A Synthetic Cell-Penetrating Dominant-Negative ATF5 Peptide Exerts Anticancer Activity against a Broad Spectrum of Treatment-Resistant Cancers

Although recent significant responses for cancer therapy have been achieved in some malignancies, these are often initially impressive, but unfortunately not durable (1, 2). For other malignancies, such as high-grade primary brain cancers, the prognosis is unfavorable even with treatment (3). Therefore, efforts continue to identify new potential targets for tumor-specific treatments as well as novel therapeutic strategies to exploit these targets.

ATF5 is an example that has been identified as a potential target for cancer treatment but for which no specific therapy has been developed (4, 5). Activating transcription factor 5 (ATF5; also termed ATFx) is a member of the activating transcription factor/cyclic AMP–responsive element-binding (ATF/CREB) family. A common feature of this family is the presence of a basic leucine zipper (bZIP) domain that promotes DNA binding via the basic region and interactions with other proteins through the leucine zipper (5–7). ATF5 binds to several different promoter elements including the nutrient-sensing report element (NRSE) and a novel motif to regulate gene transcription (6, 8). Full-length ATF5 appears to be rapidly degraded via the proteasome and it has the unusual property that it is among a small group of proteins that are selectively translated when eiF2α is phosphorylated (9, 10).

ATF5 protein levels are increased in a variety of human malignancies, including glioblastoma, breast, pancreatic, lung, and colon cancers (11). In contrast, with few exceptions (liver, prostate, and testis), ATF5 expression is low in normal tissue of the respective organs. In several tumor types, including glioblastoma and non–small cell lung cancer, ATF5 expression negatively correlates with survival (12, 13). In cell culture studies, ATF5 promotes survival by counteracting apoptosis in pro-B lymphocytes deprived of IL3 or in HeLa cells after growth factor withdrawal (14). In addition, ATF5 regulates transcription of antiapoptotic B-cell leukemia 2 (Bcl-2) and of Bcl-2 family member, myeloid cell leukemia-1 (Mcl-1), presumably thereby promoting tumor cell survival (13, 15). Conversely, interference with ATF5 expression or activity yields a marked induction of apoptosis in glioblastoma cells in vitro and in vivo without affecting astrocytes (16, 17). Moreover, in a transgenic murine model in which endogenous glioblastomas were induced by a PDGF/sh-p53–expressing virus, activation of a dominant/negative (d/n)-ATF5 blocked tumor formation and resulted in regression of formed tumors (17). Antineoplastic activity of d/n-ATF5 was also reported for breast cancer cells and pancreatic cancer cells in vitro (11, 15, 18). These findings thus suggest ATF5 as a promising target for a tailored anticancer therapy.

To provide a potential means to target ATF5 in vivo, we designed a d/n-ATF5 linked to a cell-penetrating domain (Penetratin; ref. 19). This recombinant peptide passes the blood–brain barrier, enters tumor cells, and exerts antineoplastic activity in a rodent transgenic glioma model. In this study, we assessed the activity and mechanism of action of a similar peptide (CP-d/n-ATF5-S1) that was further modified to reduce its size and that was generated synthetically. In in vitro studies and in in vivo murine xenograft models, CP-d/n-ATF5-S1 shows apoptosis induction over a broad range of recalcitrant human malignancies without apparent effects on nontransformed cells. A novel mechanism of action was found in which the peptide reduces expression of the deubiquitinating enzyme Usp9X, which in turn leads to depletion of Mcl-1 and Bcl-2 and to consequent apoptotic death. The latter findings led us to rationally design and carry out in vitro and in vivo tests of several potential combination therapies with CP-d/n-ATF5-S1 that had enhanced efficacy compared with either agent alone.

All procedures were in accordance with Animal Welfare Regulations and approved by the Institutional Animal Care and Use Committee at Columbia University Medical Center (New York, NY).

CP-d/n-ATF5-S1, mutated CP-d/n-ATF5-S1, and Penetratin were purchased from CS Bio. Recombinant TRAIL was from Peprotech. ABT263 was from Selleckchem.

Cells were grown as described previously (20, 21). Cells were obtained from the ATCC or Cell Line Services and authenticated by the manufacturer. No cell line authentication was performed by the authors and details are found in the Supplementary Section.

To examine cellular proliferation, MTT assays were performed as described previously (21).

Annexin V/propidium iodide (PI), PI, and JC-1 stainings were performed as described previously (20, 22).

