Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features.

PURPOSE: The purpose of this study was to determine whether computed tomography (CT)-based machine learning of radiomics features could help distinguish autoimmune pancreatitis (AIP) from pancreatic ductal adenocarcinoma (PDAC). MATERIALS AND METHODS: Eighty-nine patients with AIP (65 men, 24 women; mean age, 59.7±13.9 [SD] years; range: 21-83 years) and 93 patients with PDAC (68 men, 25 women; mean age, 60.1±12.3 [SD] years; range: 36-86 years) were retrospectively included. All patients had dedicated dual-phase pancreatic protocol CT between 2004 and 2018. Thin-slice images (0.75/0.5mm thickness/increment) were compared with thick-slices images (3 or 5mm thickness/increment). Pancreatic regions involved by PDAC or AIP (areas of enlargement, altered enhancement, effacement of pancreatic duct) as well as uninvolved parenchyma were segmented as three-dimensional volumes. Four hundred and thirty-one radiomics features were extracted and a random forest was used to distinguish AIP from PDAC. CT data of 60 AIP and 60 PDAC patients were used for training and those of 29 AIP and 33 PDAC independent patients were used for testing. RESULTS: The pancreas was diffusely involved in 37 (37/89; 41.6%) patients with AIP and not diffusely in 52 (52/89; 58.4%) patients. Using machine learning, 95.2% (59/62; 95% confidence interval [CI]: 89.8-100%), 83.9% (52:67; 95% CI: 74.7-93.0%) and 77.4% (48/62; 95% CI: 67.0-87.8%) of the 62 test patients were correctly classified as either having PDAC or AIP with thin-slice venous phase, thin-slice arterial phase, and thick-slice venous phase CT, respectively. Three of the 29 patients with AIP (3/29; 10.3%) were incorrectly classified as having PDAC but all 33 patients with PDAC (33/33; 100%) were correctly classified with thin-slice venous phase with 89.7% sensitivity (26/29; 95% CI: 78.6-100%) and 100% specificity (33/33; 95% CI: 93-100%) for the diagnosis of AIP, 95.2% accuracy (59/62; 95% CI: 89.8-100%) and area under the curve of 0.975 (95% CI: 0.936-1.0). CONCLUSIONS: Radiomic features help differentiate AIP from PDAC with an overall accuracy of 95.2%.

Annotated normal CT data of the abdomen for deep learning: Challenges and strategies for implementation.

PURPOSE: The purpose of this study was to report procedures developed to annotate abdominal computed tomography (CT) images from subjects without pancreatic disease that will be used as the input for deep convolutional neural networks (DNN) for development of deep learning algorithms for automatic recognition of a normal pancreas. MATERIALS AND METHODS: Dual-phase contrast-enhanced volumetric CT acquired from 2005 to 2009 from potential kidney donors were retrospectively assessed. Four trained human annotators manually and sequentially annotated 22 structures in each datasets, then expert radiologists confirmed the annotation. For efficient annotation and data management, a commercial software package that supports three-dimensional segmentation was used. RESULTS: A total of 1150 dual-phase CT datasets from 575 subjects were annotated. There were 229 men and 346 women (mean age: 45±12years; range: 18-79years). The mean intra-observer intra-subject dual-phase CT volume difference of all annotated structures was 4.27mL (7.65%). The deep network prediction for multi-organ segmentation showed high fidelity with 89.4% and 1.29mm in terms of mean Dice similarity coefficients and mean surface distances, respectively. CONCLUSIONS: A reliable data collection/annotation process for abdominal structures was developed. This process can be used to generate large datasets appropriate for deep learning.

Circulating tumor DNA as a potential marker of adjuvant chemotherapy benefit following surgery for localized pancreatic cancer.

