Cell lines

Given the difficulty in obtaining adequate biopsy material from PDA patients, numerous human PDA-derived cell lines of varying differentiation states (well, moderately and poorly differentiated) and genetic backgrounds (KRAS or TP53 wild-type or mutants) have been utilised. The most popular cell lines for functional and therapeutic studies are Panc-1 (poorly differentiated), MiaPaCa-2 (moderately differentiated) and BxPC-3 (well differentiated); other cell lines used are CaPan-1, CaPan-2 and HPAF-2. Together, studies on these cell lines have provided biological data of varying significance to the clinical situation. However, there are many limitations to cell culture models of PDA, including possible changes arising from long-term culture with multiple passages and derivation from metastases and/or propagation as xenografts (Ref. Reference Bissell and Radisky33).

Recently, three PDA cell lines (PP109, PP117, PP161) were established from primary human specimens and well characterised with respect to specific genetic alterations [KRAS, TP53, CDKN2A and SMAD4) in vitro growth characteristics, DNA profile by microsatellite fingerprinting and cytogenetics by karyotype analysis (Ref. Reference Chifenti34). Three genes were altered in all three PDA cell lines (KRAS, TP53, CDKN2A) but no mutations were detected in SMAD4. Furthermore, a complex karyotype with many chromosomal abnormalities – both structural (1p, 1q, 3p, 6q, 7q, 11p, 17p, 19q) and numerical (gains of 7 and 20; losses of 6, 12, 13, 17, and 18) – was identified. The PDA cell lines had a similar cytokeratin expression to normal ductal cells [CK7 (KRT7), CK8 (KRT8), CK18 (KRT18), CK19 (KRT19)]. The intermediate filament protein vimentin (VIM) was expressed only in PP117 and is generally associated with poorly differentiated tumours with high metastatic potential. This was confirmed by an in vitro migration assay, where PP117 had a higher motility compared with PP109 and PP161. In PP117 cells, expression of ezrin (EZR; one of the ERM protein family that ensures crosslinking between the actin cytoskeleton and the plasma membrane) was localised to membrane ruffles and short filopidia structures involved in the motility machinery, and, with coexpression of vimentin, characterises a poorly differentiated phenotype. Hence, despite a similar genetic profile in these cell lines, cellular differentiation appears to determine the aggressiveness of PDA. Therefore, dissecting the molecular mechanisms underlying the differentiation status is likely to provide insights into an improved prognostic and therapeutic signature.

Mouse models

The existence of human PDA cell lines has allowed the generation of subcutaneous flank and orthotopic (Ref. Reference Bruns35) xenograft mouse models. Despite the generally poor predictability of these mouse xenograft models they continue to be utilised as screening tools and/or proof of concept models for evaluating novel therapeutics. Although tumour growth inhibition (TGI) is an important endpoint in these studies, a better measure of activity is significantly enhanced Kaplan–Meier survival (e.g. >2 weeks) compared with control or standard therapy. For targeted therapy evaluations these models can provide a fairly robust readout of target inhibition/modulation and help develop biomarkers for Phase I clinical trials (Ref. Reference Riley36).

In addition, several genetically engineered pancreas-targeted mouse models have been generated (Refs Reference Hingorani37, Reference Aguirre38, Reference Hingorani39, Reference Izeradjene40, Reference Takaku41), based on the four most common genetic mutations (KRAS, TP53, CDKN2A, SMAD4) detected in human PDA (Ref. Reference Hruban28) (Box 1). Although these elegant mouse models appear to have histological features common to human PDA there are several drawbacks that highlight significant differences to human disease: the observed genetic changes do not reflect the chronology of mutations; the sites of metastasis do not match human disease; other histological subtypes of pancreas cancer are present; primary tumours develop in other organs (e.g. gastric cancer in KRas/Smad4 model); and numerous genes that are specific to higher vertebrates are absent in rodents [e.g. carcinoembryonic antigen-related cell adhesion molecule 6 (CEACAM6); Ref. Reference Riley36].

Xenografts derived directly from patient biopsies have also been successfully generated for several solid malignancies (colon, lung, pancreas) in the hope that they are biologically more predictive than traditional xenografts and genetically engineered mouse models (Ref. Reference Jones, Zhang and Parsons18). In a pancreas-cancer-specific study, resected tumours from 14 PDA patients were xenografted and expanded in successive groups of nude mice to develop cohorts of tumour-bearing mice suitable for drug therapy (PancXenoBank model). These xenografted tumours appear to maintain their original genotype despite serial passages, and recapitulate the genetic heterogeneity of PDA.

A summary of findings from mouse models of PDA is given in Box 1. In conclusion, none of the mouse models completely recapitulates human PDA; however, each model system has its advantages that can be utilised with caution (without over interpretation) to help decipher biological principles that can be evaluated in human clinical trials.