Protein expression was determined by Western blot analysis as described previously (23).

siRNAs were transfected as described previously (22, 24).

cDNA synthesis and RT-PCR were performed as described previously (23).

Subcutaneous xenografts were implanted as described previously (20).

Statistical significance was assessed by Student t test using Prism version 5.04 (GraphPad). P ≤ 0.05 was considered statistically significant.

CP-d/n-ATF5-S1 is a synthetic 67-amino acid peptide that was engineered to cross cellular membranes and to specifically interfere with the survival-promoting actions of ATF5 (Fig. 1A). The N-terminal end has a 16-amino acid Penetratin domain that facilitates cellular penetration (25, 26). A dominant/negative sequence follows in which the DNA-binding domain of ATF5 is substituted by an amphipathic sequence with a leucine repeat at every seventh residue and then by the human ATF5 bZIP domain truncated after the first valine (26–29). Parallel work has demonstrated that a similar recombinant tagged peptide passes the blood–brain barrier, enters intact cells both in vivo and in vitro, and promotes selective death of glioma cells (19). Four independent batches of the peptide (including one under GMP conditions) have had comparable activity. For control purposes, peptides were also synthesized with a penetratin domain alone and in which key leucine residues were mutated to glycine in the d/n portion to reduce binding to potential partners (Fig. 1A).

Western blot analyses revealed that treatment of cultured tumor cell lines (T98G, MDA-MB-436, and GBM12) with CP-d/n-ATF5-S1 leads to a dose-dependent reduction of endogenous ATF5 protein levels by 3 days (Fig. 1B and Supplementary Fig. S1B). In T98G cells, this effect was present after 48 hours, but not 24 hours (Supplementary Fig. S1A). RNAseq analysis of T98G glioblastoma cells treated with CP-d/n-ATF5-S1 for 1 to 3 days showed no significant alteration of ATF5 transcript levels (data not shown). Comparison of endogenous ATF5 levels in T98G cells treated with or without CP-d/n-ATF5-S1 in presence of cycloheximide indicates that the peptide significantly decreases ATF5 protein stability (Fig. 1C and D and Supplementary Fig. S1C). In contrast, treatment with Penetratin peptide did not affect ATF5 stability (Supplementary Fig. S1D and S1E). Thus, one action of CP-d/n-ATF5-S1 is loss of endogenous ATF5 caused at least in part by enhanced turnover.

CP-d/n-ATF5-S1 treatment for 24 hours resulted in downregulation of three known ATF5 target genes [Mcl-1 (13), Bcl-2 (15, 30), and asparagine synthetase (8)] at the mRNA level (Fig. 1E and Supplementary Fig. S2). However, for Mcl-1 and Bcl-2, this decrease was transitory with mRNA levels returning to baseline by 72 hours, presumably by compensatory mechanisms. In contrast, CP-d/n-ATF5-S1 did not decrease mRNA levels of Usp9X, a gene not described as transcriptionally regulated by ATF5 (Fig. 1E).

ATF5 is expressed in a variety of human cancers including glioblastoma (11, 15, 18). In silico analysis of the Rembrandt dataset for glioblastoma shows a significantly worse overall survival in patients harboring an amplification of the ATF5 gene compared with those with ≤1.8 copies (Fig. 1F). Several studies have also reported an inverse relationship between ATF5 protein expression and glioblastoma patient survival (5).

Inhibition of ATF5 function or expression has marked antineoplastic effects in vitro and in vivo (11, 15, 16, 18). Initially, to assess the activity of CP-d/n-ATF5-S1, T98G, U87MG glioblastoma, and HL-60 myeloid leukemia cells were treated for 72 hours with increasing concentrations of the peptide. CP-d/n-ATF5-S1 yielded a dose-dependent antiproliferative effect as determined by MTT assay (Fig. 1G and Supplementary Fig. S3) as well as marked changes in cellular morphology observed by light microscopy (Fig. 1H). To further assess the mechanism of such effects, we performed Annexin V/PI staining across a variety of therapy-refractory human cancer cell lines after treatment with increasing concentrations of CP-d/n-ATF5-S1 for 48 hours. As shown in Fig. 2A–C, Supplementary Figs. S4A, S4C, and S5A, the peptide yielded a strong and dose-dependent increase in the fraction of Annexin V–positive cells, thus indicating an apoptotic response across a wide and diverse panel of solid and nonsolid cancer cells. Moreover, this effect was attenuated by treatment with the pan-caspase inhibitor z-VAD-Fmk in T98G cells (Supplementary Fig. S6), which is consistent with studies indicating that interference with ATF5 function or expression in tumor cells promotes apoptosis (9, 12).