BACKGROUND: In early-stage pancreatic cancer, there are currently no biomarkers to guide selection of therapeutic options. This prospective biomarker trial evaluated the feasibility and potential clinical utility of circulating tumor DNA (ctDNA) analysis to inform adjuvant therapy decision making. MATERIALS AND METHODS: Patients considered by the multidisciplinary team to have resectable pancreatic adenocarcinoma were enrolled. Pre- and post-operative samples for ctDNA analysis were collected. PCR-based-SafeSeqS assays were used to identify mutations at codon 12, 13 and 61 of KRAS in the primary pancreatic tumor and to detect ctDNA. Results of ctDNA analysis were correlated with CA19-9, recurrence-free and overall survival (OS). Patient management was per standard of care, blinded to ctDNA data. RESULTS: Of 112 patients consented pre-operatively, 81 (72%) underwent resection. KRAS mutations were identified in 91% (38/42) of available tumor samples. Of available plasma samples (N = 42), KRAS mutated ctDNA was detected in 62% (23/37) pre-operative and 37% (13/35) post-operative cases. At a median follow-up of 38.4 months, ctDNA detection in the pre-operative setting was associated with inferior recurrence-free survival (RFS) [hazard ratio (HR) 4.1; P = 0.002)] and OS (HR 4.1; P = 0.015). Detectable ctDNA following curative intent resection was associated with inferior RFS (HR 5.4; P < 0.0001) and OS (HR 4.0; P = 0.003). Recurrence occurred in 13/13 (100%) patients with detectable ctDNA post-operatively, including in seven that received gemcitabine-based adjuvant chemotherapy. CONCLUSION: ctDNA studies in localized pancreatic cancer are challenging, with a substantial number of patients not able to undergo resection, not having sufficient tumor tissue for analysis or not completing per protocol sample collection. ctDNA analysis, pre- and/or post-surgery, is a promising prognostic marker. Studies of ctDNA guided therapy are justified, including of treatment intensification strategies for patients with detectable ctDNA post-operatively who appear at very high risk of recurrence despite gemcitabine-based adjuvant therapy.

Circulating tumor DNA as an early marker of therapeutic response in patients with metastatic colorectal cancer.

BACKGROUND: Early indicators of treatment response in metastatic colorectal cancer (mCRC) could conceivably be used to optimize treatment. We explored early changes in circulating tumor DNA (ctDNA) levels as a marker of therapeutic efficacy. PATIENTS AND METHODS: This prospective study involved 53 mCRC patients receiving standard first-line chemotherapy. Both ctDNA and CEA were assessed in plasma collected before treatment, 3 days after treatment and before cycle 2. Computed tomography (CT) scans were carried out at baseline and 8-10 weeks and were centrally assessed using RECIST v1.1 criteria. Tumors were sequenced using a panel of 15 genes frequently mutated in mCRC to identify candidate mutations for ctDNA analysis. For each patient, one tumor mutation was selected to assess the presence and the level of ctDNA in plasma samples using a digital genomic assay termed Safe-SeqS. RESULTS: Candidate mutations for ctDNA analysis were identified in 52 (98.1%) of the tumors. These patient-specific candidate tissue mutations were detectable in the cell-free DNA from the plasma of 48 of these 52 patients (concordance 92.3%). Significant reductions in ctDNA (median 5.7-fold; P < 0.001) levels were observed before cycle 2, which correlated with CT responses at 8-10 weeks (odds ratio = 5.25 with a 10-fold ctDNA reduction; P = 0.016). Major reductions (≥10-fold) versus lesser reductions in ctDNA precycle 2 were associated with a trend for increased progression-free survival (median 14.7 versus 8.1 months; HR = 1.87; P = 0.266). CONCLUSIONS: ctDNA is detectable in a high proportion of treatment naïve mCRC patients. Early changes in ctDNA during first-line chemotherapy predict the later radiologic response.

Activation of PI3Kα by physiological effectors and by oncogenic mutations: structural and dynamic effects.