Genomics

Altered gene function in human PDA has been studied in some depth at the DNA (genotype), mRNA (transcriptome), or protein (proteome) level in cells or tissue. Recently, a comprehensive genetic makeup of human PDA was determined by DNA sequencing of 24 patients with advanced disease. The tumours were first passaged in vitro as cell lines or in nude mice as xenografts to remove contaminating non-neoplastic cells, thus facilitating detection of mutations. DNA sequencing of mRNA transcripts (23 219), representing 20 661 protein-coding genes, revealed on average 63 genetic alterations (the majority of which were point mutations), which defined a core set of 12 partially overlapping cellular signalling pathways and processes (Ref. Reference Jones, Zhang and Parsons18) (Table 1). However, the mutated pathway components can vary widely between individual tumours. Each pathway is genetically altered in the majority (67–100%) of PDA samples with a predominance of low-frequency mutations, thus generating tumour heterogeneity. Mapping of these mutations to known three-dimensional protein structures showed that 55 of the 404 mutations were localised to domain interfaces or ligand-binding sites that are likely to interfere with normal function. Aberrant core pathways include: frequently altered genes such as KRAS and regulation of the G1–S cell cycle transition; pathways in which a few altered genes predominate, such as TGF-β signalling; and pathways in which many different genes are altered, such as in integrin signalling, regulation of invasion, homophilic cell adhesion, and small guanine triphosphatase (GTPase)-dependent signalling (Table 1).

Table 1.

Drugs targeting the core signaling pathways in pancreatic ductal adenocarcinoma

Abbreviations: ATM, ataxia telengiectasia mutated; BCL2, B cell lymphoma 2; CDK, cyclin-dependent kinase; CEACAM, carcinoembryonic antigen-related cell adhesion molecule; CHK, checkpoint kinase; FTI, farnesyl transferase inhibitor; GGTI, geranylgeranyl transferase inhibitor; JNK, Janus kinase; mAb, monoclonal antibody; MMP, matrix metalloproteinase; SHH, sonic hedgehog; TGF, transforming growth factor; TGRBR1, transforming growth factor β receptor 1; WNT, wingless; XIAP, X-linked inhibitor of apoptosis.

To analyse the transcriptome of the 24 PDA samples, SAGE (serial analysis of gene expression) (Ref. Reference Velculescu42) was performed in combination with massively parallel sequencing to provide a measure of differential gene expression. There were 541 genes that were tenfold overexpressed in >90% of the 24 PDAs [compared with normal pancreatic duct cells or the HPDE (human pancreatic duct epithelial) cell line]. Of the overexpressed genes, 54 encoded proteins predicted to be secreted or expressed on the cell surface (Ref. Reference Jones, Zhang and Parsons18). The findings indicated that expression changes contribute to dysregulation of the 12 core pathways in PDA pathogenesis.

There are several drawbacks to this analysis: alteration of gene expression when human PDA cells are cultured and/or grown in nude mice; a lack of attention to the prominent tumour–stroma interaction that appears to drive invasion and drug resistance in PDA; and no epigenetic analysis to understand the process of gene silencing in the setting of the 12 core oncogenic pathways. Nevertheless, this landmark paper (Ref. Reference Jones, Zhang and Parsons18) concludes that seeking agents that broadly target downstream mediators or key nodal points are preferable to targeting specific mutated genes. Pathways that appear to represent good targets include metabolic disturbances, neoangiogenesis, misexpression of cell-surface proteins, altered cell cycle proteins, cytoskeletal abnormalities and defective DNA repair. Table 1 highlights potential therapeutic agents that target each of the 12 core overlapping oncogenic pathways. The challenge for the future is to rationally combine these agents to obtain optimal cell kill without major off-target side effects.

Differential gene expression profiles associated with progression of PDA from localised to locally advanced to metastatic disease have also been reported (Ref. Reference Campagna23). This study showed significant gene expression changes (mRNA level) between localised and locally advanced disease (matched or unmatched samples). The genes identified to be overexpressed in locally advanced versus localised disease were pertinent to the mitogen-activated protein kinase (MAPK) and WNT pathway, cellular metabolism, immune regulation and cell–cell or cell–ECM interactions. Similar signalling pathways were identified when localised disease was compared with normal pancreas tissue (Ref. Reference Campagna23). However, this study did not evaluate differential gene expression between locally advanced versus metastatic gene expression profiles and surprisingly could not differentiate between localised and metastatic disease. The reasons for this discrepancy are not given. A second study evaluated differential gene expression profiles in lymph-node-positive versus -negative PDA and identified 194 genes with altered expression (66 overexpressed; 126 downregulated). Genes promoting tumour growth were highly expressed while antiapoptotic genes were downregulated (Ref. Reference Kim43).

Two abstracts presented at the American Society of Clinical Oncology meeting in 2008 (Refs Reference Collisson44, Reference Mori45) assessed prognostic information by gene expression profiling from resected PDA biopsies (with or without microdissection), in patients with long versus short survival. The data suggest that the disparity in outcomes seen after intended curative resection is largely intrinsic to the tumour rather than due to host factors and/or surgical technique and more likely due to tumour–stroma interactions.

The four studies discussed in this section (Refs Reference Jones, Zhang and Parsons18, Reference Campagna23, Reference Collisson44, Reference Mori45) highlight major problems and discrepancies in the study of human PDA by gene expression profiling.

Stem cells

Cancer stem or progenitor cells (CSCs) were initially identified in acute myeloid leukaemia (Ref. Reference Bonnet and Dick46). Subsequently, the first epithelial CSCs were discovered in breast cancer, with a cell-surface marker phenotype of CD44⁺, CD24⁺ and ESA^low; these cells were tumourigenic and formed tumours that recapitulated the patient's tumour in severe combined immunodeficient (scid) mice (Ref. Reference Al-Hajj47). Since then, CSCs have been identified in several solid tumours (brain, colon, head/neck, liver, melanoma and prostate) (Ref. Reference Simeone48). Pancreas-specific CSCs were recently identified in primary pancreas cancer and cell lines on the basis of CD44⁺, CD24⁺ and ESA^low surface markers, and found to constitute 0.2–0.8% of all pancreatic cancer cells (Ref. Reference Li49). A second study isolated CSCs by CD133 expression and found overlap by 14% between these cells and CD44⁺, CD24⁺ and ESA^low cells (Ref. Reference Hermann50). These pancreatic CSCs are highly tumourigenic and possess the ability to self-renew and produce differentiated progeny that reflects the heterogeneity of the patient's primary tumour. Several developmental signalling pathways are upregulated in these CSCs, including hedgehog, BMI1 (polycomb ring finger oncogene) and Notch (Ref. Reference Simeone48). Identification of drugs targeting these and other pathways in CSCs is an active area of research (see below), with Phase I trials imminent.

