Thyroid Cancers Exhibit Oncogene-Enhanced Macropinocytosis that Is Restrained by IGF1R and Promote Albumin–Drug Conjugate Response

Dysfunctional vasculature and biosynthetic demands of proliferation can promote extracellular nutrient scavenging by malignant cells to sustain their survival and growth (1–4). Macropinocytosis is defined as the non-specific endocytic uptake of extracellular material by cells in a clathrin-independent manner and is recognized as a prominent scavenging pathway in cancers with oncogenic mutations that drive constitutive Ras GTPase or PI3K pathway activities (1, 5, 6). These pathways stimulate the formation of cup-like structures on the cell surface that engulf extracellular fluid through membrane ruffling promoted by F-actin (7, 8). The structures collapse into macropinosomes that are shuttled into degradative endolysosomal vesicles. Rho family GTPases Rac1 and Cdc42 (9, 10), mTOR and AMP-activated protein kinase (AMPK) signaling (11–13), MAPK/ERK activity (6), and other pathways have all been shown to regulate macropinocytosis in various contexts (14, 15).

In principle, constitutive cancer cell macropinocytosis offers a two-fold opportunity for therapeutic targeting of tumors. Because macropinocytosis mediates nutrient uptake, its inhibition has been shown to restrict cancer cell anabolism, proliferation, and resistance to biosynthesis-disrupting chemotherapies (16). On the other hand, macropinocytosis can promote tumor uptake of extracellular drugs that are bound to serum albumin or other proteins, encapsulated in nanoparticles, or are otherwise unable to transport across the plasma membrane of cells (6, 17–19). Nanoparticulate albumin bound paclitaxel (nab-paclitaxel, Abraxane) is one such drug that has been shown in mice to exhibit enhanced efficacy in a macropinocytosis-dependent manner (6).

Despite the diversity of mechanisms known to promote macropinocytosis in various cancer types, relatively little is known about the degree to which it occurs in thyroid cancers, and through which signaling pathways it may be controlled. Papillary thyroid carcinoma (PTC) comprises roughly 80% of all thyroid cancers and is characterized by a well-differentiated, localized phenotype and favorable clinical outcomes (20). Up to 50% of PTC harbors oncogenic mutations in BRAF (v-Raf murine sarcoma viral oncogene homolog B), most frequently via V600E mutation that drives constitutive RAF/MEK/ERK mitogenic signaling (21). Seven to twenty percent of PTC harbor a gene fusion involving the receptor tyrosine kinase (RTK) RET, and such RET/PTC rearrangements promote constitutive MAPK/ERK activity (22). In contrast with PTC, follicular thyroid carcinoma (FTC) makes up 15% of all thyroid cancers and is more likely to contain mutations in RAS isoforms (HRAS, NRAS, and KRAS) rather than in BRAF (23). Roughly 5% of FTC carries a mutation in the tumor suppressor PTEN, and up to 10% of individuals harboring germline PTEN mutation eventually develop FTC (24). RAS and PTEN mutations have both been shown to promote constitutive macropinocytosis in other cancer-types, including in pancreatic and prostate cancers, respectively, but little is known about such effects in thyroid cancer (1, 6, 13).

BRAF^V600E associates with worse PTC outcomes (25) and is frequently found in anaplastic thyroid cancer (ATC; refs. 26, 27), which is characterized by an aggressive undifferentiated phenotype. ATC constitutes <10% of all thyroid cancers but is responsible for up to half of all deaths (28). Combined inhibition of BRAF and downstream MEK1/2 kinases via dabrafenib and trametinib, respectively, show efficacy in treating patients with BRAF-mutant ATC (28); nonetheless, drug resistance remains common and better treatments are needed.

We hypothesized that dysregulated signaling driven by diverse oncogenic mutations, including BRAF^V600E, could promote constitutive thyroid cancer macropinocytosis. We further hypothesized that this may be especially prominent in ATC, because aggressive cancers by nature exhibit high metabolic demands to fuel their proliferation. BRAF^V600E is a common oncogenic driver mutation present across a variety of cancer types beyond PTC and ATC, including melanoma, non–small cell lung cancer, and colorectal cancer. Nonetheless, little is known about the potential role of mutant BRAF in regulating macropinocytosis. We find that various mutations—including BRAF^V600E, RET rearrangement, KRAS mutation, and PTEN loss—can promote macropinocytosis. Through in vitro and mouse allograft experiments, combined with analysis of bulk and single-cell RNA sequencing (scRNA-seq) datasets from patients, this work establishes macropinocytosis as a metabolic signature of ATC that offers avenues for both treatment and monitoring.