PI3Kα, a heterodimeric lipid kinase, catalyzes the conversion of phosphoinositide-4,5-bisphosphate (PIP2) to phosphoinositide-3,4,5-trisphosphate (PIP3), a lipid that recruits to the plasma membrane proteins that regulate signaling cascades that control key cellular processes such as cell proliferation, carbohydrate metabolism, cell motility, and apoptosis. PI3Kα is composed of two subunits, p110α and p85, that are activated by binding to phosphorylated receptor tyrosine kinases (RTKs) or their substrates. The gene coding for p110α, PIK3CA, has been found to be mutated in a large number of tumors; these mutations result in increased PI3Kα kinase activity. The structure of the complex of p110α with a fragment of p85 containing the nSH2 and the iSH2 domains has provided valuable information about the mechanisms underlying the physiological activation of PI3Kα and its pathological activation by oncogenic mutations. This review discusses information derived from x-ray diffraction and theoretical calculations regarding the structural and dynamic effects of mutations in four highly mutated regions of PI3K p110α, as well as the proposed mechanisms by which these mutations increase kinase activity. During the physiological activation of PI3Kα, the phosphorylated tyrosine of RTKs binds to the nSH2 domain of p85, dislodging an inhibitory interaction between the p85 nSH2 and a loop of the helical domain of p110α. Several of the oncogenic mutations in p110α activate the enzyme by weakening this autoinhibitory interaction. These effects involve structural changes as well as changes in the dynamics of the enzyme. One of the most common p110α mutations, H1047R, activates PI3Kα by a different mechanism: it increases the interaction of the enzyme with the membrane, maximizing the access of the PI3Kα to its substrate PIP2, a membrane lipid.

Knock in of the AKT1 E17K mutation in human breast epithelial cells does not recapitulate oncogenic PIK3CA mutations.

An oncogenic mutation (G49A:E17K) in the AKT1 gene has been described recently in human breast, colon, and ovarian cancers. The low frequency of this mutation and perhaps other selective pressures have prevented the isolation of human cancer cell lines that harbor this mutation thereby limiting functional analysis. Here, we create a physiologic in vitro model to study the effects of this mutation by using somatic cell gene targeting using the nontumorigenic human breast epithelial cell line, MCF10A. Surprisingly, knock in of E17K into the AKT1 gene had minimal phenotypic consequences and importantly, did not recapitulate the biochemical and growth characteristics seen with somatic cell knock in of PIK3CA hotspot mutations. These results suggest that mutations in critical genes within the PI3-kinase (PI3K) pathway are not functionally equivalent, and that other cooperative genetic events may be necessary to achieve oncogenic PI3K pathway activation in cancers that contain the AKT1 E17K mutation.

Germline SMAD4 or BMPR1A mutations and phenotype of juvenile polyposis.

BACKGROUND: Juvenile polyposis (JP) is an inherited condition predisposing to upper gastrointestinal (UGI) polyps and colorectal cancer. Two genes are known to predispose to JP, SMAD4 and bone morphogenetic protein receptor type 1A (BMPR1A). The object of this study was to determine the differences in phenotype of patients with SMAD4 or BMPR1A mutations (MUT+) compared with those without (MUT-). METHODS: DNA was extracted from 54 JP probands and used for polymerase chain reaction of all exons of SMAD4 and BMPR1A. Products were then sequenced and analyzed for mutations. Medical record data were used to create a JP database, and statistical analysis was performed using Fisher's exact and unpaired t-tests. RESULTS: Nine of 54 patients had germline SMAD4 mutations, 13 had BMPR1A mutations, and 32 had neither. There were no significant differences between SMAD4+ and BMPR1A+ cases in terms of clinical factors examined, except for a family history of UGI involvement (P <.01). There was a higher prevalence of familial cases in MUT+ patients (P =.09), >10 lower gastrointestinal polyps (P =.06), and frequency of family history of gastrointestinal cancer compared with MUT- patients (P =.01). CONCLUSIONS: Patients with germline SMAD4 or BMPR1A mutations have a more prominent JP phenotype than those without, and SMAD4 mutations predispose to UGI polyposis.

The role of hPMS1 and hPMS2 in predisposing to colorectal cancer.