In general, CSCs are resistant to conventional chemotherapy and radiation except in testicular cancer, childhood leukaemias and good-prognosis Hodgkin lymphoma, where long-term cure rates are high. Treatment of pancreas cancer cells with gemcitabine with or without radiation leads to the enrichment of a CSC population (Refs Reference Li49, Reference Hermann50), providing a mechanism of escape and resistance to conventional therapy. Gene expression profiling and small interfering RNA (siRNA) technology are being used to identify genes important in pancreas CSC self-renewal in ‘tumourspheres’ (Ref. Reference Simeone48). In order to evaluate potential CSC therapeutics, primary pancreas cancer orthotopic xenograft models would be optimal, with or without cotransfer of human pancreatic stromal cells. This approach would preserve tumour heterogeneity and/or tumour–stroma desmoplasia and provide stringent evaluation of drug response (Ref. Reference Simeone48). Novel imaging agents and techniques would be valuable in this setting as would be the discovery of biomarkers of response.

Diagnosis and staging

PDA is a ‘silent’ killer in that symptoms are generally of a nonspecific nature, and at presentation the disease is often advanced. Nonspecific symptoms are fatigue, abdominal pain, heartburn, weight loss and poor appetite; specific signs are jaundice, back pain, pruritis, pale stool and dark urine. Occasionally PDA is discovered accidentally on imaging studies or through increased bilirubin (jaundice), although this is rare with localised disease and more frequent with metastatic disease, especially when the tumour is located in the body and/or tail of the pancreas. Approximately 65% of PDAs arise in the head, and the rest in the body (20%) and tail (15%) of the pancreas. Patients presenting with painless jaundice due to a mass at the head of the pancreas may be fortunate to have resectable disease. Common sites of metastasis are lymph nodes, liver, lung and bone, while rare sites are bone marrow, brain and skin (Ref. 51).

As with any solid tumour, staging is very important in making disease-management decisions and for prognosis in patients with PDA. The tumour, nodal, metastases (TNM) staging system was recently modified by the American Joint Committee on Cancer (AJCC) (Ref. 51). Table 2 describes TNM classification of PDA, with currently assigned treatment modality per staging. In the current AJCC system, patients with PDA are divided into distinct prognostic and clinical groups based on resectability: stage I and II are resectable; stage III (locoregional spread) and stage IV (distant metastasis) are deemed unresectable (Ref. Reference Katz52). For all stages, overall 1-year survival is 23% and 5-year overall survival from diagnosis is 4%. Median survival for patients with locally advanced disease is 9–12 months and, for patients with metastatic disease, the median survival is 3–6 months (Ref. 51). A recent study proposed borderline resectable PDA, a newly emerging stage, with favourable prognosis in a subset of patients treated neoadjuvantly (Ref. Reference Katz53). For patients with tumours confined to the head of the pancreas (2–5 cm) (localised), pancreaticoduodenectomy with lymphadenectomy and venous resection is the optimal surgical procedure (Whipple or modified Whipple) (Ref. Reference Glanemann54).

Table 2.

Staging, treatment strategies and outcome in pancreatic ductal adenocarcinoma

^aComputed tomography findings that preclude surgery are distant metastasis, major venous thrombosis of the portal vein (PV) or superior mesenteric vein (SMV) extending for several centimetres, and circumferential encasement of the superior mesenteric artery, celiac artery or proximal hepatic arteries (Ref. Reference Fuhrman158). Findings indicating borderline resectable patients include encasement of a short segment of the hepatic artery without evidence of tumour extension to the celiac axis, tumour abutment of the superior mesenteric artery (SMA) involving <180° of the artery circumference, and short segment occlusion of the SMV–PV confluence beneath the neck of the pancreas.

Abbreviations: M, (presence or absence of) distant metastases; N, (presence or absence of) regional lymph node metastases; T, (size and anatomic extent of) primary tumour (Tis, carcinoma in situ).

Accurate staging is achieved by noninvasive and invasive imaging methods. Contrast-enhanced computed tomography (CT) accurately predicts resectability in 80–90% of patients (for further details see Table 2 footnote). Magnetic resonance imaging (MRI) is useful in identifying small hepatic metastasis but is not superior to CT. Positron emission tomography (PET) is an evolving technique, as discussed further below. Endoluminal ultrasonography (EUS) is sensitive at detecting small pancreatic tumours, and useful for clarifying suspected vascular invasion, evaluating peripancreatic adenopathy and ascites, and, importantly, obtaining a biopsy prior to neoadjuvant therapy (Ref. Reference Ujiki and Talamonti55). Endoscopic retrograde cholangiopancreatography (ERCP) is useful for stent placement for biliary and/or pancreatic duct obstruction, brush biopsies, detection of mucin production in the PDA, and localising the extent of disease (Ref. Reference Tanaka56). For biopsies, fine-needle aspiration, EUS and ERCP may give insufficient material, and ‘tru-cut’ biopsies may be more useful for evaluation of pancreas pathology (Ref. Reference Levy57).

