Identification of two clinical hepatocellular carcinoma patient phenotypes from results of standard screening parameters

Previous work has shown that two general processes contribute to hepatocellular cancer (HCC) prognosis: liver damage, monitored by indices such as blood bilirubin, prothrombin time (PT), and aspartate aminostransferase (AST); and tumor biology, monitored by indices such as tumor size, tumor number, presence of portal vein thrombosis (PVT) and blood alpha-fetoprotein (AFP) levels. These processes may affect one another, with prognostically significant interactions between multiple tumor and host parameters. These interactions form a context that provide personalization of the prognostic meaning of these factors for every patient. Thus, a given level of bilirubin or tumor diameter might have a different significance in different personal contexts. We previously applied network phenotyping strategy (NPS) to characterize interactions between liver function indices of Asian HCC patients and recognized two clinical phenotypes, S and L, differing in tumor size and tumor nodule numbers. Our aim was to validate the applicability of the NPS-based HCC S/L classification on an independent European HCC cohort, for which survival information was additionally available. Four sets of peripheral blood parameters, including AFP-platelets, derived from routine blood parameter levels and tumor indices from the ITA.LI.CA database, were analyzed using NPS, a graph-theory-based approach that compares personal patterns of complete relationships between clinical data values to reference patterns with significant association to disease outcomes. Without reference to the actual tumor sizes, patients were classified by NPS into two subgroups with S and L phenotypes. These two phenotypes were recognized using solely the HCC screening test results, consisting of eight common blood parameters, paired by their significant correlations, including an AFP-platelets relationship. These trends were combined with patient age, gender, and self-reported alcoholism into NPS personal patient profiles. We subsequently validated (using actual scan data) that patients in L phenotype group had 1.5× larger mean tumor masses relative to S, P = 6 × 10(-16). Importantly, with the new data, liver test pattern-identified S-phenotype patients had typically 1.7× longer survival compared to L-phenotype patients. NPS integrated the liver, tumor, and basic demographic factors. Cirrhosis-associated thrombocytopenia was typical for smaller S tumors. In L tumor phenotype, typical platelet levels increased with the tumor mass. Hepatic inflammation and tumor factors contributed to more aggressive L tumors, with parenchymal destruction and shorter survival. NPS provides integrative interpretation for HCC behavior, identifying two tumor and survival phenotypes by clinical parameter patterns. The NPS classifier is provided as an Excel tool. The NPS system shows the importance of considering each tumor marker and parameter in the total context of all the other parameters of an individual patient.

Macro- and micro-environmental factors in clinical hepatocellular cancer

We previously developed a network phenotyping strategy (NPS), a graph theory-based transformation of clinical practice data, for recognition of two primary subgroups of hepatocellular cancer (HCC), called S and L, which differed significantly in their tumor masses. In the current study, we have independently validated this result on 641 HCC patients from another continent. We identified the same HCC subgroups with mean tumor masses 9 cm x n (S) and 22 cm x n (L), P<10(-14). The means of survival distribution (not available previously) for this new cohort were also significantly different (S was 12 months, L was 7 months, P<10(-5)). We characterized nine unique reference patterns of interactions between tumor and clinical environment factors, identifying four subtypes for S and five subtypes for L phenotypes, respectively. In L phenotype, all reference patterns were portal vein thrombosis (PVT)-positive, all platelet/alpha fetoprotein (AFP) levels were high, and all were chronic alcohol consumers. L had phenotype landmarks with worst survival. S phenotype interaction patterns were PVT-negative, with low platelet/AFP levels. We demonstrated that tumor-clinical environment interaction patterns explained how a given parameter level can have a different significance within a different overall context. Thus, baseline bilirubin is low in S1 and S4, but high in S2 and S3, yet all are S subtype patterns, with better prognosis than in L. Gender and age, representing macro-environmental factors, and bilirubin, prothrombin time, and AST levels representing micro-environmental factors, had a major impact on subtype characterization. Clinically important HCC phenotypes are therefore represented by complete parameter relationship patterns and cannot be replaced by individual parameter levels.