Human cell lines, plasmids, siRNA, antibodies, sources, and transfection details are listed in Supplementary Table S1. Murine TBP3743 (TPO-Cre^ER, Braf^tm1Mmcm/+, Trp53^{tm1Brn/tm1Brn}) ATC cells were as described previously (29). All were maintained in DMEM (Corning) with 10% FBS (Bio-Techne), 100 IU penicillin, and 100 μg/mL streptomycin (Invitrogen), and routinely tested for Mycoplasma (PCR, Applied Biological Materials). Lipofectamine 3000 (Invitrogen) and siRNA transfection reagent (Dharmacon) followed the manufacturer's protocols 72 hours before macropinocytosis experiments in 96-well plates (Ibidi).

Human serum albumin (HSA, Sigma) was conjugated to Alexa Fluor 555 (AF555) or AF647 via NHS-ester coupling, as described previously (6). Drug conjugate was synthesized by adding 125 μmol/L of mouse serum albumin (MSA, Abcam) in PBS to equal volume 125 μmol/L maleimide–valine–citrulline(vc)-MMAE (Selleck, 20% DMSO in PBS), stirring overnight at room temperature (RT), and purifying with PBS via 3×12,000 × g 5-minute centrifugation using 30 kDa MWCO filters (Amicon). Macrin-VT680 nanoparticles were synthesized as prior (6, 30).

Overnight, 5,000–10,000 cells per well were seeded in 96-well plates (Ibidi). One μmol/L HSA-AF647 or 100 μg/mL of 70 kDa dextran labeled with TMR or fluorescein (Invitrogen) was then added for 4 hours. Cells were washed with PBS, fixed (4% PFA, EMS), and counterstained (5 μg/mL DAPI, Sigma) before microscopy. In some experiments, cells were treated with drugs for 4–24 hours (Supplementary Table S2A) before adding HSA-AF647.

For microtubule-destabilizing monomethyl auristatin E (MMAE) and Alb-vc-MMAE, 3,000 8505c or TBP3743 cells per well were seeded in 96-well plates (Corning) overnight. After 72 hours treatment, cytotoxicity was assessed by PrestoBlue (Invitrogen). For assessing viability of adherent cells following dabrafenib and/or trametinib, 5,000 TBP3743 cells per well were seeded overnight, treated for 24 hours, stained with the live/dead Viability/Cytotoxicity Kit (Thermo Fisher Scientific) according to the manufacturer's protocol, and immediately imaged.

Animal research was conducted in compliance with the Institutional Animal Care and Use Committee at MGH. Mice fed with autoclaved food and water were maintained in ventilated cages in a light-dark cycle, temperature and humidity-controlled pathogen-free environment. A total of 5 × 10⁵ TBP3743 in 5 μL PBS were injected into the right thyroid lobe of 6–10-week-old female B6129SF1/J mice (JAX), or 5 × 10⁵ in 50 μL PBS were subcutaneously implanted in flanks. Tumor-free mice underwent sham surgery and PBS injection. During injections, 2.5% isoflurane with 2 L/min O₂ on a heated stage was used; buprenorphine was used before and for 3 days after surgery.

Nine days after tumor induction, mice were randomly assigned and treated with 1 μmol/kg MMAE (Selleck), 1 μmol/kg Alb-vc-MMAE, or vehicle through tail-vein injection. Tumor volumes (V = 0.5 × length × width²) were by caliper, and pre-determined humane endpoints included ulceration, volumes >500 mm³, and >20% body weight loss. Terminal cardiac puncture under anesthesia was used for MGH Veterinary Clinical Pathology Laboratory serum analysis. Livers underwent formalin fixation, paraffin embedding, sectioning, and hematoxylin and eosin staining, using the MGH Histopathology Core.

Fifteen mg/kg HSA-AF647 and 15 mg/kg 70 kDa dextran-TMR in 100 μL PBS, and 50 μL (1 mg/mL) of lectin-DyLight488 (Vector Laboratories) were tail-vein injected into tumor-bearing mice. Twenty-four hours after HSA/dextran and 30 minutes after lectin, mice were dissected following 20-mL PBS perfusion via cardiac puncture when anesthetized. Biodistribution used fluorescence reflectance imaging as prior (6). Where noted, animals received 1 mg/kg trametinib and/or 30 mg/kg dabrafenib in 1% methylcellulose via daily oral gavage, or 15 mg/kg AXL1717 in 10% DMSO and 90% olive oil, or 50 mg/kg metformin in PBS, via intraperitoneal injection starting 1 day before HSA.