To verify whether the effect of the peptide on survival is specifically related to the dominant-negative domain and not to the cell-penetrating domain, we treated T98G cells either with Penetratin alone or CP-d/n-ATF5-S1. In contrast to CP-d/n-ATF5-S1, Penetratin did not markedly increase the fraction of Annexin V–positive cells (Fig. 2D and E). Moreover, neither Penetratin nor CP-d/n-ATF5-S1 resulted in a significant induction of apoptosis in human fetal astrocyte cultures (Fig. 2F and G), suggesting that CP-d/n-ATF5-S1 possesses specificity toward cancer cells.

Next, we addressed whether activation of apoptosis by CP-d/n-ATF5-S1 is mediated at least in part through a mitochondrial pathway. JC1 staining revealed that peptide treatment leads to a marked reduction of mitochondrial membrane potential (Fig. 2H) and cleavage (activation) of caspase-9 (Fig. 2I) suggesting that CP-d/n-ATF5-S1 induces apoptosis in part through the mitochondrially driven apoptotic pathway.

Because our findings pointed toward involvement of mitochondria in apoptosis driven by CP-d/n-ATF5-S1, we next focused on expression of Bcl-2 family proteins. ATF5 is reported to regulate transcription of antiapoptotic Bcl-2 and Mcl-1 (13, 30). However, at least in T98G cells, there was a rebound of Mcl-1 and Bcl-2 mRNA expression by 3 days of peptide treatment (Fig. 1E), so it was important to assess protein levels at this time as well. As shown in Fig. 3A and Supplementary Fig. S7A, Mcl-1 was consistently downregulated in all lines tested (U87MG, T98G glioblastoma, and H1975 non–small cell lung cancer, PANC-1 pancreatic carcinoma, A375 melanoma, and PC3 prostate cancer) at 72 hours of peptide treatment and in some cases, by 48 hours. Bcl-2 protein was similarly downregulated in all but the PC3 cell line. Expression of Bcl-xL, a third member of the antiapoptotic Bcl-2 family, was altered in some lines (U87MG and T98G), but not in others. Despite a rebound of Mcl-1 and Bcl-2 mRNA levels in T98G cells by 3 days of CP-d/n-ATF5-S1 treatment, expression of the corresponding proteins was still decreased in these cells after 6 days of peptide treatment (Supplementary Fig. S1F).

The observed decreases in Mcl-1 and Bcl-2 proteins at 72 hours promoted by CP-d/n-ATF5-S1 under conditions in which mRNA levels appear to be unaffected led us to next assess whether the peptide affects expression of Bcl-2 family members by a posttranscriptional mechanism. For instance, Mcl-1 is stabilized by its chaperone Bcl-2–associated athanogene 3 (Bag3) on one hand and by deubiquitination through the ubiquitin-specific peptidase 9, X-linked (Usp9X) on the other (31, 32). Therefore, we determined protein levels of Bag3 and Usp9X in a panel of tumor lines following treatment with increasing concentrations of CP-d/n-ATF5-S1 for 48 to 72 hours. Usp9X expression was greatly reduced by peptide treatment in a time- and dose-dependent manner, whereas Bag3 levels fell particularly in U87MG, T98G, and PC3 cells (Fig. 3A and Supplementary Fig. S7A and S7B). CP-d/n-ATF5-S1–mediated reduction in Usp9X protein levels was not rescued by z-VAD-fmk, indicating that Usp9X depletion is most likely independent of apoptosis and activated caspases as well as of its mRNA levels (Supplementary Fig. S8 and Fig. 1E).

The consistent effect of CP-d/n-ATF5-S1 on Usp9X expression led us to examine whether silencing Usp9X by another means would be sufficient to phenocopy the proapoptotic effect of the peptide. Similarly to cells treated with CP-d/n-ATF5-S1, T98G, U251, and LN229 glioblastoma cells in which Usp9X was silenced with siRNA showed marked reduction in viability as indicated by Annexin V/PI staining (Fig. 3B and Supplementary Fig. S9A). This was accompanied by a substantial increase in cleavage of caspase-9 and -3 (Fig. 3C and Supplementary Fig. S9B). In addition, there was significant reduction in Bag3, Mcl-1, and Bcl-2 protein expression in both U251 and T98G cells. Taken together, these findings indicate that loss of Usp9X expression promoted by CP-d/n-ATF5-S1 treatment is sufficient to diminish levels of key antiapoptotic Bcl-2 family members and to induce cell death.