Phenotypic Categorization and Profiles of Small and Large Hepatocellular Carcinomas

We used a database of 4139 Taiwanese HCC patients to take a new approach (Network Phenotyping Strategy) to HCC subset identification. Individual parameters for liver function tests, complete blood count, portal vein thrombosis, AFP levels and clinical demographics of age, gender, hepatitis or alcohol consumption, were considered within the whole context of complete relationships, being networked with all other parameter levels in the entire cohort. We identified 4 multi-parameter patterns for one tumor phenotype of patients and a separate 5 multi-parameter patterns to characterize another tumor phenotype of patterns. The 2 subgroups were quite different in their clinical profiles. The means of the tumor mass distributions in these phenotype subgroups were significantly different, one being associated with larger (L) and the other with smaller (S) tumor masses. These significant differences were seen systematically throughout the tumor mass distributions. Essential and common clinical components of L-phenotype patterns included simultaneously high blood levels of AFP and platelets plus presence of portal vein thrombosis. S included higher levels of liver inflammatory parameters. The 2 different parameter patterns of L and S subgroups suggest different mechanisms; L, possibly involving tumor-driven processes and S more associated with liver inflammatory processes.

Thrombocytopenia in relation to tumor size in patients with hepatocellular carcinoma

BACKGROUND

Hepatocellular carcinoma (HCC) size at diagnosis is important in management. Without screening programs, tumor size at diagnosis is heterogeneous.

AIMS

To examine the clinical parameters related to tumor size.

METHODS

Using prospectively collected data from a 1,100-patient biopsy-proven HCC cohort presenting in the absence of screening, tumor sizes were ordered and trichotomized and the resulting terciles were compared for tumor and blood parameters.

RESULTS

The terciles were significantly different with respect to portal hypertension and thrombocytopenia, which were present in a higher percent of tercile I patients with smaller tumors. Tercile III patients with larger HCCs had the highest serum α-fetoprotein (AFP), γ-glutamyl transpeptidase (GGTP), and alkaline phosphatase (ALKP) levels and the most portal vein (PV) thrombosis. Subclassification of tercile I patients by AFP showed that patients with high serum AFP had increased numbers of tumor nodules, more PV thrombosis, higher bilirubin, ALKP, and GGTP levels, and shorter survival.

CONCLUSIONS

Smaller-tumor tercile I patients had more advanced portal hypertension with thrombocytopenia than did larger-tumor patients. Tercile I patients with higher AFP levels had more frequent PV thrombosis and worse survival than those with lower AFP levels. Elevated serum GGTP and ALKP levels appear to be associated with a more aggressive HCC phenotype. These differing patterns suggest more than one HCC pathway.

Small hepatocellular carcinomas and thrombocytopenia

BACKGROUND

Clinical phenotypes of small and large hepatocellular carcinomas (HCCs) are not well characterized.

AIM

To evaluate the characteristics of small HCCs diagnosed by screening.

METHOD

A cohort of 430 small HCCs that were diagnosed through screening, were dichotomized according to a size of ≤ 3 cm or >3 cm maximum tumor diameter and compared for radiological and blood-test parameters.

RESULTS

There were 330 males and 100 females. A higher percent of females had smaller tumors. The majority of patients had single tumors, but 15% of those with larger tumors had portal vein thrombosis (PVT) compared to 5% of those with smaller tumors. Significant differences between the tumor-size groups included alpha-fetoprotein (AFP) values and platelet counts, with thrombocytopenia and elevated bilirubin levels being associated with smaller tumors. In comparing PVT-positive and PVT-negative patients, AFP levels and platelet counts were also significantly different between the 2 groups. A mean multinomial multiple logistic regression model was developed for maximum tumor diameter plus PVT.