Mouse bone marrow–derived macrophages (BMDM) were from 8-week-old female B6129SF1/J as prior (31, 32). A total of 15,000 BMDM and/or 4,000 TBP3743 were seeded overnight in 96-well plates (Ibidi); treated 2 hours with IMDM (Gibco) containing 10% FBS, penicillin/streptomycin, 10 ng/mL murine M-CSF (Peprotech); and 500 nmol/L AXL1717, 20 μg/mL anti-IGF1R mAb (R&D Systems), and/or vehicle. Five μmol/L of Macrin-VT680 (to mark macrophages; ref. 32) and 1 μmol/L HSA-AF555 were incubated for another 4 hours before 4% PFA fixation.

Confocal microscopy used an FV1000 (Olympus) with ×20 XLUMPLFLN (NA0.95, Olympus) and filter sets as prior (6). In vitro microscopy used an IX81 (Olympus; ref. 6). Fluorescence reflectance imaging used an OV110 (Olympus; ref. 6). Liver histology used Nanozoomer (Hamamatsu). Flow cytometry followed prior protocols (33): minced tumors were digested with 1 mg/mL collagenase type I (Worthington) and type IV (Worthington), and 100 μg/mL DNase I (Sigma) in 1:2 ratio HBSS (Thermo Fisher Scientific) and RPMI-1640 (Gibco) for 30 minutes, passed through 70 μmol/L strainers (BD Falcon), and washed with RPMI-1640. Rat anti-mouse CD16/CD32 (RRID:AB_394656) and live/dead aqua (Thermo Fisher Scientific) incubation in staining buffer was followed by CD45(RRID:AB_2565884), CD11b(RRID:AB_312793), and F4/80(RRID:AB_893477) staining, washing, and analysis on an Attune Cytometer (Thermo Fisher Scientific).

In siRNA experiments, 25 nmol/L siRNA was treated to 500,000 8505c cells in 10-cm plates; 72 hours later cells were treated and lysed in RIPA (Abcam) containing protease/phosphatase inhibitor cocktail (Cell Signaling Technology, #5872S). Triplicates of 15-μg protein were resolved on 4%–12% NuPAGE gels (Invitrogen) and transferred to nitrocellulose (Invitrogen). Membranes were incubated overnight at 4°C with antibodies for p-AMPK(RRID:AB_331250), p-IGF1R(RRID:AB_10548764), p-ERK1/2(RRID:AB_2315112), IGF1R(RRID:AB_10950969), AMPK(RRID:AB_490795), ERK(RRID:AB_390779), all 1:1,000, or β-actin(RRID:AB_2242334) at 1:2,000. Secondary antibodies with horseradish peroxidase (Cell Signaling Technology) stained 1 hour at 1:2,000, RT. Restore western blot stripping buffer (Thermo Fisher Scientific) stripped membranes for 15 minutes. ECL Chemiluminescent substrate (Thermo Fisher Scientific) was scanned (Azure Biosystems). Lysate IGF1 and total protein were measured by quantikine ELISA (R&D Systems) and Pierce BCA Protein Assay (Thermo Fisher Scientific), respectively.

In transient transfection experiments, after finishing imaging of HSA and dextran, cells were incubated with 4.5% H₂O₂ and 24 mmol/L NaOH for 2 hours at RT under light exposure to bleach fluorescence following published CycIF protocols (34). A total of 0.1% Triton-X100 (Sigma) was added for 10 minutes, followed by 1 hour incubation with 5% BSA (Thermo Fisher Scientific) and 3% goat serum in PBS. Cells were incubated with p-ERK1/2(RRID:AB_2315112) or PTEN antibody (RRID:AB_390810) at 1:200 dilution in PBS containing 0.5% BSA, overnight at 4°C, washed in PBS containing 0.5% BSA, then incubated with 2 μg/mL anti-rabbit IgG AF647 (Invitrogen) in PBS containing 0.5% BSA for 2 hours before imaging.

Image processing used Fiji v2.1.0 (NIH) and flow cytometry used FlowJo 10 (Becton Dickinson). Background-subtraction and normalization to control mean fluorescence was performed in some cases. Max/min window-level and acquisition parameters were applied consistently within experiments. In Fig. 1, FTC cells were quantified in a separate cohort repeated with, normalized to, and windowed to Nthy-ori-3–1 uptake. Data are means ± standard error unless noted. Analyses were in Excel (Microsoft) or Prism 9.0 (GraphPad). P values in captions used multiple hypothesis correction, where noted and α = 0.05. The ratio of fucodian inhibition to EIPA inhibition followed Δ fucoidan/Δ EIPA = (mean_control − mean_fucoidan)/(mean_control − mean_EIPA), after normalizing mean_control = 1 for the given experiment; standard error propagation assumed independence.

scRNA-seq (GSE148673; ref. 35) was pooled across patients using published cell-type annotations, and SPRING was used for visualization (RRID:SCR_023578; ref. 36). Myeloid scRNA-seq was analyzed from the published scDVA browser (http://cancer-pku.cn:3838/CRC_Leukocyte/; ref. 37). The Cancer Genome Atlas (TCGA, Firehose Legacy) database was analyzed using cBioPotal (RRID:SCR_014555, accessed 02–2023; ref. 38). Mouse PTC and ATC gene expression (GSE55933; ref. 39) and shRNA screening data, were downloaded from the supplementary of the publications (40). WebGestalt (RRID:SCR_006786; ref. 41) analyzed gene set enrichment with gene ID mapping (ref. 42; Supplementary Table S2B–S2D).

