PerspectiveDrug Development

Pancreas Cancer Meets the Thunder God

See allHide authors and affiliations

Science Translational Medicine  17 Oct 2012:
Vol. 4, Issue 156, pp. 156ps21
DOI: 10.1126/scitranslmed.3004956


A new formulation of a natural product shows remarkable activity against pancreatic ductal adenocarcinoma across a number of preclinical model systems. These findings set the stage for a clinical trial.

Looming over every attempt to bridge the divide between scientific discovery and clinical therapy is the specter of a dismal collective success rate thus far. More than 90% of cancer patients who enroll in early-stage clinical trials experience no benefit while, nevertheless, risking exposure to potential toxicities (1). The vast majority of ideas born in basic science and matured in preclinical assays die in the clinic. With pancreatic ductal adenocarcinoma (PDA), both the need and the failure are particularly acute: Among the many ways this cancer distinguishes itself, it ranks highest in 1- and 5-year mortality rates for all malignancies. With a median survival of 6 months for metastatic disease, the most frequent stage at diagnosis, PDA is rapidly, often painfully, and almost uniformly fatal (2). Over the past 15 years, with very few exceptions, randomized phase 2 and 3 trials have failed to improve on the current standard regimen of single agent gemcitabine. Clearly, doing the same thing over and over again and expecting different results is neither appealing nor sound.

The large number of phase 3 trials that fail, particularly in pancreas cancer, suggests that the process leading up to that trial also failed. Given the tremendous costs in patient’s lives, families’ grief, caregivers’ hours, and a nation’s resources, we cannot afford to continue to fail so regularly. What should the standards be for advancing a new therapeutic strategy from idea to practice? What experimental system(s) are most predictive of the human response? Our failures, as well as successes, have taught us that we must match the most relevant platforms and end points and use the data to inform clinical application. We must also extend and deepen the correlative science that accompanies early-phase trials to apply more stringent tests of what should move forward. All of these challenges are admittedly easier to describe than overcome. Chugh et al. take up the gauntlet and perform an extensive, multiplatform, preclinical vetting of a novel therapy offering a strategy toward a more rigorous and predictive assessment of the potential to improve patient survival (3).

The authors have developed a new formulation (minnelide) of a natural product (triptolide) that has been shown to have potent therapeutic properties in a variety of studies and through its long history of use in traditional Chinese medicine. To test the suitability of this compound in treating pancreas cancer, the authors used four preclinical platforms: (i) in vitro survival assays in established human PDA cell lines; (ii) orthotopic transplants of the same cell lines into immune-deficient animals; (iii) heterotopic transplants of patient-derived tumor pieces into immune-deficient animals; and (iv) a genetically engineered mouse (GEM) model of spontaneous autochthonous PDA in the setting of an intact immune system (Fig. 1). Collectively, their experiments address two broad areas of therapeutic need in PDA: chemoprevention and treatment. Treatments for early and advanced invasive disease are the most pressing needs facing the practicing oncologist. Within the realm of treatment, results from the various platforms may also speak to specific uses in the adjuvant or neoadjuvant settings. With a rising incidence and unabated mortality, strategies that prevent precursor lesions from progressing would also be welcome. This may represent the “easier” challenge if, indeed, it is possible to identify the patients at highest risk and therefore most likely to benefit.

Fig. 1.

Preclinical parade. Salient features of pancreas cancer model systems are shown. Chugh et al. used several of these models in their studies on a novel therapeutic agent.



The perennial plant Trypterygium wilfordii Hook. f. has long been used for arthritic ailments in Chinese medicine, where it is known as lei gong teng (“thunder god vine”). T. wilfordii contains more than 100 bioactive compounds, and extracts from this plant have been shown to be superior to both placebo and the anti-inflammatory antibiotic sulfasalazine in randomized, double-blind studies in rheumatoid arthritis patients (4). Triptolide, a diterpenoid triepoxide, was first isolated from this plant and its antileukemic properties characterized by Kupchan et al. in 1972 (5). It has been shown to be anti-inflammatory and immunosuppressive both in vitro and in vivo, and its clinical use has been investigated in a number of disordered immune states, including multiple sclerosis, lupus nephritis, colitis, and transplant rejection [reviewed in (6)].