CONCLUSIONS

The finding of larger tumors being associated with normal platelets and bilirubin levels in comparison to smaller tumors having thrombocytopenia reveals 2 different patterns of HCC presentation.

Abstract 1720: Validation of a 15-gene prognostic signature in stage IB non small-cell lung cancer (NSCLC)

A prognostic 15-gene signature for NSCLC was recently published (J Clin Oncol 28, 4417, 2010). We used a novel integrated approach, the network phenotyping strategy (NPS), to evaluate the performance of this signature in our NSCLC patients and to gain insight into NSCLC heterogeneity by analyzing global mRNA/miRNA and DNA methylation microarray data (n=47 Stage IB tumors). The mRNA data was compared between cases with overall survival (OS) &lt;or&gt; 650 days. The mRNA levels of the 15 individual genes are not associated with OS. We hypothesized that the networked gene expression pattern rather than the absolute level of any 1 gene accounts for the disease outcome association. Maximal spanning tree reduction of a graph of all 105 possible pairwise expression relationships, weighted by the correlation coefficients quantifying co-regulation levels, was used to identify the 7 most informative gene pairs. For each pair, cases were dichotomized into groups where gene 1 was expressed at least 1.5x higher or lower relative to gene 2. A 7-partite graph (K7) captured these trend patterns. K7 cycle-decomposition revealed 10 different gene expression patterns (C1-C10) representing nominally different molecular NSCLC subtypes. Characterization of individual tumor expression trend patterns was computed as the difference vector (≥C) from C1-C10. A rule-based algorithm used ΔC to classify cases into 2 OS groups. We also analyzed DNA methylation and miRNA data for these 15 genes. For 2 genes, hypomethylation (≥&lt;0.1) was associated with high expression; for 3 genes, hypermethylation (≥&gt;0.5) was associated with low expression. For 10 genes with low expression and intermediate methylation, the expression data indicated the involvement of additional regulatory mechanisms. For example, HEXIM1 expression (most commonly down-regulated relative to FAM64A), is regulated by miR-623, which is itself a prognostic biomarker of OS in our study. Validation of the published signature was observed in this independent NSCLC cohort using NPS analysis. Not only was this gene signature prognostic, confirming the signature robustness, but the NPS strategy was equally effective as the originally published principal component analysis, both resulting in a significant separation of the Kaplan-Meier OS curves (p&lt;0.001). In addition, our NPS approach facilitates direct clinical interpretation of NSCLC molecular heterogeneity. The value of integrating genomic microarray data for a published gene expression signature set of 15 genes in predicting Stage IB NSCLC prognosis is illustrated. The NPS approach provided mechanistic insight into the regulation of some of these 15 genes. The significance of HEX1M1 as a member of the most common signature profile is not only detected by the gene expression pattern but is also confirmed by a significant association of a miRNA, miR-623, with OS. Supported by NIH 5P50 CA090440, P30 CA047904, UPMC Institutional Funds.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 1720. doi:1538-7445.AM2012-1720