Chemotherapy

Chemotherapy has been used as palliative therapy for unresectable PDA, and alongside surgery to improve survival following resection. Tumours deemed borderline resectable and/or unresectable (locally advanced) may be resectable after neoadjuvant chemotherapy. These patients will also benefit from postsurgical adjuvant chemotherapy. Patients with unresectable or metastatic disease are treated similarly with standard gemcitabine alone (Ref. Reference Burris13) or gemcitabine plus erlotinib (an inhibitor of epidermal growth factor receptor tyrosine kinase activity) (Ref. Reference Moore58). Other agents that may be used as front-line therapy with gemcitabine are taxols (microtubule destabilisers), including abraxane (paclitaxel), cisplatin or oxaliplatin, and xeloda. These agents may also be used in second-line treatments if patients fail gemcitabine treatment either alone or in combination with erlotinib (Ref. Reference O'Reilly and Abou-Alfa59). Third-line therapies include oxaliplatin plus 5-FU (DNA replication inhibitors) (Ref. Reference Pelzer60) or clinical trials. Thus far, none of the clinical trials of various chemotherapy combinations has resulted in meaningful responses that have significantly improved overall survival. The value of radiation therapy in PDA treatment is discussed in Box 2.

Phase III trials

In patients with advanced PDA, gemcitabine is given as a 30 min infusion. A fixed dose rate (FDR) may improve its clinical activity, but the recent ECOG E6201 study showed no survival advantage to FDR therapy (Ref. Reference Poplin61). Fifteen Phase III clinical trials for advanced PDA patients have been conducted with gemcitabine in combination with other chemotherapeutic agents [exatecan, irinotecan, 5-FU with or without leucovorin (two studies), capecitabine (three studies), cisplatin, oxaliplatin] or targeted therapy (marimistat, tipifarnib, pemetrexed, cetuximab, bevacizumab, erlotinib) (Ref. Reference Cartwright, Richards and Boehm14). The progression-free survival and overall survival range from 3.3–5.7 months and 5.5–9.0 months, respectively, with virtually no added survival advantage for all agents except for erlotinib, where overall survival was 6.24 versus 5.91 months in favour of combination therapy (Ref. Reference Moore58). A recent Phase III study in newly diagnosed metastatic PDA patients found the addition of bevacizumab to gemcitabine plus erlotinib did not significantly increase overall survival, but there was a trend towards improved survival (6.0 versus 7.1 months) and progression-free survival (3.6 versus 4.6 months) (Ref. Reference Vervenne62) compared with gemcitabine plus erlotinib. However, the possible improvements are modest, with expensive drugs and the potential for extra adverse effects. Despite these results, the addition of gemcitabine to capecitabine (7.4–9.0 months) or cisplatin (7.5 months) provides a higher survival than gemcitabine plus erlotinib.

A plethora of nongemcitabine regimens have also been explored without any significant improvement over standard gemcitabine (Refs Reference Von Hoff, Mahadevan and Bearss63, Reference Burtness64). Also, there is no consensus regarding second-line therapy after disease progression on gemcitabine plus another agent. The CONKO-003 Phase III study randomised gemcitabine-resistant PDA patients to 5-FU plus folinic acid (FA) (FF, N = 84) versus oxaliplatin plus FF (OFF, N = 76). The median progression-free survival was 13 weeks versus 9 weeks and overall survival 20 weeks versus 13 weeks for OFF versus FF, respectively (Ref. Reference Pelzer60). Unfortunately, progression-free survival and overall survival are modest for this combination, with likely compromised quality of life. Participation in appropriate targeted therapy trials are encouraged prior to considering second- and third-line therapies for advanced PDA.

Promising Phase II trials

Two new formulations of paclitaxel have given promising results in Phase II trials. The first is nab-paclitaxel (nab-pac/abraxane, Abraxis BioScience), a novel albumin nanoparticle formulation of paclitaxel that is approved by the US Food and Drug Administration (FDA) for the treatment of metastatic breast cancer. It appears to promote preferential delivery of active drug to the site of the tumour, as demonstrated in several preclinical tumour models (Ref. Reference Infante, Matsubayashi and Sato27). In a Phase I study, the combination of nab-pac and gemcitabine in PDA patients was evaluated. Pancreas cancer patients with no prior chemotherapy except a radiation sensitiser received nab-pac (100 mg/m²) plus gemcitabine (1000 mg/m²) on days 1, 8 and 15 every 28 days. Twenty patients were enrolled, with endpoints of toxicity and response. Responses by RECIST (‘response evaluation criteria in solid tumours’) (N = 16) showed partial response in nine patients, stable disease in six patients and progressive disease in one patient. The median cycles to response was 2 (1–5) and no grade 3/4 sensory neuropathy was observed (Ref. Reference Drengler65). The preliminary data suggest that nab-pac plus gemcitabine is very active in patients with advanced PDA, and based on these promising data a randomised Phase III study has been initiated. Moreover, investigation of the expression of SPARC (‘secreted protein and rich in cysteine’ – an ECM protein secreted by peritumoural stromal fibroblasts, which may bind albumin) in this cohort of PDA patients suggested improved response to nab-pac correlates with SPARC overexpression (Ref. Reference Smith66) and appears to be associated with a CA19.9 tumour marker response too.

