Research ArticleCANCER

Increased lysosomal biomass is responsible for the resistance of triple-negative breast cancers to CDK4/6 inhibition

See allHide authors and affiliations

Science Advances  17 Jun 2020:
Vol. 6, no. 25, eabb2210
DOI: 10.1126/sciadv.abb2210


Inhibitors of cyclin-dependent kinases CDK4 and CDK6 have been approved for treatment of hormone receptor–positive breast cancers. In contrast, triple-negative breast cancers (TNBCs) are resistant to CDK4/6 inhibition. Here, we demonstrate that a subset of TNBC critically requires CDK4/6 for proliferation, and yet, these TNBC are resistant to CDK4/6 inhibition due to sequestration of CDK4/6 inhibitors into tumor cell lysosomes. This sequestration is caused by enhanced lysosomal biogenesis and increased lysosomal numbers in TNBC cells. We developed new CDK4/6 inhibitor compounds that evade the lysosomal sequestration and are efficacious against resistant TNBC. We also show that coadministration of lysosomotropic or lysosome-destabilizing compounds (an antibiotic azithromycin, an antidepressant siramesine, an antimalaria compound chloroquine) renders resistant tumor cells sensitive to currently used CDK4/6 inhibitors. Lastly, coinhibition of CDK2 arrested proliferation of CDK4/6 inhibitor-resistant cells. These observations may extend the use of CDK4/6 inhibitors to TNBCs that are refractory to current anti-CDK4/6 therapies.


D-type cyclins together with their associated kinase partners, the cyclin-dependent kinases CDK4 and CDK6, drive cell cycle progression by phosphorylating the retinoblastoma protein, RB1, and RB1-related p107 and p130 proteins. During early G1 phase of the cell cycle, RB1 exists in a hypophosphorylated state and constrains cell proliferation by binding to and inhibiting the activity of E2F transcription factors. Phosphorylation of RB1 by cyclin D–CDK4/6 and later by cyclin E–CDK2 kinases functionally inactivates RB1, resulting in derepression of the E2F activity. This, in turn, allows progression of cells into the DNA synthesis phase (S phase) (1, 2).

Consistent with their growth-promoting roles, overexpression of cyclin D–CDK4/6 kinases has been documented in a large number of human malignancies (3, 4). In particular, the cyclin D1-CDK4 kinase is frequently hyperactivated in mammary carcinomas. Thus, amplification of the CCND1 gene (encoding cyclin D1) takes place in up to 20% of breast cancers, while cyclin D1 protein is overexpressed in more than 50% of cases (5). Additional breast cancer cases overexpress CDK4 or contain hyperactive cyclin D1-CDK4 complexes due to silencing of an inhibitor of CDK4 and CDK6, p16INK4a, or up-regulate cyclin D1 due to aberrant signaling within pathways that impinge on cyclin D1 (such as the estrogen receptor– or human epidermal growth factor receptor 2 (HER2)–dependent signaling) (4). Collectively, all these observations suggested that inhibition of cyclin D–CDK4/6 kinase might represent an attractive therapeutic strategy for breast cancer treatment. Indeed, our group has demonstrated that mice lacking cyclin D1, or CDK4, are completely resistant to the development of breast cancers driven by the HER2 oncogene (6, 7). We also demonstrated that a global shutdown of cyclin D1 in mice bearing mammary carcinomas halted tumor growth (8).

In a seminal study, Finn et al. (9) tested the efficacy of a chemical inhibitor of CDK4 and CDK6 kinases, palbociclib, across a wide panel of human breast cancer cell lines representing different breast cancer subtypes. They demonstrated that luminal-type breast cancer cells expressing the estrogen receptor (ER+) were most sensitive to treatment with palbociclib. In contrast, triple-negative breast cancer (TNBC) cell lines were refractory to CDK4/6 inhibition (9).

These observations paved the way for clinical trials with CDK4/6 inhibitors in patients with hormone (estrogen and progesterone) receptor–positive breast cancers. In these trials, addition of CDK4/6 inhibitors to standard endocrine therapy (an aromatase inhibitor letrozole or an ER-antagonist fulvestrant) resulted in doubling of progression-free survival (10). Moreover, a recent analysis demonstrated that cotreatment with ribociclib also significantly extended the overall survival (11). Consequently, three CDK4/6 inhibitors, palbociclib, ribociclib, and abemaciclib, received “breakthrough therapy” designation status from the U.S. Food and Drug Administration (FDA) and have been approved for treatment of ER+ breast cancers.

In contrast to hormone receptor–positive breast cancers, which initially respond to CDK4/6 inhibitor treatment, TNBC show an intrinsic resistance to CDK4/6 inhibition (12). Consequently, treatment with CDK4/6 inhibitors does not seem to represent a viable option for patients with TNBC. This breast cancer subtype is associated with particularly poor prognosis, and no targeted therapies are available to treat TNBC disease. Hence, new therapeutic approaches are urgently needed to improve the treatment of patients with TNBC.

Here, we present evidence that, in contrast to the traditional paradigm, a subset of TNBC critically requires CDK4 and CDK6 for their proliferation. However, in TNBC, palbociclib and other currently available CDK4/6 inhibitors become sequestered into tumor cell lysosomes because of the very high number of lysosomes in this cancer type. We provide several independent therapeutic strategies to overcome the lysosomal sequestration and to render TNBC fully sensitive to CDK4/6 inhibition.


A subset of TNBC depends on CDK4/6 for proliferation but is resistant to CDK4/6 inhibition

It has been assumed that TNBC are intrinsically resistant to CDK4/6 inhibition because of frequent inactivation of the retinoblastoma protein, RB1, in this tumor type (12, 13). We revisited this notion by analyzing a collection of TNBC patient-derived xenografts (PDX) from the Dana-Farber/Harvard Cancer Center collection. We observed that RB1 deficiency is present in about 45% of cases (fig. S1A). We therefore hypothesized that in the remainder of TNBC cases, a different molecular mechanism may be responsible for the resistance to CDK4/6 inhibition.

To extend these observations, we tested the response of a panel of TNBC cell lines to treatment with a CDK4/6 inhibitor, palbociclib. As expected, most of TNBC cell lines were resistant to CDK4/6 inhibition (Fig. 1A). Two of these cell lines (MDA-MB-468 and HCC1937) carry inactivating mutations of the RB1 gene, which explains their lack of response. However, several TNBC cell lines displayed resistance to CDK4/6 inhibition in the absence of any obvious abnormalities in the RB1 pathway. We verified that these cell lines were also resistant to treatment with two other FDA-approved CDK4/6 inhibitors, namely ribociclib and abemaciclib (fig. S1B).

Fig. 1 Sequestration of palbociclib into tumor cell lysosomes mediates resistance to chemical CDK4/6 inhibition.

(A) Fraction of bromodeoxyuridine (BrdU)–positive cells treated with palbociclib (PALBO) (1 μM) or dimethyl sulfoxide (DMSO) for 24 hours (means ± SD, n = 3). (B) Fraction of BrdU-positive cells transfected with anti-CDK4/CDK6 or control siRNA for 48 hours (means ± SD, n = 3; HCC1954, n = 2). (C and D) Microscopic analysis of HCC1806 cells treated with palbociclib (1 μM) or DMSO for 24 hours and stained with LysoTracker Green (LTR-green) (C), or treated with palbociclib or palbo/bafilomycin A1 (BAF) (100 nM) for 24 hours (D). PALBO auto., palbociclib autofluorescence. Scale bars, 20 μm. (E) Fraction of BrdU-positive cells treated with palbociclib (1 μM) and/or bafilomycin A1 (10nM-SUM149, 25nM-HCC1806/SUM149, 50nM-CAL120) or DMSO for 24 hours (means ± SD, n = 3, two-sided t test). (F) TNBC cells transfected with anti-ATP6AP1 or control siRNAs for 36 hours, stained with LysoSensor Green, and analyzed by fluorescence-activated cell sorting (FACS). (G) BrdU-positive fraction of ATP6AP1-depleted and control cells treated with palbociclib (1 μM) or DMSO for 24 hours (means ± SD, n = 3, two-sided t test). (H) Fraction of BrdU-positive cells treated with palbociclib (1 μM) and/or NH4Cl (50 mM) or DMSO for 24 hours (means ± SD, n = 3, two-sided t test). (I) Fraction of BrdU-positive cells treated with palbociclib, ribociclib (RIBO), abemaciclib (ABEMA) (1 μM), and/or bafilomycin A1 (25 nM) for 24 hours (means ± SD, n = 3, two-sided t test). (J) Fraction of BrdU-positive cells in nontargeting single-guide RNA (snt) or RB1-deleted cells (ΔRB1) treated with palbociclib (1 μM) and/or bafilomycin A1 (25 nM) for 24 hours (means ± SD, n = 3, two-sided t test).