Abstract 4937: Factors in NSCLC stage and outcome: Integrated genome-wide study

Genome wide data has transformed disease diagnostic and prognostic model development through the identification of molecular profiles. Enhanced molecular insight can be achieved by multiple -omic profile integration. Accordingly, we used a novel multistep integrative unsupervised approach, the network phenotyping strategy (NPS), for discovery of disease diagnostic/prognostic panels in resected NSCLC frozen tumor samples (n=81; 29 Stage IA/52 Stage IB). Whole genome gene expression and DNA methylation data was generated using Illumina BeadChips. In NPS step 1, the most common (&gt;18% of the cohort) hypo- and hyper-methylated genes (lowest/highest β-score; n=31) were identified. In step 2, a subset of these genes was found to reside in networked multiple loci, co-protected against genetic variation by extreme (high n=10 or low n=6) incorporation energy costs, based on an exome entromic analysis. In step 3, maximal spanning tree reduction of a network of all 120 possible expression level relationships for these 16 genes, weighted by the correlation coefficients quantifying co-regulation, was used to identify the most informative sub-network, an 8-partite graph K8. For each partition, cases were dichotomized into groups where gene 1 was expressed at least 1.5x higher or lower relative to gene 2. In step 4, K8 cycle-decomposition revealed only 8 distinct gene expression patterns (C1-C8) representing nominally different molecular NSCLC subtypes. Characterization of individual tumor expression patterns was computed as the patient's difference vector (ΔC) from C1-C8. Stage IA tumor gene expression patterns (overexpression of all 16 genes) most closely matched C1 (true positive rates 0.8 - 0.93 and false positive rates 0.2 - 0.07, ROC 0.9 in 10-fold cross-validation). For stage IB, these 16 gene expression levels were lower and heterogeneous. In step 5, we generated OS prediction models. The ΔCs were used in an alternating decision tree algorithm. Two rule sets were found, one for the high gene expression Stage IA tumors and the second for Stage IB. In a 10-fold cross-validation, the log-rank test for equality of survivor functions resulted in optimal OS separation at 1650 days (p&lt;0.0005). These results demonstrate that: 1) graph theory provides tools for handling complex data relationships without loss of analytic power; and 2) networks of genome loci co-protected against variant incorporation form a filter identifying the function in experimental -omic data. Here, the 16 gene panel we identified includes members of 3 common lung tumorigenesis pathways and/or were co-regulated by miRNAs independently found to be associated with prognosis in our cohort. The integrative NPS facilitated the epigenetic identification of a biomarker gene set, stage classification by overall expression levels and disease outcome prediction by the expression patterns captured by C1-C8. Supported in part by NIH 5P50 CA090440, P30 CA047904, UPMC Institutional Funds.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 4937. doi:1538-7445.AM2012-4937

Abstract 1722: Prognostic markers in early stage non small-cell lung cancer (NSCLC)

Conventional clinical and pathologic risk factors in stage I non small-cell lung cancer (NSCLC) provide limited prognostic information. Novel prognostic biomarkers are needed to identify patients with highest recurrence risk who will receive the greatest absolute risk reduction from adjuvant chemotherapy/radiation. Transcriptome, epigenome, and global genetic variation analyses of resected NSCLC frozen tumor samples (n=81; 29 stage 1A/52 stage 1B) were performed in order to develop an integrated prognostic signature. Only tumors that were resected within 3 months of the first biopsy-proven diagnosis were used to reduce lead time bias in timing of removal of the primary tumor. Here, the association of gene expression with disease free survival (DSF) is described. Whole genome expression data were quantile-normalized using the Illumina GenomeStudio 2011.1 software. Log-transformed data were then processed in Biometric Research Branch (BRB)-Array Tools. Prognostic models were developed with the top ranking genes using the supervised principal component survival algorithm (PCA). Risk group membership was then assigned based on this multivariate model using leave-one-out cross validation and class prediction modeling. Both models adjusted for age, gender and tumor histology. For stage 1A cases, a 113 gene signature selected by fitting a Cox proportional hazard model (Cox) was able to discriminate between poor and good DFS (α=0.001). Upon further analysis using a class prediction algorithm support vector machine (SVM), we identified an 18 gene classifier that predicted outcome SN=0.952, SP=0.571). A 10-fold cross validation ROC curve reported a value of 0.748. For stage 1B cases, the Cox model identified 141 genes (α=0.01), but the SVM prediction algorithm showed reduced SN/SP (0.75/0.25) for a 12 gene classifier, indicative of possible increased heterogeneity among the 1B cases. Nine genes overlapped between the PCA and class prediction model gene signatures. These nine are members of common lung tumorigenesis pathways, i.e. G-protein signaling regulation (KRAS/BRAF/RASSFI); tyrosine protein kinase signaling (EGFR/ERBB2); nucleotide excision repair (ERRC5 and ERCC6) and genes implicated in cell differentiation regulated by miR200B (HGD) or miR338 (TXNRD1), miRNAs independently found to be associated with prognosis in our cohort. These initial results demonstrate that a gene expression profile can distinguish stage I NCSLC tumors and predict prognoses, but require further validation. Supported in part by NIH 5P50 CA090440, P30CA047904, UPMC Institutional Funds.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 1722. doi:1538-7445.AM2012-1722