A second formulation of paclitaxel known as endoTAG-1, developed by MediGene, comprises cationic liposomes that encapsulate taxol and apparently bind preferentially to proliferating endothelial cells and thus interfere with the tumour vasculature. In a Phase II study, 200 advanced PDA patients received three different doses of endoTAG-1 (11, 22, 44 mg/m²) in combination with one weekly gemcitabine dose (1000 mg/m²) for 7 weeks. The 1-year survival with EndoTAG-1 plus gemcitabine nearly doubled compared with gemcitabine alone (33–36% versus 17.5%). The median survival for the combination was 9.4 months (44 mg/m²) and 8.7 months (22 mg/m²) compared with 6.8 months for gemcitabine (Ref. Reference Löhr67). It is very likely that this combination will be studied in a Phase III randomised clinical trial of advanced PDA.

Small-molecule inhibitors

Numerous investigational SMIs have been evaluated in large Phase III clinical trials and have shown no survival benefit alone and/or when combined with gemcitabine. These inhibitors are not discussed here and have been reviewed elsewhere (Ref. Reference Cartwright, Richards and Boehm14). Instead, we review here novel targets and agents that are/or should be evaluated alone and/or in combination with gemcitabine.

Monoclonal antibodies

As with SMIs, mAbs to growth factor ligands and cell-surface receptors targeting PDA, alone and/or in combination with gemcitabine, are currently not approved by the US FDA and are reviewed elsewhere (Ref. Reference Engelhardt95). Here, we review novel mAbs that may be more promising and are being evaluated alone and/or in combination with gemcitabine.

Antisense oligonucleotide therapies

First-generation antisense molecules (typically 20–25 nucleotide RNA) targeting mRNA translation of specific proteins showed some promise, but had problems with pharmacokinetics, which have been overcome with the second-generation molecules (Ref. Reference Stahel and Zangemeister-Wittke103). One target has been ‘inhibitor of apoptosis proteins’ (IAPs), which are present at high levels in many cancers, including PDA. IAPs block apoptosis by binding to and inhibiting effector caspases and have been associated with chemotherapy or radiation resistance; therefore, IAPs are thought to be excellent targets for treatment of PDA and for improving the susceptibility to conventional chemoradiation.

AEG35156, a second-generation antisense oligonucleotide to human XIAP (X-linked inhibitor of apoptosis), enhances cancer cell apoptosis preclinically as a single agent and in combination with chemotherapy (Ref. Reference Jolivet104). In a Phase I study, 29 patients with advanced solid tumours were treated with AEG35156 delivered by 2 h infusions. Twenty-three patients completed at least one 3-week treatment cycle. Treatment was generally well tolerated; the most common toxicities were grade 1 chills/pyrexia, reversible increases in activated plasma thromboplastin time and mild transaminitis. Dosing at 500 mg AEG35156 has been completed with no dose-limiting toxicities and the study is ongoing at 650 mg. Pharmacokinetics were dose-proportional and linear. Decreases in XIAP mRNA were observed in peripheral blood mononucleocytes at dose levels ≥240 mg and increased levels of circulating biomarkers of apoptosis were observed in some patients (Ref. Reference Jolivet104). Based on these promising results, a Phase II study with standard gemcitabine plus escalating doses of AEG35156 for newly diagnosed PDA is ongoing. The real potential of antisense oligonucleotide therapies to deliver clinical benefit is uncertain given disappointing results with antisense therapies to the antiapoptotic BCL2 (e.g. melanoma and chronic lymphocytic leukaemia), and none has been approved by the US FDA as a cancer therapeutic.

Tumour–stroma disruptive therapies

The desmoplastic reaction seen in PDA, which shows a high proportion of stromal cells (Ref. Reference Seymour105), involves a complex interplay between normal epithelial cells, invading ductal tumour cells, activated PSCs, an inflammatory cell infiltrate, proliferating endothelial cells, an altered ECM, and aberrant expression of growth factors that activate several oncogenic signalling pathways by autocrine and paracrine mechanisms. Here we focus on recent therapeutic developments and insights, as a substantial body of knowledge has been previously reviewed (Ref. Reference Mahadevan and Von Hoff24).

Early lesions (PanINs) in genetically engineered mouse models of PDA show infiltration of host immune cells with suppressive properties, and subsequent development of a fibroinflammatory reaction akin to the desmoplastic reaction observed in human PDA (Ref. Reference Clark106). In the KRas (G12D) model, progression from normal to PanIN histology shows progressive infiltration of CD45⁺ cells, with a ~50% leukocyte infiltrate. Flow cytometry and immunohistochemistry showed immunosuppressive cells including myeloid-derived suppressor cells (MDSCs), regulatory T cells (Tregs), and tumour-associated macrophages (TAMs). There appears to be an evolving immune response with undermining of immunosurveillance at inception of tumourigenesis and subsequent worsening in the more invasive phenotypes [KRas (G12D)/Trp53 (R172H)]. Although effector T cells predominate in invasive tumours, they remain inactive, with an associated absence of natural killer cells (Ref. Reference Clark, Beatty and Vonderheide107). Aberrant KRAS signalling elaborates an inflammatory chemokine response [CXCL8, CXCL1 (KC), CXCL2 (MIP-2), CCL2 (MCP-1)] that promotes recruitment of macrophages to the tumour site (Ref. Reference Masamune108). Therapeutic targeting or modulation of defective immunosurveillance in PDA is a viable option; for example, inhibiting FOXP3⁺ Tregs may sensitise/synergise tumours to chemotherapy and/or targeted therapies. These interesting findings need confirmation in human PDA.

