prism 7.0 (GraphPad Software). A p-value of less than 0.05 was deemed statistically significant, while a p-value greater than 0.05 was deemed not statistically significant (ns). Statistical tests, including two-tailed Student’s t-test, one-way ANOVA test with Tukey, Dunnett, or Holm-Sidak post-test for multiple comparison, Spearman correlation test, Pearson correlation test (to obtain Pearson’s R value and two-tailed t-test value), and Wilcoxon log-rank test were performed with Prism 7.0. The results and the details of the statistical analysis can be found in figure legends. ANOVA tests carried out in this study were followed by post-test for multiple comparison, and adjusted p-values for multiple comparison were reported for each ANOVA test. Measurements were taken from distinct samples (i.e. cells, mice, and organs). All data meet the assumptions and criteria for the statistical tests carried out. Furthermore, D’Agostino-Pearson omnibus normality test and F-test were performed using Prism 7.0 when appropriate to ensure data meet the normality and equal variance criteria. Samples were assigned randomly to each experimental group. For each experimental group, three or more independent biological replicates (n≥3 independent experiments) were performed, and no sample was excluded for analysis. Statistical tests were performed with averaged values calculated from biologically independent samples and experiments (i.e. cell cultures, animals). Sample sizes were based on previous studies performed in our laboratory on these and similar experimental models, and chosen to meet the current standard for in vivo and in vitro experiments^48,60. When possible, the investigators were blinded to the identities of the samples’ treatment groups during data analysis.

Materials Availability

Lead contact: Miles A. Miller, miles.miller@mgh.harvard.edu

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Miles A. Miller. This study did not generate new unique reagents.

Code Availability

This study did not generate new custom code or mathematical algorithms.

Data Availability

All data are available from corresponding authors upon request. RNA seq data used to calculate MAPK/AMPK activity score for each cell line were obtained from the Cancer Cell Line Encyclopedia (https://portals.broadinstitute.org/ccle).