Mcl-1 is a major resistance factor toward BH3-mimetics such as ABT737/263/199, and a considerable number of solid malignant tumors, including gliomas, bear high levels of Mcl-1 (33, 34). Given that CP-d/n-ATF5-S1 modulates antiapoptotic members of the Bcl-2 family, especially Mcl-1 and its interacting proteins, we examined whether the peptide may act in a complementary or synergistic fashion with BH3-mimetic agents. We accordingly treated T98G cells with CP-d/n-ATF5-S1 and Bcl-2/Bcl-xL inhibitor ABT263 or the Bcl-2/Bcl-xL/Mcl-1 inhibitor GX15-070. In both instances, combined treatment caused synergistic inhibition of cell viability as assessed by MTT assay (Fig. 4A and Supplementary Fig. S10A). In concordance, cellular morphology was markedly changed in cells subjected to the combination treatments and pointed, as anticipated, toward apoptosis as the underlying mechanism (Fig. 4B). Because GX15-070 has activities in addition to Bcl-2 family inhibition (35), we focused further combinatorial studies on ABT263. Enhancement of ABT263-mediated apoptosis by CP-d/n-ATF5-S1 was confirmed by Annexin V/PI staining. The combination treatment significantly upregulated the fraction of Annexin V–positive cells in T98G, LN229, SF188 (pediatric), NCH644 (glioma stem-like), and GBM12 glioblastoma cultures as well as in PANC-1 pancreatic carcinoma, A375 melanoma, K562 chronic myeloid leukemia (in blast crisis), and HCT116 colorectal cancer cell cultures (Fig. 4C and E and Supplementary Figs. S5B and S11). Consistent with these findings, combined treatment with CP-d/n-ATF5-S1 and ABT263 also enhanced caspase-9 cleavage in T98G cells (Fig. 4F).

In the context of these experiments, we also assessed a Penetratin-only peptide and a form of CP-d/n-ATF5-S1 mutated in the extended leucine zipper (Fig. 1A) to diminish its interaction with other proteins. In comparison with CP-d/n-ATF5-S1, the mutated peptide showed markedly less effect on T98G cell morphology either alone or in combination with ABT263 (Fig. 4B). The mutated peptide also showed much less effect on apoptosis when applied alone and minimally enhanced apoptosis when combined with the BH3-mimetic (Fig. 4D). Moreover, combined treatment with the Penetratin peptide and ABT263 resulted in an antagonistic antiproliferative effect (Supplementary Fig. S10B).

We next examined the effect of combined treatment with CP-d/n-ATF5-S1 and ABT263 on expression of antiapoptotic Bcl-2 family members. As illustrated in Fig. 4F, ABT263 alone increased expression of Mcl-1, a finding that represents a generally accepted mechanism of resistance to BH3-mimetic compounds (34). However, when ABT263 was combined with CP-d/n-ATF5-S1, Mcl-1 expression was highly suppressed, as was expression of Bcl-2 and Bcl-xL (Fig. 4F). In contrast, combined treatment with ABT263 and mutated CP-d/n-ATF5-S1 yielded only a slight decrease in Mcl-1 and Bcl-2 expression (Fig. 4F).

Part of the rationale for combining CP-d/n-ATF5-S1 with ABT263 is that unlike the latter, the former promotes Mcl-1 downregulation. To examine whether downregulation of Mcl-1, as occurs with CP-d/n-ATF5-S1, is sufficient to sensitize for ABT263-mediated apoptosis, we silenced Mcl-1 in PANC-1 cells with siRNA prior to treatment with ABT263 (Supplementary Fig. S12A). Mcl-1 knockdown combined with ABT263 yielded markedly enhanced cleavage of caspase-9 and -3. In addition, Bag3, Usp9X, and Bcl-2 expression was significantly reduced under these conditions compared with cells either silenced for Mcl-1 and/or treated with ABT263 alone. These observations were also reflected by an enhanced reduction in the fraction of viable LN229 cells remaining after silencing Mcl-1 and treating with ABT263, as compared with treating with either alone (Supplementary Fig. S12B). Similarly, when Usp9X was silenced, LN229 cells became more susceptible to the cytotoxic effects of ABT263 (Supplementary Fig. S12C).