Data generated in this study are available from corresponding authors upon request. RNA sequencing and screening data were downloaded as described above (GSE148673, GSE55933, http://cancer-pku.cn:3838/CRC_Leukocyte/, and Supplementary File download from the publication; ref. 40).

In vitro dextran uptake marks non-specific fluid-phase macropinocytosis in tumor cells (43), because carcinomas typically lack expression for receptor-mediated dextran endocytosis (refs. 44–46; Fig. 1A–C). One half of PTC cell lines showed elevated dextran uptake compared with immortalized non-malignant follicular cells (Nthy-ori-3–1; Fig. 1A–C). All FTC and ATC cell lines showed higher dextran uptake compared with PTC and Nthy-ori-3–1 cells. In ATC, punctate dextran consistent with macropinocytosis (Fig. 1D; refs. 6, 13) reduced 66% when treated with EIPA [5-(N-ethyl-N-isopropyl)amiloride], which inhibits macropinocytosis (Supplementary Fig. S1A; ref. 47).

Serum albumin is an abundant macropinocytosis substrate and an important vehicle for drug delivery (48). Fluorescent albumin was internalized by ATC cells, co-localized with co-administered dextran, and its uptake was similarly reduced 62% by EIPA (Fig. 1D–G). Both human ATC cells and cells from a genetically engineered mouse model of Braf^V600E p53^−/− ATC (TBP3743 cells), showed EIPA-dependent uptake of albumin and dextran (Supplementary Fig. S1B). Co-imaging of albumin with geminin, which accumulates through S/G₂–M cell-cycle phases, revealed slight dependence of albumin uptake on the cell-cycle (Supplementary Fig. S1C and S1D). Overall, macropinocytosis is consistently upregulated in FTC and ATC, and is a major uptake pathway of serum albumin in ATC.

We assessed the degree to which macropinocytosis contributed to albumin uptake in ATC compared to FTC, PTC, and follicular thyroid cells (Fig. 2A; Supplementary Fig. S1E). Albumin and dextran uptake did not correlate across cell lines (P = 0.27), and follicular thyroid cells surprisingly accumulated high albumin but low dextran (Fig. 2A). We hypothesized that receptor-mediated endocytosis might explain differences between dextran versus albumin uptake. Fucoidan is used to block receptor-mediated albumin endocytosis (49, 50), and we compared its effects with those of EIPA across follicular thyroid, PTC, and ATC cells (Fig. 2B; Supplementary Fig. S1F–S1I). In all cell lines, dextran uptake was reduced by EIPA but not fucoidan, indicating macropinocytosis. In contrast, fucoidan blocked albumin uptake as much as EIPA in follicular thyroid and PTC cells, indicating substantial receptor-mediated uptake. In ATC cells, albumin uptake was less affected by fucoidan than EIPA, therefore suggesting albumin uptake primarily through macropinocytosis in ATC.

We evaluated the degree to which ATC tumors accumulated albumin in vivo, compared with healthy thyroid. Albumin biodistribution was assessed 24 hours after intravenous injection, showing high orthotopic ATC allograft uptake of 8.8% injected dose per gram tissue (%ID/g) and a tumor:muscle uptake ratio of 27:1 (Fig. 2C and D). In contrast, healthy thyroids accumulated little albumin. Hepatic uptake was found and expected for albumin, which undergoes hepatic clearance (48). Compared with tumor-free mice, tumor-bearing mice showed increased albumin in the kidney, heart, and spleen (Fig. 2C and D). These data and hypoalbuminemia in tumor-bearing mice (Supplementary Fig. S2A; refs. 51, 52) are possibly suggestive of systemic capillary leak, for instance, arising from an altered systemic inflammation (53), and merits future investigation. Nonetheless, high tumor accumulation of albumin is consistent with constitutive macropinocytosis in ATC.