Hereditary nonpolyposis colorectal cancer (HNPCC) is attributable to a deficiency of mismatch repair. Inactivation of DNA mismatch repair underlies the genesis of microsatellite instability in colorectal cancer. Germline mutations in three DNA mismatch repair genes, hMSH2, hMLH1, and hMSH6, have been found to segregate in HNPCC and HNPCC-like families. The two DNA mismatch repair genes hPMS1 and hPMS2 have also been suggested to predispose to HNPCC. In this study, 84 HNPCC and HNPCC-like kindreds without known mutations in the other three known DNA mismatch repair genes were screened for germline mutations in the hPMS1 or hPMS2 gene. No clear-cut pathogenic mutations were identified. Conversion technology was used to detect a large hMSH2 deletion in two affected members of the kindred in which the hPMS1 mutation was originally reported, whereas the hPMS1 mutation was only present in one of these two individuals. Since the hPMS1 and hPMS2 genes were first reported, germline mutations in hPMS2 have been demonstrated primarily in patients with Turcot's syndrome. However, no mutation in any of the two genes has been found to segregate in HNPCC families. Until there is better evidence for an increased colorectal cancer risk associated with germline mutations in these genes, a conservative interpretation of the role of mutations in these genes is advised.

Combination bacteriolytic therapy for the treatment of experimental tumors.

Current chemotherapeutic approaches for cancer are in part limited by the inability of drugs to destroy neoplastic cells within poorly vascularized compartments of tumors. We have here systematically assessed anaerobic bacteria for their capacity to grow expansively within avascular compartments of transplanted tumors. Among 26 different strains tested, one (Clostridium novyi) appeared particularly promising. We created a strain of C. novyi devoid of its lethal toxin (C. novyi-NT) and showed that intravenously injected C. novyi-NT spores germinated within the avascular regions of tumors in mice and destroyed surrounding viable tumor cells. When C. novyi-NT spores were administered together with conventional chemotherapeutic drugs, extensive hemorrhagic necrosis of tumors often developed within 24 h, resulting in significant and prolonged antitumor effects. This strategy, called combination bacteriolytic therapy (COBALT), has the potential to add a new dimension to the treatment of cancer.

Phenotypic analysis of hMSH2 mutations in mouse cells carrying human chromosomes.

Conversion of diploidy to haploidy is a method that allows the generation of stable murine/human hybrid cell lines carrying selected human chromosomes in only a single copy. In this setting, it is possible to detect genetic mutations with greater sensitivity and reliability than in diploid cells. Using this method, we were able to identify mutations in the human mismatch repair (MMR) gene hMSH2 in hereditary nonpolyposis colon cancer families, which have escaped detection by the conventional methods. In this report, we show that such hybrid cell lines can also be a valuable tool in the study of the mutated MMR proteins, in particular the variants found in hereditary nonpolyposis colon cancer families that carry missense mutations and where it is unclear whether they predispose to colon cancer. This analysis is made possible by the fact that the human hMSH2 protein is able to complement the MMR defect in the host murine cell line.

Ferredoxin reductase affects p53-dependent, 5-fluorouracil-induced apoptosis in colorectal cancer cells.

Loss of p53 gene function, which occurs in most colon cancer cells, has been shown to abolish the apoptotic response to 5-fluorouracil (5-FU). To identify genes downstream of p53 that might mediate these effects, we assessed global patterns of gene expression following 5-FU treatment of isogenic cells differing only in their p53 status. The gene encoding mitochondrial ferredoxin reductase (protein, FR; gene, FDXR) was one of the few genes significantly induced by p53 after 5-FU treatment. The FR protein was localized to mitochondria and suppressed the growth of colon cancer cells when over-expressed. Targeted disruption of the FDXR gene in human colon cancer cells showed that it was essential for viability, and partial disruption of the gene resulted in decreased sensitivity to 5-FU-induced apoptosis. These data, coupled with the effects of pharmacologic inhibitors of reactive oxygen species, indicate that FR contributes to p53-mediated apoptosis through the generation of oxidative stress in mitochondria.

Secreted and cell surface genes expressed in benign and malignant colorectal tumors.