PSCs have been isolated from patients with PDA to ascertain their functions with respect to signalling pathways (Ref. Reference Masamune108) and promotion of tumours in mouse model systems (Ref. Reference Hwang109). In the normal pancreas PSCs are quiescent and can be identified by the presence of vitamin-A-containing lipid droplets in the cytoplasm. However, when the pancreas is injured (inflammation), PSCs transform into myofibroblast-like cells that proliferate, produce adhesive proteins [collagen type I, intercellular adhesion molecule 1 (ICAM1)] and inflammatory chemokines [interleukin (IL)-6 (IL6), IL-8, CCL2], induce angiogenesis and promote fibrosis. PSCs also express and secrete fibrinogen, a proclotting factor found in abundance in the ECM of patients with PDA. Isolated PSCs treated with fibrinogen produce IL-6, IL-8, CCL2, vascular endothelial growth factor (VEGF), angiopoietin-1 (ANGPT1), collagen type I and α smooth muscle actin (ACTA2), a marker of activation. Fibrinogen-induced cytokine production can be inhibited by antibodies to α5β1 and αvβ3 integrins expressed on PSCs, suggesting a direct role for these integrins in fibrinogen-induced cytokine production. Fibrinogen treatment of PSCs activates NF-κB, several MAPKs and AKT. SMIs to the respective target proteins inhibited IL-6 and IL-8 production mainly via NF-κB and partially by ERK (MAPK1/3) and p38 MAPK inhibition (Ref. Reference Masamune108). Although fibrinogen had no effect on PSC proliferation or migration, its direct effect on PSCs manifests a fibrogenic and inflammatory process in PDA. Therefore, breaking the interaction between the malignant ductal epithelial cells, infiltrating immunosuppressive cells and PSCs is likely to provide a key target/process that may be manipulated for therapeutic success.

The malignant stroma in PDA is recognised as a key player in promoting drug resistance through activation of PSCs. Human PSCs isolated from PDA do not grow well in cell culture unless immortalised with telomerase and SV40 large T antigen. Conditioned medium (CM) from these immortalised and nonimmortalised human PSCs (HPSCs) promoted cell proliferation (3.0–30-fold), invasion (20–40-fold) and colony formation (1.5–3.0-fold) in BxPC3 (KRAS wild-type) and Panc-1 (KRAS mutated) cells. Treatment of PDA cells with gemcitabine or radiation therapy in the presence and absence of HPSC-CM resulted in approximately a three- to fourfold resistance to apoptosis, likely due to activation of MAPK and AKT pathways. In an orthotopic PDA mouse model, a mixture of HPSCs and tumour cells enhanced tumour progression and metastasis (100%) with an increased ratio of PSCs to tumour cells compared with HSPCs alone (none) or BxPC-3 at 500 000 cells (none). However, at 1 million BxPC-3 cells, 43% of mice developed orthotopic tumours (Ref. Reference Mahadevan and Von Hoff24). A major flaw with this study is utilisation of immortalised PSCs: immortalisation is likely to change the molecular phenotype of these cells and hence affect tumour biology significantly. Also the efficacy and activity of gemcitabine was not evaluated in the orthotopic model (HPSC/BxPC-3) to see how the stroma affected drug resistance. However, despite these short comings the model provides an opportunity to further dissect cytokines and cell-surface receptors that enhance stroma-driven tumour progression.

In addition to involvement in early and late events in PDA, hedgehog signalling is implicated in desmoplasia. A transformed primary ductal-derived epithelial cell line hTert-HPNE (h-Tert/KRas mutant/HPV E6-E7/SV40 small t antigen) used to overexpress SHH ectopically produced larger primary tumours (threefold greater) with a dense desmoplastic reaction versus control in an orthotopic mouse model. Immunohistochemistry studies of human PDA tissue show tumour cells stain positive for SHH while the stromal fibroblasts stain positive for GLI, implicating a paracrine mechanism of activation of stromal fibroblasts and PSCs by the malignant ductal epithelial cells. Furthermore, SHH contributes to the dense desmoplastic reaction by enhancing the motility and invasive capacity of PSCs (Ref. Reference Bailey110). This study used Capan-2 PDA cells, which are known to express SHH and form moderately differentiated tumours, to evaluate an inhibitory antibody to SHH (5E1) in an orthotopically implanted nude mouse model. The mice treated with 500 µg of 5E1 once/week intraperitoneally for 10 weeks showed a decrease in tumour volume from ~450 mm³ to ~100 mm³ compared with a control antibody (4E11), as well as a significant decrease in desmoplasia (Ref. Reference Bailey110). However, this study did not evaluate the efficacy of 5E1 on the SHH-overexpressing hTert-HPNE orthotopic model for TGI and associated inhibition of the desmoplastic reaction. In fact, this cell line is not a true reflection of human PDA given the myriad of introduced artificial tumour-promoting genes of viral origin, making it difficult to draw major conclusions as to the effects of therapy on the desmoplastic reaction. Furthermore, the Capan-2 study did not show adequate data on TGI and a comparison with standard gemcitabine alone and as combination therapy. However, the Capan-2 orthotopic model does provide a useful system to study hedgehog inhibitors in combination with targeted therapies to tumour–stroma crosstalk to inhibit the dense desmoplastic reaction that abrogates effective therapies reaching the tumour.