We hypothesized that albumin-bound chemotherapy may efficiently block ATC progression, because albumin accumulated in ATC within ranges seen for tumor-targeted antibodies (54) and nanotherapies exploiting the enhanced permeability and retention (EPR) effect (55) in other cancer types. Microtubule-targeting paclitaxel is clinically used to treat ATC (28) but exhibits toxicities. To efficiently deliver microtubule-targeted drugs to ATC, we synthesized an albumin–drug conjugate comprising the MMAE conjugated to serum albumin with a cathepsin-cleavable linker (Alb-vc-MMAE; Fig. 2D), and compared it with unconjugated MMAE. Albumin conjugation decreased MMAE cytotoxicity in vitro, consistent with conjugation stability over the 72 hours experiment: free MMAE is permeable and does not require slower, rate-limiting transport via macropinocytosis to reach its cytoplasmic target (Supplementary Fig. S2B; ref. 56). However, in vivo free MMAE is rapidly cleared from the body, and we hypothesized that macropinocytosis of Alb-vc-MMAE could promote activity via extended circulating pharmacokinetics and ATC macropinocytic uptake. Free MMAE showed no effect on ATC tumor growth, whereas equimolar Alb-vc-MMAE shrank all tumors (>90% average volume decrease; Fig. 2E–G) and no substantial toxic effects by blood chemistry (albumin, alkaline phosphatase, alanine transaminase, glucose, bilirubin), liver histology, or weight loss were observed (Supplementary Fig. S2A, S2C–S2E). Engineered albumin binding improved responses in ATC tumors with high macropinocytosis activity.

We next examined how oncogenic mutations and phosphosignaling influence macropinocytosis in thyroid cancers. Transfection of follicular thyroid or BRAF^WT PTC cells with BRAF^V600E induced p-ERK1/2 and enhanced uptake of dextran and albumin (Fig. 3A–C; Supplementary Fig. S3A and S3B). To relate ERK activity with macropinocytosis on a cell-by-cell level, we generated a reporter BRAF^V600E ATC cell line based on a fluorescent kinase translocation reporter that accumulates in the cytoplasm or nucleus depending on whether its canonical ERK substrate is phosphorylated. The BRAF inhibitor dabrafenib, the MEK1/2 inhibitor trametinib, and their combination reduced both ERK activity and albumin uptake in ATC cells (Fig. 3D–F) that were confirmed as viable (Supplementary Fig. S3C and S3D). A total of 0.5 μmol/L dabrafenib incompletely blocked ERK activity and was less effective than 0.5 μmol/L trametinib, despite similar biochemical potencies to their targets (∼1 nmol/L; refs. 57, 58). This suggests potential dabrafenib-resistant ERK signaling, for instance, through RTKs and CRAF. Taken together, downstream MAPK/ERK pathway output controls ATC macropinocytosis.

We tested whether other prominent mutations in thyroid cancers could promote macropinocytosis. RET rearrangements stimulate MAPK/ERK signaling in PTC, and treating RET/PTC-expressing PTC cells with a kinase inhibitor targeting RET (pralsetinib) reduced p-ERK1/2 and decreased uptake of dextran and albumin by 20% and 24%, respectively (Fig. 3G–I). RAS mutations are common in FTC, and constitutively activated KRAS^G12D promotes macropinocytosis in cancer types, including pancreatic adenocarcinoma (1, 6). Ectopic KRAS^G12D enhanced p-ERK1/2 and dextran uptake compared with ectopic KRAS^WT in follicular thyroid cells, but albumin uptake was not different (Supplementary Fig. S3E–S3G), indicating that KRAS^G12D-promoted macropinocytosis and a distinct receptor-mediated albumin endocytosis response in these cells.

Because FTC cell lines showed elevated macropinocytosis, carried truncated loss-of-function PTEN mutation, but harbored no known RAS or BRAF mutation, we tested whether PTEN/PI3K perturbation could impact FTC macropinocytosis. Genetic PTEN silencing enhanced dextran and albumin uptake in follicular thyroid cells by 16% and 100%, respectively (Supplementary Fig. S3H–S3J). In FTC cells with PTEN loss, kinase inhibitors targeting PI3Kα (serabelisib and alpelisinib) showed mixed results: serabelisib reduced dextran uptake by 45% on average, but alpelisinib had no effect, despite its equimolar dose, similar potency, and shared mechanism (Supplementary Fig. S3K–S3N). Altogether, MAPK–ERK or PI3K/PTEN dysregulation via diverse oncogenic mutations can promote macropinocytosis in thyroid cancers.