To evaluate the requirement for CDK4 and CDK6 in these resistant TNBC cells, we depleted CDK4 and CDK6 using two independent sets of small interfering RNAs (siRNAs). Very unexpectedly, three of the CDK4/6 inhibitor–resistant TNBC cell lines (HCC1806, SUM149, and SUM159) showed a nearly complete proliferative arrest following CDK4/6 depletion (Fig. 1B and fig. S1C). A CRISPR screen for essential genes in a fourth cell line (CAL120) also revealed that these cells depend on CDK4 for proliferation (R.J. and M.B., unpublished observations). We made a similar observation in basal-like, HER2-positive HCC1954 cells. These cells were resistant to treatment with all three CDK4/6 inhibitors, while depletion of CDK4/6 arrested their proliferation (Fig. 1B and fig. S1, C and D). Hence, these TNBC cell lines, like hormone receptor–positive breast cancer cells, critically require CDK4 and CDK6 for their proliferation, and yet, they are resistant to treatment with all available CDK4/6 inhibitors.

To explain these findings, we hypothesized that in resistant breast cancer cells, palbociclib fails to reach its targets (CDK4 and CDK6) in the nucleus. We took advantage of our observation that palbociclib has autofluorescent properties, and we followed the localization of this compound in TNBC cells by light microscopy. Live-cell imaging of TNBC (and basal-like HCC1954 cells) revealed that upon palbociclib treatment, this compound accumulates in discrete structures in tumor cells cytoplasm. Costaining of cells with LysoTracker Green, a green fluorescent dye that stains lysosomes, revealed a complete overlap with palbociclib staining (Fig. 1C and fig. S2A).

We next asked whether inhibition of lysosomal acidification by blocking the lysosomal proton pump would trigger the release of palbociclib from lysosomes. To test this, we treated cells with bafilomycin A1, a specific inhibitor of the proton pump vacuolar (H+)–adenosine triphosphatase (ATPase) (v-ATPase), which mediates acidification of intracellular compartments (14). Inhibition of lysosomal acidification completely abrogated the lysosomal accumulation of palbociclib (Fig. 1D). Strikingly, bafilomycin A1 treatment rendered TNBC cells fully sensitive to CDK4/6 inhibition (Fig. 1E and fig. S2B), indicating that the lack of response to chemical CDK4/6 inhibition is because of the sequestration of these compounds into tumor cell lysosomes.

To confirm these findings, we used an alternative approach of inhibiting lysosomal acidification, namely, depletion of the ATP6AP1 protein, a component of the v-ATPase complex. Again, knockdown of ATP6AP1 abrogated lysosomal accumulation of palbociclib and rendered TNBC cells sensitive to palbociclib treatment (Fig. 1, F and G, and fig. S2C). Last, we treated TNBC cells with NH4Cl, which also decreases lysosomal acidification, and found that this also restored palbociclib sensitivity (Fig. 1H).

In contrast to palbociclib, other structurally related FDA-approved CDK4/6 inhibitor compounds show either no autofluorescence (abemaciclib) or only very weak autofluorescent properties (ribociclib), and hence, we could not determine their intracellular localization by microscopy. However, we observed that cotreatment of TNBC cells with bafilomycin A1 rendered these cells sensitive also to ribociclib and abemaciclib (Fig. 1I). Hence, we concluded that, in palbociclib-resistant TNBC cell lines, all three CDK4/6 inhibitors become sequestered into tumor cell lysosomes, and this can be overcome by decreasing lysosomal acidification.

To exclude an off-target effect of the drug combinations that could lead to the observed arrest, we engineered RB1−/− TNBC HCC1806 and SUM159 cells (fig. S2D). In contrast to parental, RB1-proficient cells, these cells remained refractory to CDK4/6 inhibition despite cotreatment with bafilomycin A1 (Fig. 1J). Also, naturally RB1-deficient TNBC cell lines (MDA-MB-468 and HCC1937) were insensitive to bafilomycin A1 cotreatment (fig. S2E). We concluded that the observed synergistic action of lysosomal inhibitors with palbociclib represents an on-target effect, allowing the CDK4/6 inhibitor to reach its targets.

Inhibition of lysosomal acidification is also known to inhibit autophagy. To test whether autophagy is involved in the resistance of TNBC cells to CDK4/6 inhibitors, we used CRISPR-Cas9 to knockout BECN1 or ATG7 in two TNBC cell lines. BECN1 is the master regulator of autophagy (15), while ATG7 represents an essential factor for autophagosome formation (16). We observed that ablation of autophagy had no effect on the resistance of TNBC cells to palbociclib treatment (fig. S2, F to H). Moreover, we did not observe any change in LC3 I to LC3 II ratio (indicative of altered autophagosome formation) following palbociclib treatment of either resistant or sensitive cells (fig. S2I). Collectively, these data rule out an inhibition of autophagy as the resistance mechanism in our system and indicate that sequestration of CDK4/6 inhibitors into the lysosomes underlies the resistance of TNBC cells to these compounds.

Resistant TNBC display high lysosomal biomass

We wished to understand why the lysosomal compartment of TNBC cells renders them resistant to chemical inhibition of CDK4/6. We hypothesized that either these cells contain more lysosomes or their lysosomes display an increased ability to sequester palbociclib (Fig. 2A). To distinguish between these possibilities, we treated palbociclib-sensitive and -resistant TNBC cells with palbociclib, immunoprecipitated the lysosomes (17), and quantified the amount of palbociclib by mass spectrometry. LysoTracker was used to normalize for the number of lysosomes. These analyses revealed that the amount of palbociclib per lysosome was not increased in resistant cells compared to sensitive cells (Fig. 2B). To test the hypothesis that these cells contain more lysosomes, we stained TNBC cells with LysoTracker Green and quantified the amount of lysosomes per cell by flow cytometry. We observed that the resistant cells displayed strongly increased lysosomal numbers, as compared to the sensitive cells (Fig. 2, C and D). We also quantified the numbers of lysosomes in TNBC cells by electron microscopy. Again, we observed that the resistant cell lines displayed substantially higher numbers of lysosomes per cell, as compared to sensitive cells (Fig. 2, E and F).

Fig. 2 Increased lysosomal mass in CDK4/6 inhibitor–resistant TNBC cells.

(A) Schematic representation of possible mechanisms of lysosome-mediated resistance. (B) Palbociclib/LysoTracker Red ratio in immunoprecipitated lysosomes, quantified by mass spectrometry (mean, n = 2, two-sided t test). (C and D) Representative histogram overlay of flow cytometric analysis (C) and median fluorescence intensity (MFI) (D) of TNBC cells stained with LysoTracker Green (means ± SD, n = 5, two-sided t test). (E) Representative electron microscopic images of resistant and sensitive TNBC cells. Arrowheads, lysosomes. Scale bars, 2 μm. (F) Quantification of lysosome number in electron microscopic images per 100 μm2 of cytoplasm (analyzed images: HCC1806, n = 21; SUM159, n = 16; HCC38, n = 21; HCC1143, n = 17). (G) Immunoblot analysis of TFEB protein levels in nuclear fractions. Proliferating cell nuclear antigen served as loading control (dashed line indicates splicing of nonadjacent parts of the same membrane). (H) Flow cytometric analysis of fluorescence intensity in TFEB-depleted (TFEBsi) and control cells stained with LysoTracker Green. (I) Fraction of BrdU-positive cells transfected with control or anti-TFEB siRNA for 48 hours and treated with palbociclib (1 μM) or DMSO for 24 hours (means ± SD, n = 3, two-sided t test). (J) Gene set enrichment analysis (GSEA) for GO (gene ontology) lysosomal lumen category in TNBC PDX tumors displaying intrinsic resistance to treatment with ribociclib versus in ribociclib-sensitive tumors. The response (stable/progressive disease; fig. S3) was from (19). FDR, false discovery rate; NES, normalized enrichment score.

The transcription factor EB (TFEB) represents the master regulator of lysosome biogenesis (18). Therefore, we compared the amounts of nuclear (active) TFEB between palbociclib-sensitive and -resistant TNBC cells. Consistent with an increased lysosomal mass in the resistant cells, we observed that these cells display markedly elevated levels of nuclear TFEB (Fig. 2G).

To test whether increased TFEB activity is causally linked to palbociclib resistance, we depleted TFEB in two palbociclib-resistant TNBC cell lines (fig. S2J). As expected, depletion of TFEB decreased the number of cellular lysosomes (Fig. 2H). Importantly, this rendered TNBC cells sensitive to palbociclib treatment (Fig. 2I). We concluded that the increased lysosomal biogenesis in a subset of TNBC cell lines, and the resulting lysosomal sequestration of CDK4/6 inhibitors, underlies resistance of these cells to these compounds.

To extend these observations to human TNBC tumors in vivo, we took advantage of a collection of PDXs of TNBC. Mice bearing these tumors were treated with a CDK4/6 inhibitor ribociclib, and the impact on tumor growth was evaluated (19). In our analysis, we focused on RB1-positive tumors and compared gene expression profiles of six ribociclib-sensitive versus 13 ribociclib-resistant TNBC tumors (fig. S3). Gene Ontology analysis revealed an up-regulation of transcripts belonging to “lysosomal lumen” category in the resistant tumors, pointing to an increased lysosomal biomass in the resistant TNBC in vivo (Fig. 2J).

Lysosomotropic agents that deacidify the lysosomes render TNBC cells sensitive to CDK4/6 inhibitors

As described above, we observed that cotreatment of cells with chemicals that counter the lysosomal acidification, such as bafilomycin A1 or NH4Cl, restored the sensitivity of TNBC cells to CDK4/6 inhibitors. We therefore asked whether combining CDK4/6 inhibitors with clinically relevant compounds that are known to have lysosomotropic properties might render TNBC sensitive to CDK4/6 inhibitors.