Thus, when combined with ABT263, specific knockdown of Mcl-1 and Usp9X (as seen after treatment with CP-d/n-ATF5-S1) suffices to reproduce the molecular profile of combined treatment with CP-d/n-ATF5-S1 and ABT263.

Next, we examined whether combined treatment with CP-d/n-ATF5-S1 also enhances apoptosis triggered by the extrinsic pathway. Our reasoning was that if the peptide increases sensitivity to the mitochondrial apoptotic pathway, it might complement or enhance mitochondrial-dependent and/or mitochondrial-independent apoptotic actions of a death-promoting ligand. We therefore treated T98G cells with CP-d/n-ATF5-S1 and increasing concentrations of TNF-related apoptosis-inducing ligand (TRAIL). As shown in Fig. 5A, treatment with this combination results in an enhanced antiproliferative effect in the MTT assay compared with control or single treatments. In contrast, the combination of TRAIL with mutated CP-d/n-ATF5-S1 did not show this effect. Representative microphotographs in Fig. 5B illustrate these findings at the level of morphology. Annexin V/PI staining also showed that CP-d/n-ATF5-S1 enhances TRAIL-mediated apoptosis in T98G cells as well as in LN229 glioblastoma cells and MDA-MB-436 breast cancer cells (Fig. 5C and E). In this assay, mutated CP-d/n-ATF5-S1 alone only slightly increased apoptotic cells when compared with controls and did not enhance TRAIL-induced apoptosis (Fig. 5D). In concordance with these findings, the combination therapy led to reduced expression of full-length caspase-3 in T98G cells, presumably due to elevated cleavage of this protein (Fig. 5F). Mutated CP-d/n-ATF5-S1 did not have this effect. In addition, combined treatment with CP-d/n-ATF5-S1 and TRAIL enhanced downregulation of Mcl-1 and Bcl-2 expression (Fig. 5F). Treatment with TRAIL alone in this cell line reduced expression of Bcl-xL, although this effect was neither matched nor enhanced by CP-d/n-ATF5-S1 (Fig. 5F).

Decreased expression of Mcl-1 following treatment with CP-d/n-ATF5-S1 represents a mechanism likely to contribute to the CP-d/n-ATF5-S1–mediated sensitization toward TRAIL. Because Bag3 stabilizes Mcl-1 (23) and our data indicate that CP-d/n-ATF5-S1 downregulates Bag3 in most cell lines tested (Fig. 5G), we examined whether Bag3 knockdown would phenocopy the sensitizing effect of CP-d/n-ATF5-S1 toward TRAIL. Silencing Bag3 in LN229 glioblastoma cells results in downregulation of Mcl-1 (Fig. 5G). When combined with TRAIL, Bag3 knockdown markedly increased cleavage of caspase-9 and -3. Consistent with these observations, treatment with Bag3-siRNA and TRAIL yields a significant increase in apoptosis of LN229 cells as determined by Annexin V/PI staining (Fig. 5H). Moreover, TRAIL combined with silencing of Mcl-1 with siRNA also markedly increased cleavage of caspase-9 and -3 and apoptosis in LN229 cells (Fig. 5I and J). In contrast, Bag3 and Usp9X levels were not affected by Mcl-1 knockdown alone or when combined with TRAIL (Fig. 5I). Taken together, these findings indicate that CP-d/n-ATF5-S1 sensitizes tumor cells to TRAIL and that this occurs at least in part by loss of Mcl-1 due to reduction of Bag3 expression.

We next assessed the therapeutic efficacy of CP-d/n-ATF5-S1 in multiple murine xenograft models. U87MG glioblastoma, A375 melanoma, PC3 prostate cancer, PANC-1 pancreatic cancer, and HCT116 colorectal cancer cells were implanted subcutaneously; MDA-MB-231 triple-negative breast cancer cells were implanted in the mammary fat pad; and GBM12 patient-derived xenograft cells were implanted intracranially. Once tumors formed, mice were randomized and treated with CP-d/n-ATF5-S1, vehicle, or penetratin peptide as outlined in Fig. 6A–C and Supplementary Figs. S5C, S13, and S14. Under these conditions, except for the cases of PANC-1 cells (P = 0.25) and HCT116 cells (P = 0.18), in all tumor types animals that received treatment with CP-d/n-ATF5-S1 had significantly smaller tumors than the animals treated with vehicle or Penetratin (Fig. 6B and C and Supplementary Figs. S5C, S13, and S14). Moreover, in a GBM12 intracranial patient-derived xenograft model, animals treated with CP-d/n-ATF5-S1 showed a median survival of 38 days, which was significantly prolonged compared with 22.5 days in animals receiving vehicle (Fig. 6A). Although the treatments used here affected tumor growth rate, for the most part, they did not result in statistically significant regression of tumors. However, there was a statistically significant tumor regression in mice bearing MDA-MB-231 breast cancer mammary fat pad xenografts (Supplementary Fig. S14E–S14G). This effect was not observed when mice were treated with Penetratin alone (Supplementary Fig. S14E–S14G).