HCC in older patients

BACKGROUND

Hepatocellular carcinoma (HCC) is a heterogeneous disease, with many poorly-defined prognostic patient subsets. Identification of discrete subsets will aid rational patient and treatment selection.

METHODS

A database with 778 biopsy-proven, unresectable and untransplantable HCC patients who were followed from diagnosis till death was interrogated. Using a moving average algorithm, patients were ordered by survival and then survival cohorts were analyzed according to standard liver function and CT characteristic parameters.

RESULTS

We found characteristic age clustering groupings by survival. In the older age patients, two survival sub-groups were identified, with 45-80 days in one and 330-1,250 days survival in the other group. The longer surviving group had the lowest serum bilirubin and AFP levels and the lowest tumor mass. Remarkably, the trends for both AFP and bilirubin were similar, suggesting that they were not independent variables. This idea was supported by the similar correlation of typical AFP with GGTP, ALKP and SGOT levels.

CONCLUSIONS

A large HCC cohort showed significant age clustering characteristics for survival, especially in older patients. AFP, bilirubin and age were found to be inter-related factors for HCC severity and survival.

Low alpha-fetoprotein hepatocellular carcinoma

BACKGROUND AND AIM

A large proportion of hepatocellular carcinoma (HCC) patients do not secrete elevated levels of the tumor marker alpha-fetoprotein (AFP). There is little published guide to prognostic features of this patient subset.

METHODS

We interrogated a large HCC database in which all patients had been followed until death, to examine which features might be prognostically useful.

RESULTS

We found 413 biopsy-proven unresectable HCC patients with low serum AFP values. Serum gamma glutamyl transpeptidase (GGTP) levels were one of the most significant factors for survival. This dichotomization into low and high GGTP levels separated the patients into distinctive survival ranges. Patients with GGTP levels < 110 U/100 mL and small tumors had longest survival > 795 days. Patients with GGTP > or = 110 U/mL and large tumors with the presence of portal vein thrombosis had the shortest survival range of 300-560 days.

CONCLUSIONS

Serum levels of the onco-fetal protein GGTP represent a useful prognostic parameter in HCC patients with low AFP levels.

Family networks of obesity and type 2 diabetes in rural Appalachia

The prevalence of obesity and diabetes has been studied in adolescent and adult populations in poor, medically underserved rural Appalachia of West Virginia. A web-based questionnaire about obesity and diabetes was obtained in 989 family members of 210 Community Based Clinical Research (CBPR) trained adolescent members of a network of 18 science clubs, incorporating 142 families. After age-correction in < 20 years old, 50% of both adolescents and adults were obese. The frequency distribution of obesity was trimodal. In the overall population 10.4% had type 2 diabetes, while 24% of adult, obese subjects had type 2 diabetes. A new metric, the family diabetes risk potential, identified a trimodal distribution of risk potential. In the lowest most common distribution 43% of families had a diabetic family member. In the intermediate distribution, 69% had a diabetic family member, and in the distribution with highest scores all the families had a diabetic member. In conclusion, the poorest counties of rural Appalachia are at crisis level with the prevalence of obesity and diabetes. The distribution of age-corrected obesity and family diabetes risk potential are not normally distributed. We suggest that targeting individual family units at greatest risk offers the most efficient strategy for ameliorating this epidemic.