We first focused on a widely used antibiotic azithromycin, which has lysosomotropic characteristics and decreases lysosomal acidity (20). We tested whether treatment of TNBC cells with this antibiotic could overcome palbociclib resistance. We observed that treatment of CDK4/6 inhibitor–resistant TNBC cells with azithromycin rendered these cells sensitive to palbociclib (Fig. 3A).

Fig. 3 Lysosomotropic compounds restore the sensitivity of TNBC to CDK4/6 inhibitors.

(A) Cell numbers of in vitro cultured TNBC cells at time 0 and after 6 days of culture in the presence of DMSO, palbociclib (1 μM), and/or lysosomotropic antibiotic azithromycin (AZI; 50 μM) (means ± SD; SUM159, n = 6; HCC1806, n = 4; two-sided t test). (B) Similar assay as in (A); cells were treated with palbociclib (1 μM) and/or anxiolytic/antidepressant compound siramesine (SIRA; SUM159, 3.5 μM, and HCC1806, 2 μM) (means ± SD, n = 3, two-sided t test). (C and D) Short-term ex vivo culture of a patient-derived TNBC tumor (grade 3). Tumor fragments were cultured for 48 hours in the presence of DMSO, palbociclib (1 μM), siramesine (5 μM), or palbociclib (250 nM or 1 μM) and siramesine and pulsed with BrdU (4 hours). (C) Representative images of tumor sections stained for BrdU. Scale bars, 20 μm. (D) Percentage of BrdU-positive tumor cells. (E) Cell numbers of in vitro cultured TNBC cells at time 0 and after 6 days of culture in the presence of palbociclib (250 nM or 1 μM) and/or chloroquine (CQ; SUM159, 25 μM, and HCC1806, 15 μM) or DMSO (means ± SD; SUM159, n = 3; HCC1806, n = 4; two-sided t test). (F) Tumor weights of a TNBC PDX (IDC50) following 4 days of treatment of mice with palbociclib (50 mg/kg) and/or hydroxychloroquine (HCQ; 60 mg/kg) or vehicle only (CTRL). Two-sided t test.

We next tested the effects of an anxiolytic and antidepressant compound siramesine, which was shown to increase the lysosomal pH and to destabilize the lysosomes (21). Again, cotreatment with siramesine restored the sensitivity of resistant breast cancer cells to palbociclib (Fig. 3B). We extended these observations using short-term ex vivo cultures of a patient-derived TNBC tumor. We cultured tumor fragments with palbociclib or siramesine or the combination of these two compounds and evaluated the response. We observed that cotreatment with siramesine and palbociclib nearly completely arrested tumor cell proliferation (Fig. 3, C and D). Hence, addition of siramesine rendered the patient-derived tumor sensitive to palbociclib treatment.

Last, the widely used antimalaria drug chloroquine (and hydroxychloroquine) represents another lysosomotropic agent that accumulates in the acidic lysosomes and raises their pH. Hence, we treated palbociclib-resistant TNBC cells with palbociclib in combination with chloroquine and evaluated the effect on cell proliferation. As it was the case with azithromycin and siramesine, we observed that the combination treatment nearly completely arrested cell proliferation (Fig. 3E). The growth arrest was also seen when low concentrations of palbociclib (250 nM) were used, which closely corresponds to the maximum sustained serum concentration of this compound in patients (Fig. 3E) (22). To exclude an off-target effect, we evaluated the effects of palbociclib and chloroquine treatment using matched RB1+/+ and RB1−/− TNBC cell lines engineered by CRISPR-Cas9. RB1−/− cells remained resistant to the combination treatment, indicating that the observed synergistic effect is on target (fig. S4A).

To extend these studies to an in vivo setting, we used mice bearing PDXs of a highly invasive TNBC that was resistant to chemotherapy and radiation. We treated these mice with palbociclib with or without hydroxychloroquine and evaluated the impact on tumor burden. Because of the aggressiveness of the model, mice could only be treated for 4 days. We observed that while palbociclib alone impeded tumor growth, combination treatment with hydroxychloroquine significantly potentiated this effect and resulted in a strong growth inhibition (Fig. 3F).

Coinhibition of CDK2 and CDK4/6 blocks proliferation of TNBC cells

We reasoned that in palbociclib-treated TNBC cells, not all inhibitor becomes sequestered into the lysosomes. We therefore hypothesized that these cells proliferate with limited levels of CDK4/6 activity, which may render them uniquely susceptible to inhibition of CDK2, a kinase that cooperates with CDK4/6 in driving cell cycle progression (1). Currently, no CDK2-specific inhibitors are available. To circumvent this limitation, we used an analog-sensitive CDK2 that we have engineered and characterized (23). In analog-sensitive kinases, the large hydrophobic residue in the kinase active site is mutated to a smaller amino acid, thereby generating an enlarged adenosine triphosphate (ATP)–binding pocket that is not found in any wild-type kinases (24). The resulting “analog-sensitive” kinase can be potently and highly selectively inhibited by “bulky” inhibitors, such as 3MB-PP1, that occupy the enlarged ATP-binding pocket. Importantly, these bulky inhibitors do not inhibit any wild-type kinases in the mammalian kinome (24). Accordingly, we knocked out the CDK2 gene in two TNBC cell lines and ectopically expressed analog-sensitive CDK2 (Fig. 4A). We next treated these cells with palbociclib or with 3MB-PP1 (to inhibit CDK2) or with the combination of the two compounds and evaluated the effects on cell proliferation. We observed that treatment with either of these compounds alone had little effect on cell growth. Strikingly, a combined inhibition of CDK4/6 and CDK2 essentially halted proliferation of TNBC cells (Fig. 4, B and C). These observations indicate that coadministration of CDK2 inhibitors sensitizes palbociclib-resistant TNBC cells to CDK4/6 inhibition.

Fig. 4 Coinhibition of CDK2 and CDK4/6 arrests proliferation of TNBC cells.

(A) Immunoblot analysis of CDK2 protein levels in SUM159 and HCC1806 cells following CRISPR-Cas9–mediated knockout of CDK2 and ectopic expression of Flag-tagged analog-sensitive CDK2 (CDK2AS/sg); nt, cells expressing control nontargeting sgRNA. Tubulin served as loading control. (B) Fraction of BrdU-positive control HCC1806 and SUM159 cells (nt) or CDK2 knockout HCC1806 and SUM159 cells expressing analog-sensitive CDK2 (CDK2AS/sg) treated with palbociclib (1 μM) and/or 3MB-PP1 (1 μM; an inhibitor of analog-sensitive kinases) or DMSO (means ± SD, n = 4, two-sided t test). (C) Cell numbers of in vitro cultured control (nt) or CDK2 knockout cells expressing analog-sensitive CDK2 (CDK2AS/sg) at time 0 and after 6 days. Cells were cultured in the presence of palbociclib (250 nM or 1 μM) and/or 3MB-PP1 (1 μM) (means ± SD, n = 3, two-sided t test).

Generation of novel structurally altered CDK4/6 inhibitors with less basic characteristics

All approved CDK4/6 inhibitors, namely, abemaciclib, palbociclib, and ribociclib, display weakly basic properties owing to the presence of the piperazine group in their molecular structures. We reasoned that reducing the basicity of CDK4/6 inhibitors may overcome the lysosomal sequestration, thereby rendering these compounds efficacious against TNBC. To test this prediction, we designed and synthesized two novel compounds (Cpd-1 and Cpd-2), which represent close analogs of ribociclib that incorporate a nonbasic replacement of the piperazine ring system, namely, a 4-substituted piperidine connected to the pyridyl motif through an amide linkage (Fig. 5A). We determined that Cpd-1 and Cpd-2 have pKa (where Ka is the acid dissociation constant) values of 4.73 and 4.75, respectively, indicating that these compounds are more than 10,000-fold less basic than ribociclib (which has pKa of 8.6). Importantly, we observed that Cpd-1 and Cpd-2 potently inhibited CDK4 with good biochemical selectivity over CDK2 and CDK1 (fig. S4B).

Fig. 5 Generation and characterization of novel structurally altered CDK4/6 inhibitors with less basic characteristics.

(A) Structures of ribociclib and its less basic analogs Cpd-1 and Cpd-2. (B) Top, amount of ribociclib, Cpd-1 and Cpd-2 in immunoprecipitated lysosomes quantified by mass spectrometry. Middle, the abundance of LysoTracker-Red in pulldowns indicating amount of immunoprecipitated lysosomes. Bottom, mass spectrometric quantification of ribociclib, Cpd-1, and Cpd-2 in whole-cell lysates. a.u., arbitrary units. (C) Fraction of BrdU-positive TNBC cells cultured with ribociclib (1 μM), Cpd-1 (50 μM) or Cpd-2 (25 μM), or DMSO. RB1-negative HCC1937 and RB1-positive, CDK4/6-independent HCC70 cells served as control (means ± SD, n ≥ 3, two-sided t test). (D) Cell numbers at time 0 and after 6 days of in vitro culture; cells were treated as in (C) (means ± SD, n ≥ 4, two-sided t test). (E to G) Short-term ex vivo cultures of two TNBC tumors derived from different patients (grades 3 and 2). Tumor fragments were cultured for 48 hours in the presence of DMSO, ribociclib (1 μM), Cpd-1 (50 μM), or Cpd-2 (25 μM) and pulsed with BrdU for 4 hours. (E) Representative images of tumor sections stained for BrdU (TNBC grade 3). Scale bars, 20 μm. (F and G) Percentage of BrdU-positive tumor cells in TNBC grade 3 (F) and grade 2 (G).