Serial analysis of gene expression was used to identify transcripts encoding secreted or cell surface proteins that were expressed in benign and malignant tumors of the colorectum. A total of 290,394 tags were analyzed from normal, adenomatous, and cancerous colonic epithelium. Of the 21,343 different transcripts observed, 957 were found to be differentially expressed between normal tissue and adenoma or between normal tissue and cancer. Forty-nine transcripts were elevated > or =20-fold in adenomas, 40 transcripts were elevated > or =20-fold in cancers, and 9 transcripts were elevated > or =20-fold in both. Products of six of these nine transcripts (TGFBI, LYS, RDP, MIC-1, REGA, and DEHL) were predicted to be secreted or to reside on the cell surface, and these were analyzed in more detail. The abnormal expression levels predicted by serial analysis of gene expression were confirmed by quantitative PCR analyses of each of these six genes. Moreover, the cell types responsible for the elevated expression were identified by in situ hybridization and by PCR analyses of epithelial cells immunoaffinity purified from primary tumors. This study extends knowledge of the differences in gene expression that underlie various stages of neoplasia and suggests specific diagnostic approaches that may be useful for the early detection of colorectal neoplasia.

Use of isogenic human cancer cells for high-throughput screening and drug discovery.

Cell-based screening for novel tumor-specific drugs has been compromised by the lack of appropriate control cells. We describe a strategy for drug screening based on isogenic human cancer cell lines in which key tumorigenic genes have been deleted by targeted homologous recombination. As a test case, a yellow fluorescent protein (YFP) expression vector was introduced into the colon cancer cell line DLD-1, and a blue fluorescent protein (BFP) expression vector was introduced into an isogenic derivative in which the mutant K-Ras allele had been deleted. Co-culture of both cell lines allowed facile screening for compounds with selective toxicity toward the mutant Ras genotype. Among 30,000 compounds screened, a novel cytidine nucleoside analog was identified that displayed selective activity in vitro and inhibited tumor xenografts containing mutant Ras. The present data demonstrate a broadly applicable approach for mining therapeutic agents targeted to the specific genetic alterations responsible for cancer development.

A phosphatase associated with metastasis of colorectal cancer.

To gain insights into the molecular basis for metastasis, we compared the global gene expression profile of metastatic colorectal cancer with that of primary cancers, benign colorectal tumors, and normal colorectal epithelium. Among the genes identified, the PRL-3 protein tyrosine phosphatase gene was of particular interest. It was expressed at high levels in each of 18 cancer metastases studied but at lower levels in nonmetastatic tumors and normal colorectal epithelium. In 3 of 12 metastases examined, multiple copies of the PRL-3 gene were found within a small amplicon located at chromosome 8q24.3. These data suggest that the PRL-3 gene is important for colorectal cancer metastasis and provide a new therapeutic target for these intractable lesions.

Cell surface tumor endothelial markers are conserved in mice and humans.

We recently identified genes encoding tumor endothelial markers (TEMs) that displayed elevated expression during tumor angiogenesis. From both biological and clinical points of view, TEMs associated with the cell surface membrane are of particular interest. Accordingly, we have further characterized four such genes, TEM1, TEM5, TEM7, and TEM8, all of which contain putative transmembrane domains. TEM5 appears to be a seven-pass transmembrane receptor, whereas TEM1, TEM7, and TEM8 span the membrane once. We identified mouse counterparts of each of these genes, designated mTEM1, mTEM5, mTEM7, and mTEM8. Examination of these mTEMs in mouse tumors, embryos, and adult tissues demonstrated that three of them (mTEM1, mTEM5, and mTEM8) were abundantly expressed in tumor vessels as well as in the vasculature of the developing embryo. Importantly, expression of these mTEMs in normal adult mouse tissues was either undetectable or detected only in a small fraction of the vessels. These results demonstrate conservation of human and mouse tumor angiogenesis at the molecular level and support the idea that tumor angiogenesis largely reflects normal physiological neovasculaturization. The coordinate expression of TEM1, TEM5, and TEM8 on tumor endothelium in humans and mice makes these genes attractive targets for the development of antiangiogenic therapies.

PUMA induces the rapid apoptosis of colorectal cancer cells.

Through global profiling of genes that were expressed soon after p53 expression, we identified a novel gene termed PUMA (p53 upregulated modulator of apoptosis). The protein encoded by PUMA was found to be exclusively mitochondrial and to bind to Bcl-2 and Bcl-X(L) through a BH3 domain. Exogenous expression of PUMA resulted in an extremely rapid and profound apoptosis that occurred much earlier than that resulting from exogenous expression of p53. Based on its unique expression patterns, p53 dependence, and biochemical properties, PUMA may be a direct mediator of p53-associated apoptosis.