We next tested how targeted BRAF and MEK kinase inhibition of the MAPK–ERK pathway (MAPKi) could affect albumin macropinocytosis in ATC tumors. Tumor-bearing mice were treated with MAPKi 24 hours before intravenous injection of albumin, which reduced its uptake in malignant cells by 35% and correlated with decreased MAPK–ERK activity (Fig. 4A–C). Trametinib alone, and combined with dabrafenib, decreased albumin uptake in tumors but not other tissues (Fig. 4D; Supplementary Fig. S4A and S4B). MAPKi reduced average sizes of tumors and their albumin uptake, but lectin and dextran accumulation—as markers of functional tumor vasculature and phagocytic macrophages—did not change (Supplementary Fig. S4C and S4D). Thus, reduced albumin uptake was not merely due to a reduced ability to passively access tumor tissue.

Flow cytometry quantified how MAPKi affected partitioning of serum albumin within malignant and immune cell populations. Twenty-four hours after albumin administration, tumors were digested into single-cell suspensions; immunostained for TAMs (CD45⁺ CD11b⁺ F4/80⁺), other myeloid cells (CD45⁺ CD11b⁺ F4/80⁻) and lymphocytes (CD45⁺ CD11b⁻); and quantified for cellular abundance and albumin uptake (Supplementary Fig. S4E and S4F). TAMs accumulated highest albumin on a per-cell basis (Supplementary Fig. S4F), consistent with confocal microscopy showing both malignant and especially phagocytic host cell uptake (ref. 59; Fig. 4E). Nonetheless, most albumin within control-treated ATC tumors was found in malignant cells, because they were the most abundant cell type (Fig. 4F–I). MAPKi decreased malignant cell uptake by 87% (P < 0.01; Fig. 4H), whereas TAM uptake was unchanged (P = 0.63; Fig. 4G). Treated tumors showed a greater relative TAM abundance (P < 0.01; Fig. 4I), and no difference was found by flow cytometry in overall cell viability following MAPKi (Supplementary Fig. S4G). MAPKi thus reduced albumin uptake in malignant ATC cells and shifted its relative accumulation to TAMs.

We next examined potential mechanisms to further increase macropinocytosis in ATC. We analyzed hits from a genome-wide shRNA screen of bladder cancer macropinocytosis enhancement. The screen originally reported a gene-set enrichment of its 571 hits in KEGG signaling pathways, including for insulin and MAPK (40); in a subsequent re-analysis, we found these and additional enrichment in pathways, including for endocrine resistance, AMPK, FoxO, and RAS, at lower significance levels (P < 0.05, but FDR adjusted q > 0.05; Supplementary Table S2B–S2D). We also found high-ranking enrichment in disease-associated gene-sets, including for “body weight” and “impaired glucose tolerance” (Supplementary Table S2B–S2D). We tabulated RNA expression of the top 105 screen hits (40) among 397 PTC samples in TCGA, and found IGF1R as the third highest hit based on median expression (Fig. 5A; Supplementary Table S2E). To investigate these pathways in ATC macropinocytosis, we performed a targeted pharmacologic screen to assess how 17 different compounds affected ATC albumin uptake (Fig. 5B and C) using phenformin and metformin as agents known to activate AMPK; 2-deoxy-D-glucose (2DG), lonidamine, and 3-bromopyruvate (3BP) as glycolysis disrupting agents; oxythiamine as an antimetabolite transketolase inhibitor; 6-diazo-5-oxo-L-norleucine (DON), BPTES, and V9302 as compounds modulating glutamine; and several targeted kinase inhibitors, including linsitinib and AXL1717/picropodophyllin as compounds inhibiting IGF1R (Supplementary Table S2A). Metformin, phenformin, and inhibitors targeting IGF1R were the only agents to consistently enhance albumin uptake across both human and mouse ATC cells (Fig. 5B and C). Combined AXL1717 and metformin did not elicit greater effects compared with single-agent treatments, potentially suggesting overlapping effects (Fig. 5D).

AXL1717 and/or metformin failed to affect bulk tumor accumulation of albumin in ATC (Fig. 5E), and we investigated the disparity between monoculture and in vivo results. IGF1R silencing using siRNA in ATC cells reduced IGF1R and its phosphorylation to undetectable levels, and increased both albumin and dextran uptake (Fig. 5F and G; Supplementary Fig. S5A and S5B). IGF1R siRNA did not further enhance the effects of AXL1717 on macropinocytosis, and AXL1717 slightly decreased basal p-IGF1R (Fig. 5G and H; Supplementary Fig. S5A–S5C). These data support basal IGF1R signaling as suppressing macropinocytosis in ATC monoculture.