To gauge incorporation of the new compounds into tumor cell lysosomes, we treated SUM149, HCC1806, and CAL120 TNBC cells with Cpd-1, Cpd-2, or ribociclib. We next immunoprecipitated the lysosomes and quantified the concentration of CDK4/6 inhibitors in these organelles by mass spectrometry. As expected, we observed a significant accumulation of ribociclib in tumor cell lysosomes. Strikingly, Cpd-1 and Cpd-2 failed to accumulate in the lysosomes, despite the fact that their intracellular concentration was comparable to that of ribociclib (Fig. 5B and fig. S4, C and D). These observations indicate that by reducing the basicity of the compounds, we were able to obtain CDK4/6 inhibitors that evade the lysosomal sequestration.

We then investigated the effects of these two less basic compounds on proliferation of TNBC cells that are resistant to all available CDK4/6 inhibitors. We observed that both compounds strongly inhibited the growth of TNBC cells (Fig. 5, C and D, and fig. S4E). To control for nonspecific off-target effects, we used the TNBC HCC1937 cells that harbor inactive RB1 and RB1-positive HCC70 cells, which do not require CDK4/6 for proliferation. Cpd-1 and Cpd-2 had no impact on proliferation of these cells in short-term assays (Fig. 5C), while a slight off-target effect was observed in long-term cultures (fig. S4, F and G).

We extended these observations using short-term ex vivo cultures of two patient-derived TNBC tumors. We treated ex vivo cultured tumor fragments with ribociclib or Cpd-1 or Cpd-2 and evaluated the impact on tumor cell proliferation. We observed that treatment with ribociclib had a modest effect on cell proliferation. In contrast, treatment with Cpd-1 or Cpd-2 essentially extinguished cell growth (Fig. 5, E to G). Collectively, these observations indicate that CDK4/6 inhibitor compounds with less basic characteristics escape the lysosomal sequestration, which renders them efficacious against TNBC that are resistant to all currently used CDK4/6 inhibitors.

Thr172 phosphorylation of CDK4 might be used to identify TNBC patients benefiting from improved CDK4/6 inhibition

For therapeutic purposes, it is very important to identify TNBC patients with tumors that are dependent on CDK4/6 for proliferation, since these patients might benefit from the improved CDK4/6 inhibition treatments outlined above. We took advantage of the observation made by Raspé et al. (25) that palbociclib-sensitive breast cancer cell lines are positive for phosphorylation of CDK4 at Thr172. We used immunoblotting to examine the presence/absence of CDK4 Thr172 phosphorylation in a large panel of palbociclib-resistant TNBC cell lines. We subdivided palbociclib-resistant cell lines into CDK4/6 dependent (cell cycle arrest upon CDK4/6 depletion) and CDK4/6 independent (no cell cycle arrest, for example, because of RB1 loss). We observed that all cell lines that depend on CDK4/6 for proliferation (including cell lines that are palbociclib-resistant because of lysosomal sequestration) were positive for phosphorylated CDK4 Thr172, while all CDK4/6-independent cells were negative (Fig. 6A). Immunohistochemical staining of xenografts from palbociclib-resistant, CDK4/6-dependent HCC1954 cells and palbociclib-resistant, CDK4/6-independent MDA-MB-468 cells confirmed the presence of CDK4 Thr172 phosphorylation in CDK4/6-dependent tumors (Fig. 6B). These observations suggest that immunostaining for Thr172 phosphorylation of CDK4 can be used to identify tumors that depend on CDK4/6 for proliferation and might hence benefit from improved therapeutic strategies described above.

Fig. 6 Phosphorylation of CDK4 at Thr172 as a predictor of CDK4/6-dependence of TNBC.

(A) Immunoblot analysis of phospho-Thr172 CDK4 and total CDK4 protein levels in the indicated TNBC cell lines and in the ER+ breast cancer cell line MCF7. Cells were classified as CDK4/6 dependent (cell cycle arrest upon depletion of CDK4/6) or CDK4/6 independent (no cell cycle arrest). The sensitivity (SEN) or resistance (RES) of cells to palbociclib is also indicated. GAPDH, glyceraldehyde-3-phosphate-dehydrogenase (loading control). (B) Phospho-Thr172 CDK4 immunohistochemical staining of xenografts of palbociclib-resistant, CDK4/6-dependent HCC1954 cells and palbociclib-resistant, CDK4/6-independent MDA-MB-468 cells. Scale bars, 50 μm. (C) A similar analysis as in (A), in a panel of TNBC PDXs from the Dana-Farber/Harvard Cancer Center PDX collection. GAPDH, loading control. (D and E) Ex vivo Matrigel culture of a TNBC PDX [17-01 from (C)], treated with palbociclib (1 μM) and/or chloroquine (CQ, 30 μM) or DMSO for 24 hours. (D) The fraction of 5-ethynyl-2’-deoxyuridine (EdU)–positive cells (DMSO, n = 5; palbociclib, n = 7; chloroquine, n = 3; chloroquine/palbociclib, n = 3; two-sided t test). (E) Representative images of cells stained for EdU. Scale bars, 50 μm. (F) GSEA for KEGG lysosome category in ER+/HER2 tumors displaying intrinsic resistance to palbociclib versus in palbociclib-responsive tumors. Gene expression data were from NeoPalAna trial.

To obtain some measure of the frequency of CDK4 Thr172 phosphorylation-positive TNBC cases, we analyzed the TNBC PDX collection from the Dana-Farber/Harvard Cancer Center. Immunoblot analysis revealed that a substantial fraction of human TNBC express Thr172-phosphorylated CDK4 (Fig. 6C).

To further confirm the validity of CDK4 phosphorylation as a marker of the therapeutic response, we chose one CDK4 Thr172 phospho-positive PDX (PDX 17-01) and cultured tumor cells ex vivo in the presence of palbociclib and/or chloroquine. While treatment with either compound alone did not significantly affect tumor cell proliferation, the combined treatment arrested cell growth (Fig. 6, D and E). Hence, analysis of CDK4 Thr172 phosphorylation status in TNBC tumors might allow one to identify tumor cases that would show a therapeutic response to the improved CDK4/6 inhibition.

A lysosomal gene signature in luminal-type, hormone receptor–positive breast cancers predicts resistance to palbociclib treatment

In contrast to TNBC, most of luminal-type, hormone receptor–positive breast cancers are initially sensitive to treatment with CDK4/6 inhibitors. However, a small subset of these tumors is inherently resistant to palbociclib treatment, despite their RB1-positive status. We wondered whether the increased lysosomal mass in these tumors might be responsible for their preexisting resistance to palbociclib. To test this hypothesis, we interrogated transcript microarray data from NeoPalAna clinical trial (26). The trial evaluated the response to palbociclib in combination with an aromatase inhibitor anastrozole as neoadjuvant therapy in patients with clinical stages 2 and 3 hormone receptor–positive breast cancer. Five patients in this trial displayed preexisting resistance to palbociclib treatment, despite the RB1-positive status of their tumors. Comparison of the expression patterns between palbociclib-resistant versus palbociclib-sensitive tumors revealed a significant enrichment of the Kyoto Encyclopedia of Genes and Genomes (KEGG) “lysosome” signature in the resistant tumors (Fig. 6F). These findings indicate that the increased expression of lysosomal genes, likely reflecting an increased lysosomal mass, is also seen in hormone receptor–positive breast cancers that are resistant to treatment with CDK4/6 inhibitors. We propose that the therapeutic strategies to overcome the lysosomal sequestration, presented here for TNBC, might render CDK4/6 inhibitors also efficacious against intrinsically resistant hormone receptor–positive tumors.


CDK4/6 inhibition in conjunction with endocrine therapy represents the standard of care for advanced, hormone receptor–positive breast cancers. In contrast to these tumors, TNBC are intrinsically resistant to CDK4/6 inhibitors (9). This has been attributed to a particularly frequent loss of RB1 in this tumor type (12). Moreover, a recent study revealed that CDK4/6 inhibitor–resistant TNBC cells exit mitosis and enter the G1 phase with relatively high levels of cyclin E1 and high CDK2 activity (27), which might entirely bypass the requirement for CDK4/6 in cell cycle progression. These findings have raised a possibility that the cell cycle machinery may operate in a distinct fashion in TNBC cells, thereby rendering the proliferation of these cells CDK4/6 independent.

In this study, we demonstrate that RB1-positive TNBCs critically require CDK4/6 for proliferation, like hormone receptor–positive breast cancers, and yet, these tumors are largely resistant to treatment with all available CDK4/6 inhibitors. We found that the resistant tumors display markedly elevated levels of nuclear (active) TFEB, the master regulator of lysosomal biogenesis. Consequently, the resistant tumors contain elevated numbers of lysosomes. Importantly, the increased lysosomogenesis is causally linked to the intrinsic resistance of TNBC to CDK4/6 inhibitors, as depletion of TFEB reduced the lysosomal numbers and rendered TNBC sensitive to these compounds. We also showed that the increased lysosomal gene expression signature, likely reflecting increased lysosomal numbers, underlies the intrinsic resistance of TNBC and a subset of hormone receptor–positive tumors to CDK4/6 inhibition.

