Galgo: a bi-objective evolutionary meta-heuristic identifies robust transcriptomic classifiers associated with patient outcome across multiple cancer types.

MOTIVATION: Statistical and machine-learning analyses of tumor transcriptomic profiles offer a powerful resource to gain deeper understanding of tumor subtypes and disease prognosis. Currently, prognostic gene-expression signatures do not exist for all cancer types, and most developed to date have been optimized for individual tumor types. In Galgo, we implement a bi-objective optimization approach that prioritizes gene signature cohesiveness and patient survival in parallel, which provides greater power to identify tumor transcriptomic phenotypes strongly associated with patient survival. RESULTS: To compare the predictive power of the signatures obtained by Galgo with previously studied subtyping methods, we used a meta-analytic approach testing a total of 35 large population-based transcriptomic biobanks of four different cancer types. Galgo-generated colorectal and lung adenocarcinoma signatures were stronger predictors of patient survival compared to published molecular classification schemes. One Galgo-generated breast cancer signature outperformed PAM50, AIMS, SCMGENE and IntClust subtyping predictors. In high-grade serous ovarian cancer, Galgo signatures obtained similar predictive power to a consensus classification method. In all cases, Galgo subtypes reflected enrichment of gene sets related to the hallmarks of the disease, which highlights the biological relevance of the partitions found. AVAILABILITY AND IMPLEMENTATION: The open-source R package is available on www.github.com/harpomaxx/galgo. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Heat stress induces epithelial plasticity and cell migration independent of heat shock factor 1.

Current cancer therapies including cytotoxic chemotherapy, radiation and hyperthermic therapy induce acute proteotoxic stress in tumour cells. A major challenge to cancer therapeutic efficacy is the recurrence of therapy-resistant tumours and how to overcome their emergence. The current study examines the concept that tumour cell exposure to acute proteotoxic stress results in the acquisition of a more advanced and aggressive cancer cell phenotype. Specifically, we determined whether heat stress resulted in an epithelial-to-mesenchymal transition (EMT) and/or the enhancement of cell migration, components of an advanced and therapeutically resistant cancer phenotype. We identified that heat stress enhanced cell migration in both the lung A549, and breast MDA-MB-468 human adenocarcinoma cell lines, with A549 cells also undergoing a partial EMT. Moreover, in an in vivo model of thermally ablated liver metastases of the mouse colorectal MoCR cell line, immunohistological analysis of classical EMT markers demonstrated a shift to a more mesenchymal phenotype in the surviving tumour fraction, further demonstrating that thermal stress can induce epithelial plasticity. To identify a mechanism by which thermal stress modulates epithelial plasticity, we examined whether the major transcriptional regulator of the heat shock response, heat shock factor 1 (HSF1), was a required component. Knockdown of HSF1 in the A549 model did not prevent the associated morphological changes or enhanced migratory profile of heat stressed cells. Therefore, this study provides evidence that heat stress significantly impacts upon cancer cell epithelial plasticity and the migratory phenotype independent of HSF1. These findings further our understanding of novel biological downstream effects of heat stress and their potential independence from the classical heat shock pathway.

Multiple combination bactericidal antibiotic testing for patients with cystic fibrosis infected with multiresistant strains of Pseudomonas aeruginosa.

We developed a rapid in vitro antibiotic susceptibility test to screen double- and triple-antibiotic combinations for bactericidal activity against 75 multiresistant Pseudomonas aeruginosa isolates referred from 44 cystic fibrosis (CF) patients. When used alone, the most effective intravenous antibiotic, meropenem, was bactericidal against only 44% of the isolates. High-dose tobramycin (200 microg/ml; concentrations achievable by aerosol administration) was bactericidal against 72% of isolates. Adding a second antibiotic significantly improved bactericidal activity. The most effective double-antibiotic combinations contained high-dose tobramycin plus meropenem, piperacillin/tazobactam, or ciprofloxacin, and were bactericidal against 88 to 94% of the isolates. Excluding high-dose tobramycin, the most effective intravenous double-antibiotic combinations contained meropenem plus ciprofloxacin, tobramycin (4 microg/ml), or cefipime, and were bactericidal against 85%, 71%, and 70% of isolates, respectively. Adding a third antibiotic did not significantly improve inhibition in vitro. We conclude that double-antibiotic combinations containing meropenem or high-dose tobramycin show the best bactericidal activity in vitro against multiresistant strains of P. aeruginosa. Addition of a third antibiotic to these double-antibiotic combinations may be unnecessary.

Differential mechanisms of neutrophil and monocyte adhesion on neuroblastoma cells: CD18 and VLA-4 integrins mediate adhesion to SK-N-SH, but not to SK-N-MC cell line.