Significance of increased serum GGTP levels in HCC patients

A large dataset of patients with biopsy-proven and unresectable HCC who were prospectively followed from diagnosis till death, was examined for the prognostic significance of serum GGTP levels. Their survival-ordered baseline clinical parameter data was examined. Two cohorts of patients with both low AFP and bilirubin levels were identified, who had serum GGTP levels with an increasing trend with increasing AFP levels. The survival of these patients also decreased as their GGTP levels trended up, as did their tumor masses. By contrast, in patients with elevated serum AFP levels, the GGTP trends were less informative, although patients with the highest AFP levels had a worse prognosis if their GGTP levels were simultaneously elevated. The results suggest that following trends in serum GGTP levels, in the context of specific ranges of AFP and bilirubin levels, might be useful in diagnosis of small HCCs.

Network-based analysis of survival for unresectable hepatocellular carcinoma

Primary hepatocellular cancer (HCC) classification systems are based on histopathology and radiology, yet clinical intuition and experience suggest that natural history and disease progression have distinctive clinical features, consistent with a cluster of homogeneous entities within a heterogeneous cohort. We built a rigorous, sequenced, graph-based strategy of network phenotyping analysis (NPA) to combine data smoothing to minimize stochasticity, multivariate analysis to identify ambiguity and prioritize key variables, and k-partite graphs to visualize coherence. In 890 unresectable HCC patients, we selected 13 baseline clinical variables. After rank ordering by survival, we found data structure that exhibited coherence between variables, implying heterogeneity in which survival varied depending on the associated variable profile. The NPA data compression identified five distinctive clinical phenotypes, based only on gender, age, and levels of serum bilirubin, serum alpha-fetoprotein (AFP), and serum gamma glutamyl transpeptidase (GGTP). The phenotypic profiles for gender and age were substantially different. Young, male patients had a low survival, while elderly women had a long survival. Novel clusters included young men, with high AFP levels for their level of bilirubin, as well as women of all ages, with high GGTP values for their level of AFP. The identification of phenotypic groups of HCC may be of value in studies designed to understand their underlying pathophysiology. We conclude that NPA offers a new useful tool in the reclassification of HCC.

HCC in young adults

A large cohort of unresectable and untransplantable biopsy-proven HCC patients was rank-ordered for survival. Non-random clustering by age was noted, with 3 sub-cohorts of younger patients with survival in the range of 90-360 days. One sub-cohort had a predominance of females. Tumor numbers were well monitored by serum AFP, but tumor mass was better monitored by serum GGTP. In contrast to the older patients, the probability of hepatitis appeared to have a major impact on their survival and these patients tended to have larger numbers of smaller tumors, consistent with the idea of a hepatitis-mediated carcinogenic field defect.

Abstract 2010: A Network Phenotyping Strategy (NPS) for HCC identifies distinct coherent survival groupings from observed clinical traits

Background: HCC is a complex disease with tumor growth replacing adjacent liver. Our initial hypothesis was that tumor traits and loss of liver function are independent processes. We examined this, seeking evidence of coherent discrete structure with respect to survival. Methods: NPS was used to analyze 970 patients with biopsy-proven, non-surgical HCC, all followed prospectively until death. We assumed stochasticity with respect to timing of diagnosis and ambiguity of each variable with respect to survival and used a combination of k-partite graph-based analysis with statistical processing of pairwise dependencies of clinical data ordered by survival, each characterized by the mean of clinical parameters of the 100 patients with the closest survival, for 678 survivals. Results: NPS clearly identified heterogeneity, with 3 distinctive phenotypes based on coherence of multiple variables. Each had different survival ranges with trends of collinear relationships between bilirubin, AFP, alkaline phosphatase, SGOT and GGTP. In patients &gt;70 years, the larger the tumor the higher the biochemical values, with the trends decreasing with longer survival. Above 80 years, bilirubin and AFP were normal, with long survival, 330-1250 days. There were 2 groups of younger patients (&lt;55 years). One, with 30-45 days survival, had a similar profile to the older group. In the other, with 90-330 days survival, the profile was for multiple small tumors to trend with biochemical values. A striking finding was the close correlation in trends for AFP against bilirubin, alkaline phosphatase, GGTP or SGOT values, suggesting that liver damage parameters and tumor parameters were not independent factors, but might be inter-twined, especially in older patients. Conclusions: This NPS analysis does not support the hypothesis that tumor growth and cirrhotic damage are independent, but are intertwined processes. Furthermore, we have evidence based on coherence between variables that the growth rates for unifocal tumors of equivalent size are faster in the young than in the elderly. NPS provides useful insights to design future genomic studies to compare different phenotypes and understand the different pathophysiological processes involved.