A major triumph of 19th-century pathology was establishing that neoplastic processes are distinct from infectious and inflammatory ones (7). From late in the 20th century, however, it became increasingly clear that these are not as unrelated as the macroscopic appearance and the histopathology had suggested. The growing awareness that inflammatory processes are pro-carcinogenic and clear evidence that anti-inflammatory agents can be anti-neoplastic (8) make it highly appropriate to test specific known anti-inflammatories for their preventive or therapeutic role in cancer. In that light, it would be useful to know whether patients with inflammatory disorders treated long-term with T. wilfordii extracts have a lower cancer incidence than expected; this is established, at least for colorectal cancer, in rheumatoid arthritis patients on standard anti-inflammatories (although they also show an excess of lymphomas) (9).

Triptolide inhibits proliferation and/or induces apoptosis in a wide variety of epithelial and hematological cancer cell lines, and an extensive range of potential mechanisms have been invoked (10). It may even have potential as a male contraceptive because of its reversible effects on sperm count and maturation. Triptolide inhibits the synthesis of a number of inflammatory cytokines, including interferons 1, 2, 6, and 8 and tumor necrosis factor (TNF)–α; enzymes involved in inflammation such as PGE-2, iNOS, and metalloproteinases; transcription factors, including NFκB; and RNA polymerase II. The authors’ own prior in vitro and orthotopic studies with human pancreas cell lines have shown that triptolide suppresses HSP70 at the mRNA and protein levels and induces apoptosis by both caspase-dependent and -independent means (11). They have reported similar effects in neuroblastoma (12) and additional mechanisms of cell death in cholangiocarcinoma cell lines (13). It is fair to say that the specific mechanism(s) of action in most of these contexts have yet to be clearly delineated. Indeed, analysis by the COMPARE algorithm of triptolide’s effects on the NCI-60 panel of cell lines suggests a unique and as-yet-undiscovered pharmacological mode of action, distinct from conventional cytotoxics and antimetabolites (14).

Despite the compelling array of diseases that may be treated with this compound, the clinical utility of triptolide is limited by its poor solubility in aqueous solutions and poor stability. Prodrug formulations have been developed to overcome these limitations and enable parenteral use. One formulation (PG490-88), a sodium succinate salt, has previously been tested against solid tumors in phase 1 trials (15). The prodrug is hydrolyzed by plasma esterase to release the active compound, a process that proved to be relatively slow and fraught with unpredictable pharmacokinetics, presumably because of varying circulating enzyme levels in patients (16). The present team of investigators has synthesized a phosphonooxymethyl form of the drug dubbed minnelide (Minnesota + triptolide). The compound undergoes a two-step bioconversion in vivo: The first and rate-limiting step is dephosphorylation by alkaline phosphatase, which is followed by rapid nonenzymatic breakdown to release triptolide and a hydroxymethyl group. The latter serves as a precursor for formaldehyde, which itself is rapidly converted to formic acid and eliminated. The release of formaldehyde in molar stoichiometry may raise concern; however, the same process is involved in the metabolism of other molecules in the body, including glycine, serine, and choline, and it is estimated that ~50 g of formaldehyde is turned over per day in humans (17). Thus, the authors cleverly solve the problems of solubility and commercialization (natural compounds are not patentable) in a process that is also likely to provide good bioavailability. The discernible toxicities thus far appear to be related to sites of high alkaline phosphatase activity and include bone, liver, kidney, placenta, and gut (although, in the experiments performed here, no significant toxicities were observed).


For their in vitro and orthotopic studies, the authors used four different and complementary established human PDA lines: MIA PaCa-2, AsPC-1, S2-013, and S2-VP10, derived, respectively, from a primary tumor, from malignant ascites, and as two independent subclones from the SUIT-2 liver metastasis cell line (3). The in vitro studies demonstrated growth inhibitory concentration ranges for minnelide similar to those seen previously for triptolide and also showed that exposure to alkaline phosphatase was required for minnelide activity. MIA PaCa-2 cells were most sensitive in vitro, which the authors attribute to differences between primary and metastatic cell lines; however, they have also previously reported different mechanisms of cell death for minnelide in these two groups of cell lines (18). Apart from MIA PaCa-2, the cell lines did not show a dose response in the range tested (100 to 200 nM). Intraperitoneal (i.p.) delivery of minnelide was also effective at preventing tumor growth from MIA PaCa-2 or S2-013 cells embedded in Matrigel and injected orthotopically into the pancreatic tail of athymic mice. Once again, no dose response was observed in survival analyses of recipient mice. Although formal pharmacokinetic (PK) studies were not reported, twice-daily dosing (BID) appeared to be superior in suppressing tumor growth to once per day (QD) for the same total dose, providing some information on this topic. Local and metastatic spread by S2-013 cells, manifested clinically by the development of jaundice and ascites, were also inhibited.