We examined the adhesion of monocytes and polymorphonuclear leukocytes (PMNLs) to the neuroblastoma (NB) cell lines SK-N-SH and SK-N-MC, which have some distinct differentiation characteristics. Monocytes adhered to SK-N-SH and SK-N-MC to the same extent (20 +/- 1.4% and 24 +/- 0.8% of monocytes added). Monocyte adhesion to SK-N-SH but not SK-N-MC was partially inhibited by treating monocytes with a mAb to the CD18 (beta2) integrin chain. The adhesion was further inhibited when monocytes were treated with a combination of mAb to CD18 and VLA-4. Treatment of both NB cell lines with interleukin-1alpha (0.5 ng/ml), tumor necrosis factor alpha (100 U/ml), interferon gamma (200 U/ml), or their combinations increased monocyte adhesion to SK-N-SH and SK-N-MC. With each condition, monocyte adhesion to SK-N-SH was partially blocked by mAb to CD18. The inhibition of adhesion to IL-1alpha- or TNFalpha-treated SK-N-SH cells was greater when the monocytes were treated with mAb to both CD18 and VLA-4. In contrast, monocyte adhesion to IL-1alpha or IFNgamma treated SK-N-MC was only slightly inhibited with a combination of mAb to CD18 + VLA-4 and there was no inhibition at all to TNFalpha-treated SK-N-MC. Spontaneous PMNL adhesion to SK-N-SH was almost negligible but increased by treating the cell line with IL-1alpha, TNFalpha, IFNgamma or their combinations. A mAb to CD18 blocked this increase in each case. The pattern of adhesion of PMNLs to SK-N-MC was totally different. PMNL adhesion to unstimulated SK-N-MC was very high (24 +/- 1.3%), was not inhibited by mAb to CD18, and did not increase by stimulating the cell line with IL-1alpha, TNFalpha, IFNgamma or their combinations. Overall, these results suggest two distinct patterns of monocyte and PMNL interaction with neural cells, such as the SK-N-SH and MC cell lines. While monocyte and PMNL adhesion to SK-N-SH is mainly via CD18/VLA-4 or the CD18 mechanisms, respectively, leukocyte adhesion to SK-N-MC is CD18- and VLA-4-independent. Thus, leukocyte-neural cell interactions share some mechanisms common also to leukocyte-endothelium interaction, but there are also unique mechanisms which may be neural cell and differentiation specific.

Chemokine production and adhesion molecule expression by neural cells exposed to IL-1, TNF alpha and interferon gamma.

We investigated the effect of TNF alpha, IL-1alpha and IFN gamma on two neuroblastoma (NB) cell lines (SK-N-SH and SK-N-MC). These lines responded differentially to IL-1alpha, TNF alpha and IFN gamma for MCP-1 and IL-8 production and expression of the ICAM-1 and VCAM-1 adhesion molecules. None of the cytokines induced MCP-1 or IL-8 on SK-N-MC cells. Both chemokines were produced in response to IL-1alpha by SK-N-SH cells, while TNF alpha induced mainly MCP-1 production. Addition of IFN gamma decreased IL-8, but not MCP-1 production. These responses correlated with monocyte and neutrophil chemotactic activity in NB culture supernatants. This activity was neutralized by antibodies to IL-8 and MCP-1. The expression of ICAM-1 on SK-N-MC was up-regulated by TNF alpha or IFN gamma, while IL-1alpha also upregulated ICAM-1 on SK-N-SH cells. VCAM-1 expression on SK-N-SH was induced by IL-1alpha and TNF alpha and IFN gamma synergized with TNF alpha in this respect on both NB cell lines. These results suggest that mechanisms for chemokine production and VCAM-1 and ICAM-1 upregulation by inflammatory cytokines differ and IFN gamma, in conjunction with TNF alpha, stimulate neural cell responses (high MCP-1 and VCAM-1 and decreased IL-8) favouring mononuclear cell recruitment.

Adhesion molecule mechanisms mediating monocyte migration through synovial fibroblast and endothelium barriers: role for CD11/CD18, very late antigen-4 (CD49d/CD29), very late antigen-5 (CD49e/CD29), and vascular cell adhesion molecule-1 (CD106).

Monocytes migrate through vascular endothelium, and then in connective tissue. As a model of this process, we investigated adhesion molecules involved in monocyte migration through HUVEC and a barrier of human synovial fibroblasts (HSF). Minimal spontaneous monocyte migration (6-7%) occurred through either cell barrier, but this increased markedly (27-35% of added monocytes) when a C5a chemotactic gradient was present. Migration across unstimulated HUVEC was partially inhibited (40%) by mAb to CD18 (beta2 integrin) and completely blocked by anti-CD18 plus anti-alpha4 (CD49d; very late Ag-4 (VLA-4)) mAbs. In contrast, migration across HSF induced by C5a or monocyte chemoattractant protein-1 was not inhibited by mAb to CD18 and was only partially inhibited (33%) in combination with anti-alpha4 mAb. The CD18- and VLA-4-independent migration across HSF was completely inhibited by mAb to alpha5 of VLA-5. The inhibitory effect of mAbs to VLA-4 and VLA-5 was on the monocyte and required blockade of CD11/CD18 to be observed. In contrast to HSF, no role for VLA-5 in monocyte transendothelial migration was detected. Both HSF and IL-1-stimulated HUVEC expressed vascular cell adhesion molecule-1 (VCAM-1). However, VLA-4-mediated monocyte migration across HSF was only partially dependent on VCAM-1, in contrast to transendothelial migration, which was completely blocked by anti-VCAM-1 mAbs. In conclusion, unlike transendothelial migration, for which VLA-4 is the alternative mechanism to CD11/CD18 on monocytes, both VLA-4 and VLA-5 can mediate monocyte migration through fibroblast barriers. In addition to VCAM-1, other ligand(s) on HSF are also involved in the VLA-4-mediated migration.

