Prognostic value of deep learning-derived body composition in advanced pancreatic cancer-a retrospective multicenter study

BACKGROUND

Despite the prognostic relevance of cachexia in pancreatic cancer, individual body composition has not been routinely integrated into treatment planning. In this multicenter study, we investigated the prognostic value of sarcopenia and myosteatosis automatically extracted from routine computed tomography (CT) scans of patients with advanced pancreatic ductal adenocarcinoma (PDAC).

PATIENTS AND METHODS

We retrospectively analyzed clinical imaging data of 601 patients from three German cancer centers. We applied a deep learning approach to assess sarcopenia by the abdominal muscle-to-bone ratio (MBR) and myosteatosis by the ratio of abdominal inter- and intramuscular fat to muscle volume. In the pooled cohort, univariable and multivariable analyses were carried out to analyze the association between body composition markers and overall survival (OS). We analyzed the relationship between body composition markers and laboratory values during the first year of therapy in a subgroup using linear regression analysis adjusted for age, sex, and American Joint Committee on Cancer (AJCC) stage.

RESULTS

Deep learning-derived MBR [hazard ratio (HR) 0.60, 95% confidence interval (CI) 0.47-0.77, P < 0.005] and myosteatosis (HR 3.73, 95% CI 1.66-8.39, P < 0.005) were significantly associated with OS in univariable analysis. In multivariable analysis, MBR (P = 0.019) and myosteatosis (P = 0.02) were associated with OS independent of age, sex, and AJCC stage. In a subgroup, MBR and myosteatosis were associated with albumin and C-reactive protein levels after initiation of therapy. Additionally, MBR was also associated with hemoglobin and total protein levels.

CONCLUSIONS

Our work demonstrates that deep learning can be applied across cancer centers to automatically assess sarcopenia and myosteatosis from routine CT scans. We highlight the prognostic role of our proposed markers and show a strong relationship with protein levels, inflammation, and anemia. In clinical practice, automated body composition analysis holds the potential to further personalize cancer treatment.

Apparent Diffusion Coefficient (ADC) predicts therapy response in pancreatic ductal adenocarcinoma

Recent advances in molecular subtyping of Pancreatic Ductal Adenocarcinoma (PDAC) support individualization of therapeutic strategies in this most aggressive disease. With the emergence of various novel therapeutic strategies and neoadjuvant approaches in this quickly deteriorating disease, robust approaches for fast evaluation of therapy response are urgently needed. To this aim, we designed a preclinical imaging-guided therapy trial where genetically engineered mice harboring endogenous aggressive PDAC were treated with the MEK targeting drug refametinib, which induces rapid and profound tumor regression in this model system. Multi-parametric non-invasive imaging was used for therapy response monitoring. A significant increase in the Diffusion-Weighted Magnetic Resonance Imaging derived Apparent Diffusion Coefficient (ADC) was noted already 24 hours after treatment onset. Histopathological analyses showed increased apoptosis and matrix remodeling at this time point. Our findings suggest the ADC parameter as an early predictor of therapy response in PDAC.

Structure, chromosomal localization, and expression of the gene for mouse ecto-mono(ADP-ribosyl)transferase ART5

Mono(ADP-ribosyl)transferases regulate the function of target proteins by attaching ADP-ribose to specific amino acid residues in their target proteins. The purpose of this study was to determine the structure, chromosomal localization, and expression profile of the gene for mouse ecto-ADP-ribosyltransferase ART5. Southern blot analyses indicate that Art5 is a single copy gene which maps to mouse chromosome 7 at offset 49.6 cM in close proximity to the Art1, Art2a and Art2b genes. Northern blot and RT-PCR analyses demonstrate prominent expression of Art5 in testis, and lower levels in cardiac and skeletal muscle. Sequence analyses reveal that the Art5 gene encompasses six exons spanning 8 kb of genomic DNA. The 5' end of the Art5 gene overlaps with that of the Art1 gene. A single long exon encodes the predicted ART5 catalytic domain. Separate exons encode the N-terminal leader peptide and a hydrophilic C-terminal extension. Sequencing of RT-PCR products and ESTs identified six splice variants. The deduced amino acid sequence of ART5 shows 87% sequence identity to its orthologue from the human, and 37 and 32% identity to its murine paralogues ART1 and ART2. Unlike ART1 and ART2, ART5 lacks a glycosylphosphatidylinositol-anchor signal sequence and is predicted to be a secretory enzyme. This prediction was confirmed by transfecting an Art5 cDNA expression construct into Sf9 insect cells. The secreted epitope-tagged ART5 protein resembled rat ART2 in exhibiting potent NAD-glycohydrolase activity. This study provides important experimental tools to further elucidate the function of ART5.