Raising the bar further, after establishing that tumor growth could also be prevented in mice with orthotopically transplanted AsPC-1 cells, the authors discontinued therapy after ~3 months in a subset of animals and followed them for an additional 9 months, at which point palpable tumors had still not emerged. As encouraging as these results are, regression of established tumors represents a greater challenge and implies a cytotoxic rather than merely cytostatic capability. To this end, after establishing palpable orthotopic tumors with AsPC-1 cells, the authors showed that daily i.p. dosing with minnelide caused essentially complete tumor eradication, and all animals were alive at 75 days; control animals had very large tumors and a median survival of 36 days.

Having largely exhausted what could be gleaned from orthotopically injected cell lines, the authors explored patient-derived xenografts. In these experiments, pieces from patient-derived tumors are serially expanded under the skin of immune-deficient mice. Serial transplantation expands the tumor epithelium, generating sufficient tissue for subsequent experiments, albeit at the expense of a corresponding depletion of the associated human stromal elements. Thus, the investigators expanded tumor chunks from two different primary human PDAs in the flanks of SCID mice and began treating the host animals with minnelide after the third passage. Tumors that averaged 300 mm3 became clinically undetectable after 40 days of treatment and remained so even after 65 days off therapy (although four animals succumbed to thymomas, perhaps reflecting immune system effects). Even more impressively, 1000 mm3 tumors could also be eradicated, this time within 35 days; half the daily dose of minnelide achieved the same end point with slightly slower kinetics (50 days).

Acknowledging the likely importance of a native tumor microenvironment (TME) and intact immunity to cancer development, the authors also tested their compound in the KrasLSL-G12D/+;Trp53LSL-R172H/+;Pdx-1-Cre (KPC) GEM model of PDA (19). KPC mice develop the full spectrum of preinvasive ductal lesions, or pancreatic intraepithelial neoplasms (PanINs), and these lesions spontaneously progress to invasive and widely metastatic PDA. The PanINs manifest shortly after birth, become locally invasive by 10 to 12 weeks, and culminate in a metastatic burden to which the mice succumb at a median age of 5 to 6 months. The histological progression of both the epithelial and stromal compartments as well as the aberrantly activated signaling pathways faithfully recapitulate the human disease. The animals also manifest a clinical syndrome—including cachexia, jaundice, and ascites—that mirrors the patient experience. KPC mice at 4 to 6 weeks of age (that is, at the preinvasive stage of disease) were treated with either 0.3 mg/kg minnelide or saline i.p. QD and sacrificed at an average age of 4 months. Significantly lower tumor weights and volumes were observed in minnelide-treated animals. This chemoprevention trial points to potential uses of minnelide in patients who develop high-risk preinvasive pancreatic neoplasms, either because of genetic predisposition or underlying pancreatic disease.

Finally, in the only experiment studying the current standard for PDA treatment, the authors compared the effects of gemcitabine and minnelide in the MIA PaCa-2 orthotopic model and found minnelide to be superior in preventing tumor growth. Indeed, gemcitabine was completely ineffective in this setting, unlike significant responses seen in prior studies (20), a consequence, perhaps, of the use of Matrigel in the present study as opposed to buffered saline alone to create the injected cell suspension; it may be that the tissue density and physicomechanics of human and murine PDA are better approximated by cells suspended in a matrix than in aqueous solution.

Collectively, these data indicate that minnelide has remarkable efficacy across a range of experimental platforms and types of responses including in vitro inhibition of proliferation, prevention as well as regression of tumor formation in orthotopically grown invasive cells, regression of transplants from patient-derived whole-tumor pieces, and chemoprevention of preinvasive disease in a GEM model of spontaneous autochthonous PDA. The only study lacking from the set is a treatment trial of invasive and metastatic disease in the same GEM model (see Fig. 1).