Quantification of studies of blood flow in growing and irradiated tumors in C3H mice.

An approach to quantification of blood flow in growing malignant tumors, by relating the specific rate of change of flow to the specific rate of change of tumor weight or volume, was discussed. This model, previously noted to hold for many animal tumors, offers a good description of blood flow in the KHT sarcoma in C3H mice. The same tumor responds to radiation therapy, and simple expressions were given of the relationship of flow per unit volume, at 2 points in time, to the original radiation dose. Since most animal tumors show a nonviable or necrotic volume, a description of the phenomenon was given for the C3H/Bi mammary carcinoma by noting that the percent of nonviable tissue approaches a limiting (asymptomic)value.

Comparison of growth functions within and between lines of mice selected for large and small body weight.

Several criteria have been suggested for comparing different nonlinear growth functions to determine which function gives the best quantitative description of a given set of observed sigmoid growth curves. These criteria were then used to compare the logistic, Gompertz and Bertalanffy functions within and among lines of mice: a control line (C 1) and lines selected for large (H 6) and small (L 6) body weight at six weeks of age.A general comparison of the three growth functions was based on the differences in residual variances of the respective functions fitted to the growth data of individual mice. Since the three functions differ primarily in the fixed proportion of the asymptotic weight at which the inflexion point occurs, the growth function which will provide the minimum residual variance among the three considered is the one which most closely approximates the observed proportion. The results of this comparison indicated that the logistic function gave the best fit for both sexes of the H 6 and C 1 lines. While no significant differences in residual variances were evident in L 6 males, the Bertalanffy function had the smallest residual variance in L 6 females.The four derived traits of each growth function analyzed individually were the asymptote (A), age at inflexion (t (*)), rate at which a logarithmic function of body weight changes with time (k) and mean absolute growth rate with respect to body weight increase (v). The coefficient of variation among individuals within full-sib families was used to compare the relative variability of the analogous traits estimated from the three growth functions. The coefficients of variation of A, t (*) and k calculated from the logistic function were significantly (P < .01) smaller than those from both the Gompertz and Bertalanffy functions in all three lines, while there were no significant differences in the relative variability of v among the three lines. The genetic and phenotypic correlations between the analogous traits estimated from two different growth functions were sufficiently high in most cases to conclude that the same trait was being measured by the three growth functions. Each derived trait was analyzed for variation in lines, sexes, seasons and respective interactions. The sources of variation generally exhibited similar levels of significance for the analogous traits estimated by the three functions, although a few exceptions were found. These results suggest that although the logistic function provided the best description of the growth data, the same general conclusions about differences within and among the three lines would have been reached with any of the three functions. The four derived traits of the logistic curve were used to describe quantitatively the differences in growth among the H 6, L 6 and C 1 lines.

Rate, composition and efficiency of growth in mice selected for large and small body weight.

Mice selected for high (H6) and low (L6) 6-week body weight and a randombred control population (C1) were characterized for rate, composition and efficiency of growth. Individual body weights were obtained from birth to 8 weeks of age on 682 mice representative of the three lines. Individual whole carcass determinations of water, fat, ash and protein (residual) were obtained for 180 mice sampled weekly from 3-8 weeks of age. Efficiency of feed utilization was estimated from individual body weight and feed consumption data obtained on 189 mice from 3-8 weeks of age. Growth curves for body weight and gain in body weight, constructed by line and sex, showed a temporary retardation of maximum growth rate in the L6 line, which was attributed in part to an extended depression in growth following weaning. The composition of growth yielded no evidence that the more rapid growth rate in the H6 line resulted from an increase in fat deposition relative to the other carcass components. A decrease in fat percent at 7 weeks of age in the H6 and C1 lines was not evident in the L6 line until 8 weeks of age. Females had a higher percentage carcass fat than did males during the 4-7 weeks growth period, but this difference was essentially reduced to zero by 8 weeks of age. Percentage water was highly correlated negatively with percentage fat. Percentages protein and ash were essentially constant across lines and ages. A positive relation between rate and efficiency of growth was observed between lines. Consistent sex differences, males more efficient than females, were observed prior to 6 weeks of age, but were not evident in the later (6-8 week) data.