Imaging

Various imaging approaches are used in PDA diagnosis and management as outlined above. However, these methods have their limitations in distinguishing pancreatic cancer from chronic mass-forming pancreatitis and in differentiating viable tumour from post-therapeutic changes during follow-up (Refs Reference Bipat111, Reference Dewitt112, Reference Farma113, Reference Levy114). CT is the standard imaging modality in diagnosing PDA, but this method has low sensitivity (53%) (Ref. Reference Mertz115) and fails to detect early-stage pancreatic cancer (Refs Reference Chari116, Reference Canto117, Reference Fukushima118). EUS provides high sensitivity (93%) in detecting pancreatic lesions (Ref. Reference Chang119), but cannot differentiate between chronic pancreatitis and PDA (Refs Reference Brand120, Reference Rosch121), and is operator dependent, leading to heterogeneous findings (Refs Reference Mertz115, Reference Meining122, Reference Meining123). A complete discussion about these imaging modalities is beyond the scope of this article. Here we describe novel targeted imaging techniques and agents for detecting PDA at an early stage.

PET is a safe, noninvasive molecular imaging technique used to detect changes in metabolic activity, which is an early sign of disease. PET imaging using [¹⁸F]fluorodeoxyglucose (FDG) is now increasingly being used for diagnosing, staging, and monitoring therapeutic response and recurrence in various cancers. As with glucose, FDG enters the cell and is phosphorylated by hexokinase to FDG-6-phosphate, but further catabolism of FDG-6-phosphate is not possible, as this would require the presence of an oxygen atom on the C-2 position, and FDG-6-phosphate is thus trapped in the cell. Because of these characteristics, FDG can be used to investigate glucose uptake of the cell. The spatial resolution of PET is somewhat limited (10 mm), which restricts the precise anatomic localisation of PET-positive findings in the body. However, with the advent of the PET–CT hybrid, both functional and morphological imaging can be combined into a single test (Ref. Reference Beyer, Townsend and Blodgett124). A study of PET–CT in PDA demonstrated a 91% positive predictive value, 64% negative predictive value, and higher specificity for diagnosis than contrast-enhanced multidetector CT, with a 81% sensitivity and 100% specificity for the detection of distant metastasis to the liver, retroperitoneal lymph nodes, and other sites (Ref. Reference Heinrich125). Since FDG is also taken up rapidly in dividing cells, it is difficult to differentiate malignancy from inflammation. Therefore, several novel PET tracers are being developed to detect PDA at an early stage, differentiate it from inflammatory conditions and evaluate postsurgical recurrence.

Two potential candidates are sigma-receptor ligands and 3′-deoxy-3′-[¹⁸F]fluorothymidine (FLT). Sigma-2 receptors are overexpressed in PDA cell lines and allografts in mice, and the selective sigma-2 receptor ligands SV119 and WC26 have been used with microPET (Ref. Reference Kashiwagi126). Interestingly, these ligands induced apoptosis in mice and human PDA cells in vitro and in vivo, suggesting sigma-2 receptor ligands may be used as potential diagnostic as well as therapeutic agents in PDA (Ref. Reference Kashiwagi126). The thymidine analogue FLT is a new radiopharmaceutical for clinical PET that specifically accumulates in proliferating tissues (Ref. Reference Shields127). In a pilot study involving 31 patients with undefined pancreatic lesions, FLT was taken up exclusively by malignant pancreatic tumours and not by benign pseudotumours (Ref. Reference Herrmann128). This suggests a potential role for FLT as a diagnostic imaging PET agent for differentiating cancer from benign pancreatic lesions. In addition, a novel peptide-based PET tracer targeted to the H⁺/peptide transporter – [¹¹C]glycylsarcosine – has recently been developed (Ref. Reference Mitsuoka129). After intravenous injection of the tracer into athymic mice transplanted with human pancreatic, prostate and gastric cancer cells, all three tumour xenografts were well visualised and there was no accumulation of this tracer in inflammatory sites in comparison to FDG-PET (Ref. Reference Mitsuoka129).

Advances in imaging technology, tagged to those in molecular oncology, have led to several new methods to image molecular interactions in real time – elucidating the intracellular signalling network differences in cancer compared with normal counterparts, angiogenesis, tumour–host responses, tumour–stromal interactions and in vivo gene and protein expression, as well as predicting the pharmacokinetics and flux rates of small molecules and biological agents. Novel technologies in optical imaging include fluorescence reflectance imaging for rapid screening of molecular events on tissue surfaces, tomographic fluorescence imaging (FMI) in combination with structural imaging (FMI-CT, FMI-MRI) in quantifying fluorochrome reporters in three-dimensional space, and bioluminescence imaging for gene and protein expression profiling in vivo. These novel imaging technologies and developments have been recently reviewed (Refs Reference Weissleder and Pittet130, Reference Min and Gambhir131). Although these technologies are still under development, eventually they are likely to be translated into pancreatic cancer management. The use of phage display screening to develop plectin-1 (PLEC1) as a potential new biomarker for PDA illustrates progress towards this goal (Ref. Reference Kelly132). This membrane-localised protein links intermediate filaments to microtubules and microfilaments and also anchors the cytoskeleton to the plasma and nuclear membranes. Conjugation of plectin-1-targeted peptides to magnetofluorescent nanoparticles in conjunction with intravital confocal microscopy and MRI enabled detection of small PDAs and precursor lesions (PanIN) in engineered mouse models.