We propose three therapeutic strategies to overcome the resistance of TNBC to CDK4/6 inhibition: First, coadministration of azithromycin, siramesine, or chloroquine rendered resistant cells fully sensitive to CDK4/6 inhibitor treatment. Azithromycin is used for treatment of various types of infections and is considered to be very safe. Siramesine, an agonist of the σ2 receptor, has been previously used in clinical trials for treatment of depression and anxiety, but it failed to show any efficacy (28, 29). This treatment did not result in any adverse side effects. Chloroquine is widely used for prevention and treatment of malaria. Hence, each of these compounds could be used, in combination with CDK4/6 inhibitors, to treat patients with breast cancer. Our findings are consistent with a recent report that all three CDK4/6 inhibitors (palbociclib, ribociclib, and abemaciclib) accumulate in the lysosomes. This study also reported that palbociclib sensitivity is increased upon lysosomal basification with chloroquine or ammonium chloride (30).

A synergistic combination of palbociclib with chloroquine has been described before, and the growth-suppressive effect has been attributed to inhibition of autophagy (31). However, our results indicate that autophagy inhibition is not the mediator of this effect, since deletion of the master regulator of autophagy Beclin-1 (BECN1) (15) or the mediator of autophagosome formation ATG7 (16) had no influence on the response of cells to palbociclib. Hence, the observed ability of lysosomotropic compounds to sensitize cells to CDK4/6 inhibition is mediated by their action on lysosomes.

The second strategy to overcome the resistance is provided by our observation that coinhibition of CDK4/6 and CDK2 arrested the growth of TNBC tumor cells, while inhibition of CDK2 or CDK4/6 alone had no effect. These findings are in agreement with the observations that CDK2 is dispensable for proliferation of human tumor cells (32) and that CDK4/6 and CDK2 play redundant and overlapping roles in cell proliferation (1).

Last, we developed novel, less basic CDK4/6 inhibitors that were designed to escape the lysosomal sequestration. We verified that these inhibitors indeed do not accumulate in tumor cell lysosomes. Importantly, these compounds show a markedly improved efficacy against TNBC, as compared to all existing FDA-approved compounds. It should be noted that the basic properties of CDK4/6 inhibitors facilitate cell permeability, as basic compounds become less ionized in physiologic pH compared to acidic structures. For this reason, in our analyses, we had to use higher concentrations of the novel compounds, as compared to those of the FDA-approved CDK4/6 inhibitors, to achieve the comparable intracellular concentration of the compounds. Future in vivo therapeutic application of these compounds will require overcoming their limited cell permeability. One possible solution would be to generate prodrugs that retain their basic characteristics, which are then intracellularly cleaved by cellular enzymes to release the mature acidic inhibitor compound (33). While these issues need to be resolved, our study provides a proof-of-concept demonstration of the utility of developing less basic inhibitors and indicates that these compounds would potently inhibit the growth of TNBC that are resistant to all currently used CDK4/6 inhibitors.

Palbociclib and ribociclib, both very specific CDK4/6 inhibitors, have failed to show a significant antitumor activity in advanced breast cancers when used as single agents (4). This is in contrast to in vitro and preclinical analyses, where palbociclib or ribociclib treatment can arrest cell proliferation (9). The increased lysosomal mass in tumors may explain this discrepancy. Lysosomal biogenesis is increased in hypoxic and nutrient-deprived conditions (34), and advanced cancers are characterized by high lysosomal content (35). The well-documented synergistic effect of antiestrogens with CDK4/6 inhibitors might be explained by the fact that the former reduce the levels of cyclin D1, thereby further inhibiting CDK4/6 kinase (36). Since tumors that sequester CDK4/6 inhibitors into the lysosomes display reduced, but not extinguished, CDK4/6 kinase activity, inhibition of the ER-dependent pathways may allow to diminish CDK4/6 activity below the levels needed to sustain cell proliferation.

In summary, we have shown that the increased lysosomal mass is responsible for the intrinsic resistance of TNBC and a subset of hormone receptor–positive tumors to one of the most promising breast cancer therapies in decades. Hence, inhibition of the lysosomal function may allow to substantially extend the range of breast tumors that respond to anti-CDK4/6 therapy and may block tumor cell proliferation and possibly trigger senescence in the resistant TNBC tumors. Since high lysosomal content is generally seen in aggressive, advanced cancer cases, this paradigm may apply to a wide variety of tumors that present clinical resistance to therapy with CDK4/6 inhibitors.


Experimental design

The overall objective of this study was to determine whether RB1-proficient TNBCs are a target for CDK4/6 inhibition. To this end, the first objective was to identify the mechanism of resistance of RB1-proficient TNBC cells to CDK4/6 inhibitors. The second objective was to identify therapeutic strategies to overcome the resistance. Both objectives were met by identifying the lysosomal biomass as mediator of resistance and determining combinatorial treatment options with lysosomal inhibitors, as well as generating structurally modified CDK4/6 inhibitors.

Cell lines

HCC1806, HCC38, HCC1143, HCC70, HCC1187, and HCC1937 cells were obtained from American Type Culture Collection (ATCC) and cultured in RPMI media supplemented with 10% inactivated fetal calf serum. CAL120 and CAL51 cells, obtained from DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen - German Collection of Microorganisms and Cell Cultures), and human embryonic kidney (HEK) 293T cells, obtained from ATCC, were cultured in Dulbecco’s Modified Eagle Medium (DMEM) media with 10% inactivated fetal calf serum. MDA-MB-157 and MDA-MB-468 cells were obtained from ATCC and cultured in McCoy’s 5A media supplemented with 10% inactivated fetal calf serum. SUM149 and SUM159 cells were obtained from S. Ethier (University of Michigan) and were cultured in DMEM/F12 media supplemented with 10% inactivated fetal calf serum, hydrocortisone (1 μg/ml; STEMCELL Technologies), and insulin (5 μg/ml; Thermo Fisher Scientific). PMC42 cells were obtained from R. Whitehead (Ludwig Institute for Cancer Research, Melbourne, Australia) and were cultured in RPMI media with 10% inactivated fetal calf serum, hydrocortisone (5 ng/ml), and insulin (10 μg/ml). All cells were maintained at 37°C and 5% CO2. All media were obtained from Thermo Fisher Scientific. Fetal calf serum was purchased from Sigma-Aldrich.

BrdU-incorporation assay

Cells were pulsed with bromodeoxyuridine (BrdU) (10 μM; BD Biosciences) for 1 hour (HCC1806, SUM159, SUM149, CAL51, and MDA-MB-468) or 2 hours (CAL120, HCC38, HCC1143, HCC70, HCC1187, HCC1937, and MDA-MB-157) and subsequently trypsinized, washed with phosphate-buffered saline (PBS), and fixed with 80% ethanol (EtOH) overnight at 4°C. Next, DNA was hydrolyzed by incubation with 2 N HCl/0.5% Triton X-100 for 30 min at room temperature (RT) and neutralized by adding 0.1 M Na2B4O7 (pH 8.5). Subsequently, cells were incubated in 100 μl of PBS/1% bovine serum albumin (BSA) containing fluorescein isothiocyanate–conjugated anti-BrdU antibody (556028, BD Biosciences) 1:5 for 30 min at RT. Next, cells were washed twice with PBS, stained with 20 μl of 7-AAD (7-Amino-Actinomycin D) (BD-Biosciences), resuspended in PBS, and analyzed by flow cytometry. Data acquisition was performed on a BD LSRFortessa Flow Cytometry Analyzer, and data analysis was done with Cytobank.

Growth analysis

For long-term growth analysis, 4 × 104 SUM159, SUM159 non-targeting (CRISPR sg) (nt), and SUM159 ΔRB1 cells or 6 × 104 HCC1806, SUM149, CAL120, HCC70, and HCC1937 cells were seeded in 6-cm cell culture dishes. Next day, cell number was determined on one plate, and the remaining plates were treated. For chloroquine, azithromycin, and siramesine treatment, medium was changed every 2 days. On day 6, the total cell number on each plate was determined using a Beckman Coulter Z series dual threshold analyzer.

siRNA transfection

siRNAs were transfected using Lipofectamine RNAiMAX (Life Technologies; standard forward transfection) according to the manufacturer’s instructions with a final siRNA concentration of 10 nM. The following siRNAs were used (Thermo Fisher Scientific): ctrl siRNA, Silencer Select, Negative Control #1 siRNA; ATP6AP1, assay ID s1812; TFEB, assay ID s224819; CDK4si1, assay ID s2822; CDK4si2, assay ID 44756; CDK6si1, assay ID s51; CDK6si2, assay ID 99.