Molecular characterization and expression of the gene for mouse NAD+:arginine ecto-mono(ADP-ribosyl)transferase, Art1

Mono(ADP-ribosyl)transferases regulate the function of target proteins by attaching ADP-ribose to specific amino acid residues in the proteins. We have characterized the gene for mouse arginine-specific ADP-ribosyltransferase, Art1. Southern blot analyses indicate that Art1 is a single-copy gene. Northern blot and reverse transcription-PCR analyses demonstrate prominent expression of Art1 in cardiac and skeletal muscle, and lower levels in spleen, lung, liver and fetal tissues. While human ART1 is not represented in the public expressed sequence tag (EST) database, the database contains 14 mouse Art1 ESTs. The Art1 gene encompasses four exons spanning 20 kb of genomic DNA. The deduced amino acid sequence of Art1 exhibits the characteristic features of a glycosylphosphatidylinositol-anchored membrane protein. It shows 75-77% sequence identity with its orthologues from the human and rabbit, and 33-34% identity with its paralogues from the mouse, Art2-1 and Art2-2. Separate exons encode the N- and C-terminal signal peptides, and a single long exon encodes the entire predicted native polypeptide chain. We expressed Art1 in 293T cells as a recombinant fusion protein with the Fc portion of human IgG1. This soluble protein exhibits enzyme activities characteristic of arginine-specific ADP-ribosyltransferases. The availability of the Art1 gene provides the basis for applying transgene and knockout technologies to further probe the function of this gene product.

Two novel human members of an emerging mammalian gene family related to mono-ADP-ribosylating bacterial toxins

Mono-ADP-ribosylation is one of the posttranslational protein modifications regulating cellular metabolism, e.g., nitrogen fixation, in prokaryotes. Several bacterial toxins mono-ADP-ribosylate and inactivate specific proteins in their animal hosts. Recently, two mammalian GPI-anchored cell surface enzymes with similar activities were cloned (designated ART1 and ART2). We have now identified six related expressed sequence tags (ESTs) in the public database and cloned the two novel human genes from which these are derived (designated ART3 and ART4). The deduced amino acid sequences of the predicted gene products show 28% sequence identity to one another and 32-41% identity vs the muscle and T cell enzymes. They contain signal peptide sequences characteristic of GPI anchorage. Southern Zoo blot analyses suggest the presence of related genes in other mammalian species. By PCR screening of somatic cell hybrids and by in situ hybridization, we have mapped the two genes to human chromosomes 4p14-p15.1 and 12q13.2-q13.3. Northern blot analyses show that these genes are specifically expressed in testis and spleen, respectively. Comparison of genomic and cDNA sequences reveals a conserved exon/intron structure, with an unusually large exon encoding the predicted mature membrane proteins. Secondary structure prediction analyses indicate conserved motifs and amino acid residues consistent with a common ancestry of this emerging mammalian enzyme family and bacterial mono(ADP-ribosyl)transferases. It is possible that the four human gene family members identified so far represent the "tip of an iceberg," i.e., a larger family of enzymes that influences the function of target proteins via mono-ADP-ribosylation.

Use of the EST database resource to identify and clone novel mono(ADP-ribosyl)transferase gene family members

We searched the database of expressed sequence tags (dbEST) for relatives of the known human and murine mono(ADP-ribosyl)transferases (mADPRT), poly(ADP-ribosyl)polymerases (PARP), ADP-ribosyl cyclases, and ADP-ribosylarginine hydrolases (ARH). By May 31, 1996, all of the known enzymes except for RT6 were represented in dbEST by exact sequence matches from mouse and/or human tissues. Several ESTs show significant sequence similarity but not identity to known mADPRTs. We isolated, cloned, and sequenced the corresponding genes. Our results show that seven human ESTs stem from a novel gene, provisionally designated LART, which is specifically expressed in lymphatic tissues. Five human ESTs stem from a novel gene, here designated TART1, which is specifically expressed in testis. This gene is also represented by a single mouse EST. One other mouse EST stems from a distinct gene, here designated TART2, which is also expressed in testis. These genes have similar exon/intron structures. The predicted LART and TART1 gene products contain hydrophobic N- and C-terminal signal peptides characteristic for GPI-anchored surface proteins, TART2 lacks the GPI-anchor signal peptide. The predicted native proteins show 28-42% sequence identity to one another. They each contain four cysteine residues that probably form conserved disulfide bonds. They each also contain a conserved glutamic acid residue within the proposed active site motif LART and TART1 show interesting deviations from the surrounding consensus sequence.