To detect possible toxic effects due to peptide treatment, histologic analysis was performed on various tumor-free tissues of animals treated with either vehicle or CP-d/n-ATF5-S1 according to the dosing schedule described in Fig. 6C. No tissue alterations in brain, lung, kidney, heart, liver, spleen, and intestine were found (Supplementary Fig. S15A). Moreover, the body weights of the animals did not vary between the treatment groups toward the end of the experiment (Supplementary Fig. S15B).

Our in vitro studies indicated that combined CP-d/n-ATF5-S1 and ABT263 treatment enhanced tumor cell death due to additive and complementary effects on antiapoptotic Bcl-2 family members. To assess whether the combination is more effective in vivo than either treatment alone, we utilized a U251 heterotopic glioblastoma xenograft model and a HCT116 heterotopic colorectal cancer xenograft model (Fig. 6D and Supplementary Fig. S5C). The mice with xenografted tumors were divided into four groups: vehicle, ABT263, CP-d/n-ATF5-S1, or the combination of ABT263 and CP-d/n-ATF5-S1. As shown in Fig. 6D, in the U251 model, at the endpoint of the study, animals that received the combination treatment had significantly smaller tumors compared with those treated either with ABT263 or CP-d/n-ATF5-S1 alone and showed a decrease in tumor size over time when compared with the beginning of treatment. Similarly, in the HCT116 model, the combination treatment led to significant reduction in tumor growth rate when compared with vehicle or single-agent treatments (Supplementary Fig. S5C). The combination treatment also showed no clinical signs of toxicity, indicating that although the combined treatment is more efficient, it does not increase the occurrence of evident side effects.

Cancer cells typically develop primary or secondary resistance to apoptosis (36). Therefore, means to manipulate the apoptotic machinery are pivotal to restore therapeutic sensitivity. Deregulation of the apoptotic machinery is mediated through numerous factors, such as the Bcl-2 family of proteins, the Inhibitor of Apoptosis Proteins, and expression of death receptors, initiator caspases, and endogenous caspase inhibitors (37, 38). The aberrant expression of such molecules is regulated by various means, including the actions of transcription factors. ATF5 is an example of a transcription factor with oncogenic potential that affects Bcl-2 family member expression (13, 16). ATF5 is upregulated in various malignancies, including highly prevalent tumors, such as breast carcinoma (11), but it is also increased in less common malignancies, such as low- and high-grade gliomas (13, 16). In the context of low- and high-grade gliomas, ATF5 expression levels are not only increased, but also correlate with survival (13). Thus, ATF5 represents a potential target in treatment-refractory cancers.

Here, we show that a novel synthetic cell-penetrating dominant-negative ATF5 peptide induces apoptosis in a broad range of tumor types, including glioblastoma, triple-negative breast cancer (MDA-MB-436), hormone-refractory prostate cancer (PC3 and DU145), EGFR kinase inhibitor resistant non–small cell lung cancer (H1975), BRAF (V600E)-mutated melanoma (A375), and pancreatic carcinoma (PANC-1). CP-d/n-ATF5-S1 also showed in vivo efficacy in reducing growth of a range of tumor types in xenograft models. We have yet to optimize dosing or regimens of administration. Our in vitro and in vivo studies indicate that the peptide does not kill nontransformed cells and causes no evident histologic or behavioral signs of toxicity in mice at levels up to at least 150 mg/kg. With respect to specificity, a Penetratin peptide and a peptide in which key leucine residues in CP-d/n-ATF5-S1 were mutated showed no or little apoptotic activity alone or in combination studies. The fact that ATF5, as a prosurvival factor is overexpressed in cancer cells, may cause a state of cellular ATF5 dependency in the sense of oncogene addiction. This would explain the selective response toward CP-d/n-ATF5-S1 treatment in cancer cells as a consequence of a sudden CP-d/n-ATF5-S1–mediated loss of ATF5 function.