CRISPR-Cas9–mediated knockout

CRISPR guides targeting RB1, ATG7, and BECN1 were designed using CRISPOR (37): RB1-gRNA1: TCCTGAGGAGGACCCAGAGC; RB1-gRNA2: CGGTGGCGGCCGTTTTTCGG; RB1-gRNA3: GGACAGGGTTGTGTCGAAAT; ATG7-gRNA1: GAAGCTGAACGAGTATCGGC; ATG7-gRNA2: CTTGAAAGACTCGAGTGTGT; BECN1-gRNA1: CCTGGATGGTGACACGGTCC; and BECN1-gRNA2: ATTTATTGAAACTCCTCGCC. Guides were cloned into lentiCRISPRv2 as previously described (38). LentiCRISPRv2 was a gift from F. Zhang (Addgene plasmid no. 52961). The CDK2-specific CRISPR guide (CAGAAACAAGTTGACGGGAG) was described before (23). To achieve efficient RB1 knockout, all three guides were used on the same cells.

Generation of CDK2AS lentiviral expression vector

Mouse analog-sensitive CDK2 (CDK2AS-F80G) was subcloned from p3xFlag-CMV-10 (23) into the lentiviral expression vector pLenti-FUGW-3xFLAG-ires-Neo (modified from Addgene plasmid no. 14883; a gift from D. Baltimore).

Lentiviral transduction

Lentiviruses were produced by transfecting HEK293T cells with the respective lentiviral plasmid (plentiCRISPRv2-RB1sg/ATG7sg/BECN1sg/CDK2sg-puro, pLJC5-Tmem192-3xHA-puro, pLJC5-Tmem192-3xFlag-puro, and pLenti-FUGW-3xFLag-CDK2AS-ires-Neo) and lentiviral envelope (vesicular stomatitis virus glycoprotein) and packaging (Δ8.9) plasmids using PolyFect (Qiagen). Transfection medium was changed the next day and replaced by DMEM containing 10% inactivated fetal calf serum and 1% BSA. Medium with virus particles was collected 48 hours later and passed through a 0.45-μm syringe filter followed by virus concentration using Amicon Ultra-15 100 kDa centrifugal columns (Millipore). Concentrated virus was added to 1 × 105 cells together with polybrene (8 μg/ml; AmericanBio). Forty-eight hours after infection, medium was changed, and cells were placed under selection with puromycin [SUM159 and CAL120 (3 μg/ml); HCC1806, SUM149, and HCC1143 (2 μg/ml)] or neomycin [HCC1806 (400 μg/ml) and SUM159 (1.2 mg/ml)].

Immunoblot analysis and antibodies

Cell lysates were prepared in ice-cold radioimmunoprecipitation assay (RIPA) lysis buffer [50 mM tris (pH 8), 150 mM NaCl, 1% NP40, 0.5% Na-deoxycholate, and 0.1% SDS] supplemented with HALT Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific). Protein content was quantified using the BCA (bicinchoninic acid) Protein Assay Kit (Pierce). Protein (20 μg) were resolved by SDS–polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes (Thermo Fisher Scientific). The following primary antibodies were used: anti-CDK4 (1:1000; MS616P1, Thermo Fisher Scientific or 1:100; sc-23896, Santa Cruz Biotechnology), anti-CDK6 (1:1000; ab54576, Abcam), anti-ATP6AP1 (1:1000; sc-81886, Santa Cruz Biotechnology), anti-phospho RB1-S807/811 (1:1000; 8516, Cell Signaling Technology), anti-RB1 (1:500; 9313, Cell Signaling Technology), anti-TFEB (1:1000; 4240, Cell Signaling Technology), anti–proliferating cell nuclear antigen (1:250; sc-9857, Santa Cruz Biotechnology), anti–pCDK4-Thr172 (1:1000; STJ29359, St John’s Laboratory), anti-ATG7 (1:1000; 2631, Cell Signaling Technology), anti-BECN1 (1:1000; 4122, Cell Signaling Technology), anti-LAMP1 (lysosomal associated membrane protein 1) (1:1000; 9091, Cell Signaling Technology), anti-LAMP2 (1:1000; sc-18822, Santa Cruz Biotechnology), anti–cathepsin C (1:250; sc-74590, Santa Cruz Biotechnology), anti–cathepsin D (1:1000; 2284, Cell Signaling Technology), anti-LC3 (1:1000; 2775, Cell Signaling Technology), anti–voltage-dependent anion channel (1:1000; 4661S, Cell Signaling Technology), anti-calreticulin (1:1000; 12238, Cell Signaling Technology), anti–golgin-97 (1:1000; 97537, Cell Signaling Technology), anti-p70 S6 kinase (1:1000; 2708, Cell Signaling Technology), anti- PEX19 (1:500; ab137072, Abcam), anti-CDK2 (1:1000; sc-163, Santa Cruz Biotechnology), anti–glyceraldehyde phosphate dehydrogenase (1:5000; 5174, Cell Signaling Technology), and anti-tubulin (1:5000; T5168, Sigma-Aldrich).

Nuclear fractionation

Extraction of the nuclear cell fraction was done using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific) according to the manufacturer’s instructions.

Microscopic imaging of palbociclib autofluorescence

To visualize palbociclib accumulation in lysosomes, 6 × 104 cells were cultured on coverslips in 12-well cell culture plates. Next day, cells were treated with dimethyl sulfoxide (DMSO), palbociclib (1 μM), or palbociclib and bafilomycin A1 (100 nM) for 24 hours. Subsequently, cells were incubated with LysoTracker Green DND-26 (Cell Signaling Technology) 1:10,000 for 5 min at 37°C when indicated. Next, the coverslips were removed from the plates and placed on a microscope slide with 15 μl of PBS (cells facing downward) and analyzed using a Nikon Eclipse E600 microscope immediately. Autofluorescent palbociclib is visible in the blue channel. LysoTracker Green DND-26 staining appeared weak. Brightness was therefore adjusted in images showing LysoTracker Green staining using ImageJ. Adjustment was similar in all LysoTracker Green images.

Lysosome immunoprecipitation

Lysosome immunoprecipitation (IP) was performed as described previously (17). Each IP was processed separately. For each IP, HCC1806, SUM149, CAL120, SUM159, and HCC1143 cells expressing either Tmem192-3xHA or Tmem192-3xFlag were grown to 80% confluency in 15-cm cell culture plates. To determine drug amount in lysosomes, SUM159 and HCC1143 cells were treated with palbociclib (1 μM) for 20 hours and HCC1806, SUM149, and CAL120 cells with ribociclib (1 μM), Cpd-1 (50 μM), and Cpd-2 (25 μM) for 8 hours. LysoTracker DND-99 (Thermo Fisher Scientific) (4 nM) was added for the last hour of drug treatment. For each condition, lysosome IP was performed in duplicate/triplicate for TMEM192-3xHA cells and on one plate of TMEM192-3xFlag cells; TMEM192-3xFlag cells were used as a control to determine the amount of nonspecific drug and LysoTracker binding. Pulldown of hemagglutinin (HA)–tagged lysosomes was performed using anti-HA magnetic beads (Thermo Fisher Scientific). To extract the drug from immunoprecipitated lysosomes, beads with bound lysosomes were resuspended in extraction buffer (80% methanol and 20% water containing internal standards), and the extract was cleared by centrifuging at 16,100g for 10 min. Extracts and corresponding whole-cell lysates extracted with the same extraction buffer were submitted to the Metabolite Profiling Core Facility at the Whitehead Institute for Biomedical Research for measurement of the corresponding drug (palbociclib, ribociclib, Cpd-1, and Cpd-2) and LysoTracker content. Metabolite profiling was conducted on a Q Exactive bench top Orbitrap mass spectrometer equipped with an Ion Max source and a HESI II probe, which was coupled to a Dionex Ultimate 3000 high-performance liquid chromatography system (both from Thermo Fisher Scientific). Five microliters of lysosome IP was injected onto a Kinetex C18 50-mm by 2.1-mm analytical column (2.6-μm particle size, Phenomenex). The column oven and autosampler tray were held at 30° and 4°C, respectively. The following conditions were used to achieve chromatographic separation: Buffer A was 0.1% formic acid, and buffer B was 0.1% formic acid in acetonitrile. The flow rate was 0.4 ml/min. The chromatographic gradient was run as follows: 0 to 3.5 min, linear gradient of 5 to 80% B; 3.6 to 4.5 min, the gradient was held at 98% B; 4.6 to 6 min, the gradient was held at 5% B. The mass spectrometer was operated in a positive mode, and data acquisition was performed using two narrow-range full scans from and 70 to 250 and 390 to 500 mass/charge ratio (m/z) to include both palbociclib and the internal standard phenylalanine-13C9-15N. An additional targeted selected ion monitoring scan was centered on 400.2115 m/z to improve detection of LysoTracker Red-DND-99. The resolution was set at 70,000, the AGC (automatic gain control) target was 1 × 106, and the maximum injection time was 50 ms. Absolute quantification of palbociclib, ribociclib, Cpd-1, and Cpd-2 was performed using XCalibur QuanBrowser 2.2 (Thermo Fisher Scientific) using a 5–parts per million mass tolerance and referencing a calibration curve made in 80% methanol containing concentrations of palbociclib, ribociclib, Cpd-1, and Cpd-2 ranging from 1 to 100 μM and the internal standard phenylalanine-13C9-15N (500 nM).