We hypothesized that ATC response in vivo could be affected by growth factors supplied by other cells in the tumor microenvironment. Macrophage-derived IGF1 has been reported as redirecting phagocytosis in other contexts (60), and we detected IGF1 in ATC tumors (Supplementary Fig. S5D). Therefore, we tested whether IGF1 addition could block effects of IGF1R inhibition. AXL1717 and IGF1R siRNA treatments failed to enhance macropinocytosis in ATC cells in the presence of recombinant IGF1, which is known to bind both IGF1R and insulin receptor (Fig. 5F). AXL1717 failed to reduce exogenous IGF1-stimulated IGF1R phosphorylation (Fig. 5G and H), and published data show that AXL1717 only partially inhibits IGF-stimulated p-IGF1R in other cancer types (61). IGF1 treatment decreased p-AMPK regardless of AXL1717 or siRNA treatment (Fig. 5I; Supplementary Fig. S5C). In effect, IGF1 neutralized the ability of IGF1R inhibition to enhance macropinocytosis in ATC.

We analyzed possible sources of IGF1 signaling in thyroid cancers. From bulk tissue, IGF1R protein and mRNA levels in TCGA data revealed that IGF1R levels were 2-fold higher in BRAF-mutant compared with BRAF wild-type thyroid tumors (Supplementary Fig. S5E and S5F). IGF1 was expressed at lower levels than IGF1R, and did not correlate with BRAF mutation (Supplementary Fig. S5G and S5H). In genetically engineered mouse models, Igf1r and Igf1 were both highly expressed in Braf-mutant PTC and ATC tumors (Supplementary Fig. S5I). IGF1 and IGF1R expressions in ATC tumors from five patients, using published scRNA-seq (35), revealed both malignant and myeloid cell populations expressed IGF1, particularly among myeloid cells exhibiting M2-like macrophage gene expression patterns (CD206+ cells in ATC consistent with C1QC⁺ macrophages identified in thyroid cancer and other cancer types; Fig. 6A and B; Supplementary Fig. S6; ref. 37). In contrast, IGF1R was more highly expressed in malignant but not myeloid cells in ATC and other cancer types (Fig. 6C and D; Supplementary Fig. S6). IGF1R is thus highly expressed in BRAF-mutant thyroid cancers, including ATC, and mixed IGF1 expression is detected in both malignant and myeloid cells within the ATC tumor microenvironment.

To assess whether TAMs could directly influence IGF1R-dependent ATC macropinocytosis, we co-cultured IL4-polarized BMDMs [IL4 BMDMs, known to express CD206 (32) and IGF1 (62)] with murine ATC (TBP3743) cells. AXL1717 and an IGF1R-blocking antibody, both separately and in combination, increased albumin uptake more in TBP3743 monoculture than in BMDM co-culture (Fig. 6E). These data show how IGF1, macrophages, and potentially other macrophage-supplied signaling factors influence ATC macropinocytosis to suppress the effects of IGF1R inhibition (Fig. 6F).

Thyroid cancer often carries MAPK/ERK and PI3K pathway mutations and is the fifth most common cancer in women worldwide, but little has been described about macropinocytosis in this disease. Here, we identified elevated constitutive macropinocytosis in 5/5 combined FTC and ATC cell lines tested, but macropinocytosis was lower and more variable in the two PTC cells tested. We focused particularly on aggressive BRAF^V600E ATC, and treatment with kinase inhibitors targeting BRAF or MEK reduced macropinocytosis in BRAF-mutant ATC. ATC accumulated serum albumin via macropinocytosis, syngeneic ATC tumors in mice accumulated >30-fold greater albumin than healthy thyroid, and combined BRAF and MEK inhibition reduced macropinocytosis in cancer cells but not macrophages or other tumor-associated leukocytes. Macropinocytic albumin uptake thus depended on BRAF^V600E-driven MAPK/ERK activity in ATC, and these data raise questions about whether BRAF^V600E similarly promotes macropinocytosis in other cancer types where BRAF^V600E is common.

Despite our focus on BRAF^V600E, its presence or absence was not the sole determinant of macropinocytosis in thyroid cancer. Ectopic BRAF^V600E stimulated macropinocytosis in follicular thyroid cells, but the mere existence of BRAF^V600E failed to induce macropinocytosis in BRAF^V600E BCPAP PTC cells. BCPAP are reported to exhibit lower constitutive p-ERK1/2 compared with the more aggressive BRAF^V600E ATC cell lines studied here (63). Although this suggests that downstream MAPK/ERK output correlates with macropinocytosis, we found counter-examples in comparing RET/PTC versus BRAF-mutant PTC. BRAF^V600E associates with higher MAPK/ERK activity than as promoted by RET rearrangement in PTC (64), but TPC1 cells harboring RET fusion exhibited higher macropinocytosis than BCPAP, suggesting additional regulatory factors. In contrast with PTC, FTC frequently harbors RAS and PTEN mutations (23–24), and we found that both PTEN silencing and ectopic mutant KRAS could promote macropinocytosis in follicular thyroid cells. Taken together, we found consistent macropinocytosis in FTC and ATC, promoted through diverse routes of phosphosignaling dysregulation.