The stepwise progression of experiments in model systems of increasing complexity and, arguably, fidelity to the human disease provides potentially useful insights to inform clinical applications. Of course, assessing the relative reliability of each of these platforms in predicting clinical behavior would have been easier had there been some interplatform disagreement in results. The tumor regression experiments in orthotopic and heterotopic systems suggest that established primary invasive and locoregional extension of disease may be treated effectively by minnelide, provided an intact immune system and fully developed TME are not prohibitive. The ability to prevent metastases suggests roles in the adjuvant setting or possibly in expanding the use and investigation of neoadjuvant regimens by decreasing the attendant risks of disease dissemination during the treatment course (21). The ability of minnelide to arrest preinvasive disease progression invites testing in high-risk cohorts. Given its potent inhibitory effects on both neoplastic epithelium and inflammation, minnelide may be an especially effective chemopreventive in patients with chronic pancreatitis, one of the highest risk factors for PDA.

There are important caveats, however, that can confound these simple extrapolations. First, both orthotopic and patient-derived xenograft systems involve introducing a wound, with its attendant inflammatory reactions and a neovasculature that, as part of wound healing, is quite unlike the anatomy of a native cancer. The fact that triptolide has been shown to be antiangiogenic may therefore be more therapeutically relevant in these settings than in the generally hypovascular environment of autochthonous PDA (20, 22). Also, unlike most patients, the host animals lack the intact immune response that is often complicit in both cancer progression and resistance. In addition, passaging of human PDA tumor xenografts depletes the accompanying stromal elements, enriching for and reorganizing the epithelial compartment (23). In this regard, formal histological and immunohistochemical analyses on recovered tissues from the various experiments would have been welcome and, in future work, may provide information on cellular effects. Indeed, no formal assessment of histology was presented for any experiment to illuminate specific effects on epithelial cells, stromal cells, immune cells, and vasculature.

One of the major advances of 20th century oncology, albeit based squarely in 19th-century evolutionary biology, was the advent of combination chemotherapy: combining cytotoxic agents with distinct mechanisms of action to provide synergistic benefits, reducing the likelihood of clonal escape and increasing response rates. The approach has achieved notable successes in hematologic malignancies and select solid tumors. We are now moving into an era in which combined therapies are directed not just at different molecular targets within the tumor epithelium but also against different compartments within the complex tumor organ (24). One important hurdle therefore was omitted, and a potential opportunity missed, in not testing the compound in KPC mice with invasive and metastatic PDA. As discussed, the GEM models recapitulate a clinical and pathophysiological progression from earliest precursor lesions to invasive and metastatic disease, including the accompanying evolution of complex mesenchymal, endothelial, and immune responses that culminate in a faithful rendering of a three-dimensional pancreas cancer. Considerable barriers to drug delivery, diffusion, and convection are erected by a fully elaborated stromal reaction in PDA (22, 25), and it remains to be seen whether effective intratumoral concentrations of minnelide or triptolide can be achieved by this strategy. Thus, the GEM models can be highly informative, not only to study chemoprevention as in the new work by Chugh et al. (3) but also in the treatment of late-stage disease.

No formal pharmacokinetic, pharmacodynamic, or toxicology data were provided in this study, and the drug was administered intraperitoneally, a route not likely to be used in patients. These types of studies with intravenous dosing have been or presumably will be completed in anticipation of the planned phase 1 trial alluded to by the authors. Given that patients with PDA frequently have biliary obstruction, hepatic dysfunction, and elevated alkaline phosphatase levels, there may be variability in effective dose from patient to patient that will need to be monitored and may require that patients with abnormal liver function have independent PK assessments. Potential effects on the male and female reproductive systems will also have to be followed, and the cluster of observed thymomas, to which these mice are admittedly predisposed, may nevertheless warrant attention, given the reported effects noted above of triptolide on lymphocytes. Finally, elucidating the mechanism(s) of action in PDA will enable development of rational combination treatment strategies as well as ways to obviate or overcome the inevitable emergence of resistance.

Chugh et al. (3) have updated an ancient remedy and cross-tested it in a complementary series of traditional and modern experimental systems. Almost exhausting the current repertoire of preclinical platforms, they have elevated the standard for advancing a therapeutic strategy to use in patients. Perhaps just as crucially, their multiplatform effort invites a rational and focused discussion about what constitutes the best approach to integrated studies from cell to cage to clinic and what best defines a translational program where preclinical success is more than just a prelude to clinical failure.

References and Notes

  1. Acknowledgments: : We thank S. Thorsen for expert assistance with manuscript preparation. Funding: This work was supported by NCI grants CA129537, CA161112, CA152249, and CA159240 to S.R.H and support to S.R.H. from the Giles W. and Elise G. Mead Foundation.
View Abstract

Navigate This Article