To determine the relative amount of palbociclib per lysosome, first, nonspecific palbociclib and LysoTracker signal determined in the IP from Tmem192-3xFlag cells was subtracted from each Tmem192-3xHA signal. Next, LysoTracker signal was normalized against palbociclib signal for each IP. To determine the total amount of ribociclib, Cpd-1, and Cpd-2 per lysosome, the amount of drug determined in the respective IP from Tmem192-3xFlag cells was subtracted from each corresponding Tmem192-3xHA IP.

To ensure the quality of lysosomal IP (HA-IP), immunoblot analysis for lysosomal, as well as mitochondrial, endoplasmic reticulum, peroxisome, cytoplasmic, and Golgi proteins, was performed for the Tmem192-HA and Tmem192-Flag HA-IP. To obtain lysosomal lysates, beads with bound lysosomes were resuspended in RIPA lysis buffer supplemented with HALT Protease and Phosphatase Inhibitor Cocktail. Lysates were cleared by centrifugation at 16,100g for 10 min and subjected to immunoblot analysis.

LysoTracker Green DND-26 and LysoSensor Green DND-189 flow cytometry

To determine the lysosomal mass of TNBC cells, semiconfluent 6-cm dishes of naïve cells or cells that were transfected with anti-TFEB or control siRNA were trypsinized, and equal cell numbers were stained with 50 nM LysoTracker Green DND-26 (Cell Signaling Technology) for 15 min at 37°C. Subsequently, cells were washed and resuspended in PBS supplemented with 2% fetal calf serum and directly analyzed by flow cytometry. An unstained sample of each cell line or condition was used to adjust autofluorescence to similar intensity for all cell lines/conditions measured. To assess lysosomal acidification, cells were trypsinized, and equal cell numbers were stained with 1 μM LysoSensor Green DND-189 (Thermo Fisher Scientific) for 30 min at 37°C. An unstained sample of each condition was used to adjust autofluorescence to similar intensity for all conditions measured. Data acquisition was performed on a BD LSRFortessa Flow Cytometry Analyzer, and data analysis was done with Cytobank.

Electron microscopy

Cells (2.5 × 105) were seeded in six-well culture plates. Next day, cells were fixed by adding fixative (2.5% paraformaldehyde, 5% glutaraldehyde, and 0.06% picric acid in 0.2 M cacodylate buffer) 1:1 to growth media. Cells were fixed for 2 hours at RT, washed in 0.1 M cadodylate buffer, and postfixed with 1% OsO4/1.5% KFeCN6 for 1 hour at RT. Subsequently, cells were washed twice in water and once in maleate buffer and incubated in 1% uranyl acetate in maleate buffer for 1 hour at RT followed by one wash in maleate buffer and two washes in water. Next, cells were dehydrated in grades of alcohol (10 min each; 50, 70, and 90%; 2×, 10 min, 100%), and cells were removed from the plates in propyleneoxide, pelleted, and infiltrated in a 1:1 mixture of propyleneoxide and TAAB Epon (TAAB Laboratories Equipment) overnight. The following day, the cell pellets were embedded in TAAB Epon and polymerized at 60°C for 48 hours. Ultrathin sections (~60 nm) were cut on a Reichert Ultracut S microtome, collected onto copper grids, stained with lead citrate, and examined in a JEOL 1200EX transmission electron microscope. Images were taken using an AMT 2k charge-coupled device camera. Sample preparation and image collection were done by the Harvard Medical School Electron Microscopy Core Facility without knowledge of resistant and sensitive status of the respective cell line. Lysosome counting on images was also done by the Core Facility without knowledge whether a cell line is resistant or sensitive. For every cell line, between 16 and 21 pictures were taken and counted.

Determination of cytoplasmic area from electron microscopy pictures

To normalize the number of counted lysosomes from the electron microscopy pictures, the total cytoplasmic area for each electron microscopic image was determined. The cytoplasmic area was defined as the total cellular area minus the area of the nucleus. The area was measured in square pixels using ImageJ and converted into square micrometer using the scale on the image.

Analysis of NeoPalAna trial and TNBC PDX

Microarray gene expression data (GSE93204) were obtained from the tumor biopsies from the NeoPalAna trial (26). In the NeoPaLAna study, resistance to palbociclib was defined as Ki67 of above 2.7% after 15 days of treatment with palbociclib and anastrozole. Log2-normalized expression data from baseline samples were used for the analysis. The gene set enrichment analysis (GSEA) software was used for calculating enrichment scores.

For analyses of the response of TNBC PDX to ribociclib, classification of responses (stable disease and progressive disease) and RNA sequencing data were from (19). Classification of tumors into TNBC category was from PRoXe ( Expression data from baseline PDXs were used for the analysis. Differential expression analysis was done using DESeq2 v1.24 (39). GSEA analysis was performed using the Broad GSEA application.

Synthesis and characterization of Cpd-1 and Cpd-2

Cpd-1 and Cpd-2 were synthesized by the Global Discovery Chemistry group of Novartis Institutes for Biomedical Research and were characterized by standard analytical methods. Cpd-1, 7-cyclopentyl-2-((5-(4-hydroxypiperidine-1-carbonyl)pyridin-2-yl)amino)-N,N-dimethyl-7H-pyrrolo[2,3-d]pyrimidine-6-carboxamide. 1H NMR (nuclear magnetic resonance) (400 MHz, DMSO-d6): δ = 9.96 (s, 1H), 8.85 (s, 1H), 8.40 to 8.30 (m, 2H), 7.80 (dd, J = 8.6 and 2.4 Hz, 1H), 6.65 (s, 1H), 4.82 to 4.69 (m, 2H), 4.08 to 3.56 (m, 3H), 3.23 (ddd, J = 13.0, 9.3, and 3.3 Hz, 2H), 3.06 (s, 6H), 2.49 to 2.38 (m, 2H), 2.06 to 1.93 (m, 4H), 1.77 (s, 2H), 1.72 to 1.59 (m, 2H), and 1.45 to 1.33 (m, 2H); liquid chromatography–mass spectrometry (LCMS) >95%; high-resolution mass spectrometry (HRMS) (m/z): [M + H]+ calcd for C25H32N7O3, 478.2576; found, 478.2552.

Cpd-2, 7-cyclopentyl-2-((5-(4,4-difluoropiperidine-1-carbonyl)pyridin-2-yl)amino)-N,N-dimethyl-7H-pyrrolo[2,3-d]pyrimidine-6-carboxamide. 1H NMR (400 MHz, DMSO-d6): δ = 10.00 (s, 1H), 8.85 (s, 1H), 8.45 to 8.35 (m, 2H), 7.87 (dd, J = 8.7, 2.4 Hz, 1H), 6.65 (s, 1H), 4.76 (p, J = 8.8 Hz, 1H), 3.63 (s, 4H), 3.06 (s, 6H), 2.48 to 2.37 (m, 2H), 2.12 to 1.95 (m, 8H), and 1.72 to 1.61 (m, 2H); LCMS >95%; HRMS (m/z): [M + H]+ calcd for C25H30F2N7O2, 498.2429; found, 498.2383.

The median inhibitory concentration (IC50) against different cyclin-CDKs were determined as follows. Human CDK4/cyclin D1 was expressed in Sf21 cells via baculovirus infection. An assay for monitoring CDK4/cyclin D1–catalyzed phosphorylation of RB1 at the Ser780 site was performed using time-resolved fluorescence resonance energy transfer in a 384-well format and was used for IC50 determination and kinetic analysis. The reaction was carried out in a 30-μl volume containing 0.25 nM CDK4/cyclin D1, 150 nM biotin-RB1 (773-924), 3 μM ATP, and 1.3% DMSO (or compound in DMSO) in the assay buffer [50 mM Hepes-Na (pH 7.5), 5 mM MgCl2, 1 mM dithiothreitol, 0.02% Tween 20, and 0.05% BSA]. ATP (3 μM) was added last to initiate the reaction. The reaction was quenched with 10 μl of 240 mM EDTA-Na (pH 8.0) after 60 min of incubation at 22°C. The signal was developed by the addition of 40 μl of detection solution containing 40 nM streptavidin-allophycocyanin, anti–phospho-RB1 (S780) antibody (143 ng/ml), and 2 nM Eu-W1024 anti-rabbit immunoglobulin G antibody in the detection buffer [50 mM Hepes-Na (pH 7.5), 60 mM EDTA-Na (pH 8.0), 0.05% BSA, and 0.1% Triton X-100]. After the 60-min incubation in the dark, the plate was read on EnVision (2102-0010, PerkinElmer). The inhibition of human CDK1/cyclin B and human CDK2/cyclin A (catalog nos. 14-450 and 14-448, respectively, Millipore) was monitored using IMAP-FP assays (Molecular Devices, CA) by following the phosphorylation of Tamra-Histone H1-derived peptide (catalog no. R7385, Molecular Devices). The final reaction volume was 20 μl and contained 0.25 nM CDK1/cyclin B or 0.3 nM CDK2/cyclin A, 100 nM Tamra-H1, and 20 μM ATP in 1X reaction buffer (R8139, Molecular Devices). The reactions were run for 2 hours at 22°C. Quenching and detection were carried out following the protocols for the peptide substrate provided by the vendor.