CP-d/n-ATF5-S1–mediated cell death was accompanied by depletion of endogenous ATF5 protein, suggesting that CP-d/n-ATF5-S1 may induce cell death in part through depletion of total ATF5 levels. In that context, earlier results showed that depletion of ATF5 by siRNA/shRNA leads to cell death in a broad variety of tumor cells (11, 13, 15). It remains to be determined by what underlying mechanisms CP-d/n-ATF5-S1 controls ATF5 protein levels. Our observations suggest that this is a nontranscriptional event. Given that ATF5 possesses a short-half life (39) and is stabilized by chaperones (39), it is conceivable that CP-d/n-ATF5-S1 enhances ATF5's degradation through disruption of its interactions with other binding partners.

Consistent with its activation of the intrinsic apoptotic pathway, CP-d/n-ATF5-S1 modulated the expression of the antiapoptotic Bcl-2 protein family, including Mcl-1, Bcl-2, and, in some instances, Bcl-xL. Particularly, at earlier time points, there is an occasional CP-d/n-ATF5-S1–mediated increase in Bcl-xL, which, however, for most cell lines (except for PANC1) tested is not sustained. Additional research will shed light on why CP-d/n-ATF5-S1 causes this biphasic modulation of Bcl-xL. Nevertheless, one possible consequence might be that the increase in Bcl-xL might render tumor cells more sensitive to Bcl-xL inhibitors, such as ABT263. These are known to counteract Bax-dependent apoptosis by preventing mitochondrial outer membrane permeabilization and subsequent cytochrome c release and caspase-9 activation.

Although Bcl-2 and Mcl-1 have been described as transcriptional targets of ATF5, our findings indicate that CP-d/n-ATF5-S1 causes only a transient decrease in their transcript levels and that Mcl-1, in particular, is subject to sustained downregulation at the posttranscriptional level by CP-d/n-ATF5-S1. Several molecules have been described that control Mcl-1 levels posttranslationally. Examples include MULE (40), Bag3 (31), and Usp9X (32). Bag3 is a co-chaperone of Hsp70 (41) and binds Mcl-1 to prevent its degradation, whereas Usp9X is a deubiquitinase that removes ubiquitin chains from Mcl-1, rendering it resistant to proteasomal degradation and thereby in turn increasing its half-life (32). Both Bag3 and Usp9X have been shown to counteract intrinsic apoptosis. For instance, Bag3 is upregulated in malignant gliomas and interferes with Bax-mediated apoptosis (42), whereas Usp9X knockdown enhances ABT263-mediated cell death in glioblastoma (20, 33). Usp9X interacts with a variety of molecules in addition to Bcl-2 proteins that may also affect cell survival (43, 44) and these too may thus play a role in tumor cell death promoted by CP-d/n-ATF5-S1.

Our findings show that in PC3, PANC-1, T98G, H1975, A375, and U87MG cells, CP-d/n-ATF5-S1 significantly affects protein levels of Usp9X starting as early as 48 hours and continuing at 72 hours after treatment, when apoptotic death is manifest. To assess the impact of Usp9X depletion by CP-d/n-ATF5-S1, we transfected glioblastoma cells with Usp9X siRNA and found that this was sufficient to induce significant apoptotic death. Mechanistically, Usp9X knockdown caused concomitant suppression of Bag3, Mcl-1, and Bcl-2 expression, which remarkably recapitulates the effects of CP-d/n-ATF5-S1 on these molecules. These observations suggest that CP-d/n-ATF5-S1–mediated suppression of Usp9X levels may be an instrumental mechanism by which it mediates death of neoplastic cells. Although Usp9X is known to modulate Mcl-1 expression, our observed effects of Usp9X manipulation on Bag3 and Bcl-2 have not been previously described and may suggest that Usp9X also interacts with Bag3 and Bcl-2. The mechanism(s) by which CP-d/n-ATF5-S1 decreases Usp9X expression remain to be explored. Our data indicate a posttranscriptional mechanism in that the peptide does not affect Usp9X mRNA levels. The strong antineoplastic activity related to Usp9X downregulation warrants further studies directed at identification of specific small-molecule inhibitors of Usp9X function.