Biomarkers

Specific and sensitive biomarkers combined with imaging techniques will aid early diagnosis and monitoring of response to therapy. Current biomarkers, including CA19-9 (cancer antigen 19-9), have low sensitivity and specificity in differentiating PDA from benign conditions, such as chronic pancreatitis and benign tumours. In recent years, several possible biomarkers have been identified in the serum, urine and pancreatic juice (Table 3), and validation studies are under way. In biomarker development, biological processes such as immune response to a developing tumour, and hormone/cytokine/growth-factor changes induced by the tumour and stroma, as represented in pancreas juice, serum and urine are analysed by proteomic platforms and mass spectroscopy. Methylation profiling of target genes, comparative genomic hybridisation and gene expression profiling are also useful methods but need to be tailored to obtain a limited number of specific genes that can differentiate normal from preinvasive and overt cancer.

Table 3.

Promising serum biomarkers for pancreatic cancer

Abbreviations: CEACAM1, carcinoembryonic antigen-related cell adhesion molecule 1; EIF4A3, eukaryotic translation initiation factor 4A, isoform 3; GPI, glycosylphosphatidylinositol; MIC-1, macrophage inhibitory cytokine 1 (GDF15); MPF, megakaryocyte-potentiating factor; MUC1/MUC16 mucin 1/16; PGK1, phosphoglycerate kinase 1; TGF-β, transforming growth factor β.

In PDA, autoimmunity has been shown against several tumour-specific cellular proteins, including mucin 1 (MUC1) (Ref. Reference Hamanaka133), p53 (Ref. Reference Raedle134), RAD51 (Ref. Reference Maacke135), and DEAD-box protein 48 (EIF4A3) (Ref. Reference Xia136). Tumour-specific proteins that may show aberrant expression in the serum of patients with PDA include osteopontin (Ref. Reference Koopmann137), macrophage inhibitory cytokine 1 (MIC-1; GDF15) (Ref. Reference Koopmann138), phosphoglycerate kinase 1 (PGK1) (Refs Reference Koopmann138, Reference Koopmann139) and CEACAM1 (Ref. Reference Simeone140). MIC-1 is a member of the TGF-β superfamily, and is overexpressed in many epithelial cancers including PDA. MIC-1 may be a more sensitive marker than CA19-9 in PDA, as 90% of resectable patients had MIC-1 levels >2 standard deviations above controls, while 62% had elevated CA19-9. Furthermore, MIC-1 levels were independent of TNM stage (Ref. Reference Koopmann139). Use of the surface-enhanced laser desorption/ionisation (SELDI) technique to differentiate protein profiles in sera among patients with resectable PDA, normal healthy volunteers and chronic pancreatitis revealed two discriminating SELDI protein peaks, which differentiated patients with PDA from healthy controls with a sensitivity of 78% and specificity of 97% (Ref. Reference Koopmann141). In this study, a combination of three prominent SELDI protein peaks and CA19-9 was superior to CA19-9 alone in distinguishing individuals with PDA from the combined pancreatic disease controls and healthy subject groups. SELDI protein peaks were also better than CA19-9 in differentiating patients with cancer from those with chronic pancreatitis (Ref. Reference Koopmann141). Another serum marker of interest is mesothelin, which was also found to be expressed in PDA tissue (microarrays); detection in suspicious pancreatic lesions may aid in cytopathological diagnosis (Ref. Reference Argani142).

Furthermore, study of proteins initially identified as upregulated in a mouse model of PDA identified five as significantly elevated in sera of PDA patients compared with control groups: neutrophil gelatinase-associated lipocalin (LCN2), lithostathine 1 (REG1A), regenerating islet-derived protein 3 (REG3A; also known as HIP/PAP), tissue inhibitor of metalloproteinase 1 (TIMP1) and insulin-like growth factor binding protein 4 (IGFBP4) (Ref. Reference Faca143). These results were confirmed by studying sera of 26 patients involved in the Carotene and Retinol Efficacy trial (CARE).

Biomarker research on pancreatic juice has focused on genetic (mutant KRAS and TP53) and epigenetic factors such as DNA methylation and mitochondrial DNA mutations (Ref. Reference Goggins144). Mass spectrometric profiling methods have also been used to analyse pancreatic juice biomarkers: one study identified 170 unique proteins in pancreatic cancer by analysing pancreatic juice from three patients, including several known proteins overexpressed in pancreatic cancers (CEACAM5, MUC1, REG3 and LCN2) (Ref. Reference Gronborg145); other studies identified 30 proteins that were at least twofold elevated in pancreatic juice in PDA compared with normal controls (Refs Reference Chen146, Reference Chen147). MALDI-TOF mass spectroscopy identified seven protein peaks that were differentially expressed in pancreatic juice from PDA patients in comparison with normal controls (Ref. Reference Zhou148). Further studies are required to reproduce these results and, if so, validation in a large cohort of PDA patients is needed.

Several groups have attempted to differentiate pancreatic cancers from benign conditions by profiling protein glycosylation patterns via lectin microarrays. In one study, many distinct oligosaccharides were identified from the serum of patients with PDA as compared with serum from normal controls, as well as increased sialylation and fucosylation in haemopexin (HPX), kininogen-1 (KNG1), antithrombin-III (SERPINC1), and haptoglobin-related protein (HPR) but decreased sialylation on plasma protease C1 inhibitor (SERPING1) (Ref. Reference Zhao149). Although it is unclear how these findings could be utilised in patient management, further studies are warranted.

Finally, alterations in microRNAs – small RNAs that regulate mRNA expression – have been analysed in PDA (Ref. Reference Goggins144). Most microRNAs are reduced in cancer but several are overexpressed and thus could be potential targets for early detection (Ref. Reference Lu150).