Ex vivo culture of patient-derived TNBC tumors

Fresh tumor tissue was received from Brigham and Women’s Hospital after signed consent was obtained from patients with TNBC undergoing surgery, according to a research protocol approved by Dana-Farber/Harvard Cancer Center Institutional Review Board. The study is compliant with all relevant ethical regulations regarding research involving human participants. Samples were deidentified before transport to the laboratory. To avoid contamination of subsequent ex vivo culture, tumor tissue was incubated 2 × 10 min in PBS containing penicillin/streptomycin and amphotericin B (both from Thermo Fisher Scientific) upon receipt. Subsequently, any apparent fat tissue was removed from the specimen, and tumor was cut in small, very thin sections (approximately 1 mm3) using sterile scalpel and microdissection scissors. Following, tumor pieces were incubated in 1.5 ml of medium [DMEM/F12 containing 10% inactivated fetal calf serum, 2% penicillin/streptomycin, insulin (10 μg/ml), and hydrocortisone (10 μg/ml)] and the indicated compounds for 48 hours in a 12-well cell culture plate at 37°C and 5% CO2. For each condition, multiple tumor pieces were used. BrdU (20 μM) was added for the last 4 hours of drug incubation. Subsequently, tumor pieces were fixed in 10% neutral-buffered formalin (Thermo Fisher Scientific) for 24 hours at RT, followed by transfer to 70% ethanol. Paraffin-embedding, sectioning, and staining with hematoxylin and eosin were performed by the Brigham and Women’s Hospital Specialized Histopathology Core. BrdU immunohistochemistry staining was carried out by Servicebio ( Counting of BrdU-positive cells on tissue sections was performed in a blinded fashion.

Ex vivo culture of PDX

PDXs were derived from fresh human cancer specimens that were received from Brigham and Women’s Hospital after signed consent was obtained from patients undergoing surgery, according to a research protocol approved by Dana-Farber/Harvard Cancer Center Institutional Review Board. Fragments of the cancer tissue specimen were implanted orthotopically into mammary fat pads of female NOG mice (Taconic) and serially passaged as subcutaneous implants of tumor fragments. For ex vivo culture of PDX 17-01, a tumor fragment was cut into small pieces and sliced thinly. Slices were subsequently incubated in dissociation solution [DMEM with collagenase type IV (16,000 U/ml, 1:50; Thermo Fisher Scientific), hyaluronidase (10,000 U/ml, 1:100; Sigma-Aldrich), and deoxyribonuclease I (10,000 U/ml, 1:77; Sigma-Aldrich)] at 37°C until sufficient dissociation was detected under the microscope. Next, cells were passaged through a 100-μm filter and centrifuged at 600g for 5 min. Cell pellet was resuspended in DMEM/F12 medium supplemented with 5% inactivated fetal calf serum, and cells were plated on Matrigel-coated eight-well chamber plate (Corning Matrigel hESC-qualified matrix according to the manufacturer’s instructions). Cells were incubated in a cell culture incubator (37°C, 5% CO2) until cells reached 80% confluency. Cells were treated with DMSO, palbociclib (1 μM), chloroquine (30 μM), or palbociclib and chloroquine for 24 hours. Subsequently, cells were cultured with 10 μM 5-ethynyl-2′-deoxyuridine (EdU) for 4 hours, stained with the Click-iT EdU Alexa Fluor 488 Imaging Kit according to the manufacturer’s instructions to visualize EdU incorporation, and analyzed using a Nikon Eclipse E600 microscope.

Immunohistochemistry staining

Formalin-fixed, paraffin-embedded tissue sections of HCC1954 and MDA-MB-468 xenografts were deparaffinized, boiled in 10 mM sodium citrate (pH 6.0) for 10 min, and treated with 3% H2O2 for 10 min. Following, slides were blocked with 10% goat serum in tris-buffered saline with 0.1% Tween 20 and 1% BSA and incubated with pCDK4 Thr172 antibody overnight at 4°C in tris-buffered saline with 0.1% Tween 20, 5% goat serum, and 1% BSA. pCDK4 Thr172 antibody (STJ29359, St John’s Laboratory) was used at a dilution of 1:1000. Subsequently, slides were incubated with a biotinylated secondary antibody diluted in tris-buffered saline with 0.1% Tween 20 and 1% BSA for 30 min at RT. Antibody was visualized using the VECTASTAIN Avidin-Biotin Complex Kit (PK-4000, Vector Laboratories) and ImmPACT diaminobenzidine substrate (SK-4105, Vector Laboratories) and counterstained with Vector Hematoxylin QS (H-3404, Vector Laboratories). All reagents were used according to the manufacturer’s instructions. Subsequently, sections were dehydrated and mounted with Permount.

Xenograft studies

In vivo studies of IDC50X (generation of the model was described by Shu et al. (40)) were conducted using 12-weeks-old female immunodeficient NOD.Cg-PrkdcscidIl2rgtm1Sug/JicTac (NOG) mice (Taconic). Cells (2 × 105) in 50% Matrigel (Thermo Fisher Scientific) in DMEM/F12 were injected bilaterally into the inguinal mammary fat pads. When tumors became palpable, mice were randomized to treatment groups with five animals in control or hydroxychloroquine group and eight animals in palbociclib or palbociclib and hydroxychloroquine group. The median tumor volume in each group at the start of treatment was 175 mm3. Tumor volumes were calculated using the formula V = (L × W × W)/2, where L is tumor length and W is tumor width. Palbociclib was administered at 50 mg/kg by daily oral gavage and hydroxychloroquine at 60 mg/kg by daily intraperitoneal injection (in PBS). Control animals received only vehicle [10% 0.1 N HCl, 10% Cremophor EL, 20% PEG300 (polyethylene glycol, molecular weight 300), and 60% 50 mM citrate buffer (pH 4.5)]. Animal studies were performed according to a protocol approved by the Dana-Farber Cancer Institute Animal Care and Use Committee.

Statistical analysis

Statistical significance was defined as a P value of less than 0.05 using the appropriate statistical test method. The statistical test used is indicated in the figure legends. Error bars are presented as means ± SD.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank members of the Sicinski laboratory; S. Goel, S. Grinstein, and P. Ostrowski for helpful discussions and suggestions; and the Whitehead Institute Metabolite Profiling Core Facility for help in acquiring and analyzing the data presented. We also thank J. Köhler for help with nuclear fractionation and S. Knobloch-Bollmann for generation of the schematic representation of possible lysosome-mediated resistance mechanisms. Funding: This work was supported by R01 CA202634, P50 CA168504, and the Dana-Farber/Novartis Drug Discovery Program (to P.Si.); R01 CA103866, R01 CA129105, R37 AI47389, and the Lustgarten Foundation (to D.M.S.); R35 CA210057 and Breast Cancer Research Foundation (to J.J.Z.); and a Charles King Fellowship (to M.A.-R.). Author contributions: A.F. and P.Si. designed the project. A.F. performed the experiments with the help from collaborators. I.S., D.B., W.Micho., T.O., P.St., K.N., B.Ji., and M.C. helped with the experiments. C.B., W.Micha., Q.S., and A.L. contributed Cpd-1 and Cpd-2. M.A.-R. and D.M.S. helped with immunoprecipitation of lysosomes. J.B. and J.J.Z. contributed TNBC PDX. R.T. helped with analyses of PDX. B.Jo. and K.P. contributed IDC50 PDX. M.E. helped with electron microscopy and analyses of electron micrographs. N.G. helped supervise parts of the project. C.D., R.S., R.J., and M.B. contributed analyses of NeoPalAna study. A.S.F. and R.J. contributed GSEA analysis of TNBC PDX. D.D. contributed patient-derived TNBC tumors and histopathological analysis. K.P. provided scientific input and guidance throughout the project. A.F. and P.Si. wrote the manuscript with the help of K.P. Competing interests: C.B. has stock and options in Novartis and is an inventor on the patent US 8,962,630 B2, which is related to compounds 1 and 2. R.S. received research grants from AstraZeneca, GlaxoSmithKline, Gilead Sciences, and PUMA Biotechnology; has been a consultant to Eli Lilly; and serves on a scientific advisory committee of MacroGenics. N.G. is a founder, science advisory board (SAB) member, and equity holder in Gatekeeper, Syros, Petra, C4, B2S, Aduro, and Soltego (board member). The Gray laboratory receives or has received research funding from Novartis, Takeda, Astellas, Taiho, Janssen, Kinogen, Voronoi, Her2llc, Deerfield, and Sanofi. D.D. is on the advisory board for Oncology Analytics Inc. and consults for Novartis. R.J. has received research funding from Pfizer. M.B. receives sponsored research support from Novartis and serves as consultant to H3 Biomedicine and serves on the scientific advisory boards of Kronos Bio and GV20 Therapeutics. K.P. serves on the scientific advisory boards of Mitra Biotech and Acrivon Therapeutics. P.Si. has been a consultant at Novartis, Genovis, Guidepoint, The Planning Shop, ORIC Pharmaceuticals, Syros, and Exo Therapeutics; his laboratory receives research funding from Novartis. R.T. is a shareholder of Novartis. All authors declare that they have no other competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. All cell lines can be obtained from ATCC, DSMZ, or P. Sicinski, except for SUM149 and SUM159 (from S. Ethier, University of Michigan) and PMC42 (from R. H. Whitehead, Melbourne, Australia).

Stay Connected to Science Advances

Navigate This Article