Given our observation that CP-d/n-ATF5-S1 strongly affects the intrinsic apoptotic machinery and the possibility that treatment with a single drug may fall short in the clinic, we investigated whether rational drug combination therapies could enhance the efficacy of CP-d/n-ATF5-S1. For that purpose, we utilized the orally available BH3-mimetic ABT263 (43). This class of compounds has received great attention because they target both Bcl-2 and Bcl-xL, which are upregulated in many malignancies, especially in hematologic malignancies, such as follicular lymphoma (45), and also in solid neoplasms, such as glioblastoma (46). Although certain tumors demonstrate remarkable sensitivity for BH3-mimetics, others reveal resistance, which in the vast majority of cases is attributed to high-levels of Mcl-1 expression (47). Therefore, means to counteract high Mcl-1 protein levels may sensitize resistant tumors to BH3-mimetic treatment (48, 49). In the current case, we found that CP-d/n-ATF5-S1 strongly depresses expression of two Mcl-1–interacting proteins, Bag3 and Usp9X, and this in turn leads to Mcl-1 depletion. These considerations led us to assess the combination of CP-d/n-ATF5-S1 and ABT263 in vitro and in in vivo xenograft tumor models. We found enhanced cell death by this combination in a variety of tumor cell lines, including LN229 and A375 that are relatively resistant to ABT263. In the in vivo models, the combination was highly effective compared with the single treatments and blocked/reduced tumor growth over the course of the studies.

Because the Bcl-2 family is also implicated in extrinsic apoptosis, we tested whether CP-d/n-ATF5-S1 overcomes resistance to TRAIL. TRAIL has received attention for its ability to kill a broad variety of cancer cells in vitro and in vivo. One main obstacle for TRAIL-related therapies is that although a subset of tumors respond, the majority display resistance (50). Therefore, efforts have aimed to identify treatments that sensitize cancer cells to TRAIL therapeutics. Our results suggest that CP-d/n-ATF5-S1 is a potent sensitizer for TRAIL-mediated apoptosis. Mechanistically, this is most likely linked to the ability of CP-d/n-ATF5-S1 to suppress Bag3 and Mcl-1, as specific knockdown of Bag3 or Mcl-1 was sufficient to sensitize TRAIL-resistant LN229 glioblastoma cells to apoptosis.

Overall, our results serve as a proof-of-principle and suggest that the strategy of treatment with a cell-penetrating dominant-negative form of ATF5 is efficacious, selective, and nontoxic and therefore holds promise for cancer therapy, either alone or in a multitargeting approach.

L.A. Greene is a consultant/advisory board member for and reports receiving commercial research grants from Sapience Therapeutics. L.A. Greene and J.M. Angelastro are listed as coinventors on patents entitled “Methods for promoting apoptosis and treating tumor cells by inhibiting the expression or function of the transcription factor ATF5” and “Methods for inhibiting the differentiation of proliferative telencephalic cells in vitro by addition of ATF5,” which are owned by Columbia University, and on a provisional patent application for “Compositions and Methods for Inhibiting Tumor Cells by Inhibiting the Transcription Factor ATF5,” which is owned by Columbia University and University of California, Davis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G. Karpel-Massler, B.A. Horst, L.A. Greene, J.M. Angelastro, M.D. Siegelin

Development of methodology: G. Karpel-Massler, L.A. Greene, J.M. Angelastro, M.D. Siegelin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Karpel-Massler, C. Shu, L. Chau, T. Tsujiuchi, M.D. Siegelin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Karpel-Massler, L. Chau, L.A. Greene, M.D. Siegelin

Writing, review, and/or revision of the manuscript: G. Karpel-Massler, B.A. Horst, J.N. Bruce, P. Canoll, L.A. Greene, J.M. Angelastro, M.D. Siegelin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Karpel-Massler, B.A. Horst, C. Shu, L. Chau, J.N. Bruce, M.D. Siegelin

Study supervision: L.A. Greene, M.D. Siegelin

The authors thank Yanping Sun for excellent assistance with the MRI studies.

This work was supported by a scholarship from the Dr. Mildred Scheel Foundation of the German Cancer Aid (to G. Karpel-Massler) and American Brain Tumor Association, Translational Grant 2013 (ABTACU13-0098), the 2013 AACR-National Brain Tumor Society Career Development Award for Translational Brain Tumor Research (13-20-23-SIEG), NIH NINDS award K08NS083732 (to M.D. Siegelin), and NIH NINDS grant R01NS083795 (to J.M. Angelastro and L.A. Greene).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.