Note: This abstract was not presented at the AACR 101st Annual Meeting 2010 because the presenter was unable to attend.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 2010.

Abstract 147: DNA methylation patterns are associated with cisplatin and docetaxel response and survival among NSCLC cases

Cisplatin and docetaxel have historically been considered important chemotherapeutic agents with activity against NSCLC, both in the palliative treatment of metastatic (stage IV) disease and combined-modality therapy of stage III disease. However, de novo and acquired drug resistance is common and patients may have only a partial response to therapy. DNA methylation has recently been recognized as a useful marker for diagnosing and monitoring the progression of many types of cancer. CpG island sites frequently become hypermethylated during tumorigenesis, and often exhibit distinctive tumor type and stage related patterns. Here, we examined the DNA methylation profile of 1505 autosomal CpG loci in 807 genes, in relation to response to therapy, time to progression and overall survival in 154 stage III and IV NSCLC cases. The Illumina GoldenGate Methylation Cancer Panel I with DNA isolated from peripheral blood mononuclear cells from patients who were treated with cisplatin and docetaxel. Response to therapy was assessed using RECIST guidelines. Stable disease was noted in 49 cases, 29 cases showed partial response and 54 had progressive disease (22 cases not evaluated). The mean time to progression was 5.2 + 4.5 months (range 3.4-27 months) and mean overall survival was 11.4 + 8.2 months (range 0.1-35 months). Due to assay sensitivity limits, CpG loci with β scores &lt; 0.25 or &gt; 0.75 in at least 75% of the samples were excluded from the analysis. Similarly, in order to screen for differential methylation patterns, we excluded the loci with small β score ranges which were defined as the mean score ± 12.5% for at least 75% of the cases. Following this screening process, 259 loci were included in the association analyses. Using these loci in an unsupervised correlation-based machine learning feature selection algorithm, we further reduced the relevant data pool to 6 significant loci. β scores of these 6 loci were then used in a neural network analysis and predictions significantly correlated with observed time to progression (r=0.74, p&lt;0.0001). Although previous studies have reported an association between the methylation index of specific candidate DNA damage response genes in relation to platinum based chemotherapy response in ovarian cancer, a significant association with the promoter methylation status for BRCA1, MGMT, MLH1, RASSF1 and p16 was not observed among these NSCLC cases in a secondary analysis. These data suggest that methylation status may not only represent a useful biomarker for early detection of NSCLC, but may also aid in the identification of genes differentially regulated that closely correlate with patient survival, which in turn could lead to novel strategies for molecular targeted therapies and patient quality of life. A larger study population with extended follow-up data is required to further validate the observed correlation between DNA methylation and survival. [NIH 5U01CA114771]

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 147.