ATC accumulated high levels of serum albumin by macropinocytosis, and such albumin uptake can serve as an amino acid nutrient source in RAS-mutant cancers, including KRAS^G12D pancreatic adenocarcinoma (1–6); nonetheless, multiple mechanisms beyond macropinocytosis govern serum albumin transport and uptake throughout the body, including via receptor-mediated endocytosis, transcytosis, and FcRn (neonatal Fc receptor) recycling. Both class A and B scavenger receptors can mediate the uptake or transcytosis of albumin and chemically modified versions thereof (48). Comparing the effects of fucoidan with EIPA showed how multiple routes of albumin transport explain the imperfect correlation between cellular accumulation of dextran compared with albumin across thyroid cells, and reinforced macropinocytosis as a prominent uptake pathway in ATC (Fig. 2A and B). In vivo, nonspecific convection/diffusion processes, including factors of EPR seen heterogeneously in patients, also influence the transport of albumin and other macropinocytosis substrates. Comparing albumin uptake in the normal thyroid with uptake in ATC, along with multichannel imaging and flow cytometry of the ATC tumor microenvironment, all showed data consistent with cancer cell macropinocytosis as a key driver of albumin tumor accumulation.

Constitutive macropinocytosis has been therapeutically targeted as a pathway that supports malignant cell survival and proliferation (1–4), and here we demonstrate that clinically used BRAF and MEK kinase inhibitor therapy efficiently downregulates macropinocytosis in ATC. Future studies may more closely examine mechanisms of how MAPK/ERK inhibition affects F-actin dynamics to limit macropinocytosis at the cell surface and may consider strategies to capitalize on potential drug sensitivities created by reduced macropinocytosis. On the other hand, high macropinocytosis may promote the delivery of highly albumin-bound, nanoparticle-based, or otherwise cell impermeable therapies that rely on macropinocytosis for their delivery (6, 17–19). As a proof of principle, we demonstrated how covalent albumin-binding of the chemotherapy MMAE enhanced efficacy in treating ATC allografts; such albumin conjugation confers advantages in extending the circulating pharmacokinetic half-life of its therapeutic payload, and facilitates tumor uptake through factors, including EPR effects and macropinocytosis. BRAF^V600E-driven macropinocytosis may thus facilitate more efficient drug delivery while minimizing off-target uptake and toxicity. This result motivates future studies to quantify the degree to which macropinocytosis occurs in patients with ATC, for instance, via radiologic imaging of albumin uptake by PET (48, 65), or by histologic assessment of labeled albumin uptake during tumor resection (66). Such studies may inform biomarker strategies to predict patients likely to respond to macropinocytosis-dependent therapy.

H. Hu reports grants from China Scholarship Council during the conduct of the study. R. Li reports grants from American Cancer Society during the conduct of the study. M.A. Miller reports grants from US National Institutes of Health during the conduct of the study. M.A. Miller also reports other support from January Therapeutics, non-financial support from Pfizer and Genentech/Roche, and grants from Ionis Pharmaceuticals outside the submitted work; M.A. Miller also reports a patent for Massachusetts General Hospital pending and issued and a patent for Massachusetts Institute of Technology issued. No disclosures were reported by the other authors.

H. Hu: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. T.S.C. Ng: Visualization, methodology, writing–review and editing. M. Kang: Data curation, visualization, writing–review and editing. E. Scott: Data curation, visualization, methodology, writing–review and editing. R. Li: Data curation, visualization, writing–review and editing. J.M. Quintana: Visualization, methodology, writing–review and editing. D. Matvey: Visualization, methodology, writing–review and editing. V.R. Vantaku: Visualization, writing–review and editing. R. Weissleder: Resources, writing–review and editing. S. Parangi: Resources, supervision, funding acquisition, writing–review and editing. M.A. Miller: Conceptualization, resources, supervision, funding acquisition, investigation, visualization, writing–original draft, project administration, writing–review and editing.

This work was supported in part by NIH grants DP2CA259675 and R00CA207744 (to M.A. Miller), T32CA079443 (to J.M. Quintana), U01CA206997 (to R. Weissleder), R01GM138790 (to M.A. Miller), and the President and Fellows of Harvard College (to S. Parangi). H. Hu was supported by the China Scholarship Council 201906370164. M. Kang received support from the National Research Foundation of S. Korea. R. Li was supported in part by American Cancer Society Postdoctoral Fellowship PF-20–106–01-LIB. The authors are grateful to Dr. M.A. Garlin, Yoshiko Iwamoto, and the MGH Histopathology Research Core for assistance with liver histology, as well as Gregory Wojtkiewicz for assistance through the MGH-CSB MIP program.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.