Feasibility of adolescents to conduct community-based participatory research on obesity and diabetes in rural Appalachia

Community-Based Participatory Research (CBPR) has been advocated to translate advances in health care sciences to the community. We describe a novel approach applied to obesity management and diabetes prevention. This takes advantage of a network of science clubs organized by the Health Sciences and Technology Academy (HSTA) for extracurricular activity of disadvantaged high school students in rural Appalachia. Physician scientists and educators provided an intensive summer course on CBPR, ethics, and study design on obesity management and diabetes prevention. Ethical certification for CBPR investigation was obtained for 210 students and 18 mentors for a study on the prevalence of obesity and Type II diabetes within their community. Over a 6-month period, 989 had a collection of complete analyzable data, of which 103 had diabetes. The proportion with obesity (BMI > or = 30) was over 50%. The frequency of diabetes was related to increasing BMI. When BMI > or = 40, the frequency approached 50%, and exhibited a clear familial distribution. We conclude that trained adolescents can effectively conduct CBPR, and obesity and diabetes are more prevalent than previously reported in this community. This experience provides encouragement to conduct future studies to infl uence weight management from high-risk populations in this medically disadvantaged community.

Prediction of response to alkylator-based chemotherapy in metastatic melanoma (MM) using gene expression and promoter methylation signatures

9009

Background: Temozolomide and dacarbazine (TMZ and DTIC) remain the mainstay of alkylator-based chemotherapy for MM, despite response rates of 10–15% and the absence of any impact on survival. Classification of patients according to responsiveness can guide the individualization of therapy and inform approaches to abrogate mechanisms of chemotherapy resistance. Epigenetic mechanisms play an important role in regulation of genes associated with resistance and were evaluated in tandem with gene expression profiling in biological samples from MM patients (pts) to refine our understanding of the epigenomic-genomic-phenotypic interplay. Methods: We examined promoter methylation and gene expression in tumor tissues of 21 pts with MM treated with TMZ or DTIC, using high throughput technologies (Illumina Inc). The cases were divided into responder (R) and non-responder (NR) groups based on clinical response. The data were analyzed using Prediction Analysis of Microarrays (PAM) from BRB array tools. Results: Differential promoter methylation analysis revealed that 63.6% of promoter sites were hypomethylated in tumors obtained from R pts (p<0.0001). PAM analysis of gene expression data revealed that a classifier set consisting of 82 genes was able to predict NRs from Rs with 83% sensitivity and 89% specificity. Promoter methylation profiling did not independently correlate with R status. A simultaneous analysis of the promoter methylation and gene expression values first stratified into 3 data-driven categories and then combined into a 3 by 3 matrix allowed us to identify a common gene expression/methylation signature of 15 genes that classified both NR and R groups accurately 100% of the time. Conclusions: Gene expression signatures independently predict response to chemotherapy in MM, however promoter methylation profiling alone does not. Analysis of combined gene expression and promoter methylation in a well- annotated clinical data set dichotomized according to response identified a highly predictive signature. The findings from this study are qualified by the relatively small sample size and are currently being validated in an expanded sample set. Supported in part by the ECOG Paul Carbone, MD, Fellowship Award.

No significant financial relationships to disclose.

Tumor and liver determinants of prognosis in unresectable hepatocellular carcinoma: a case cohort study

BACKGROUND AND AIMS

A total of 967 patients with unresectable and untransplantable, biopsy-proven hepatocellular carcinoma (HCC) were prospectively evaluated at baseline and followed up till death.

METHODS

Survival was the end-point for all analyses.

RESULTS

We found in our overall analysis, that male gender, ascites, cirrhosis, portal vein thrombosis (PVT), elevated alpha-fetoprotein (AFP) or bilirubin or alkaline phosphatases were each statistically significant adverse prognostic factors. Patients with normal AFP survived longer than those with elevated AFP, in the presence of PVT, large or bilobar tumors or cirrhosis. We used a bivariate analysis to separate patient subgroups based on poor liver function and aggressive tumor characteristics. In subgroup analysis based on these subsets, there was clear discrimination in survival between subsets; in addition both cirrhosis and presence of PVT were significant, independent but modest risk factors. The results of this large dataset show that amongst nonsurgical HCC patients, there are clear subsets with longer survival than other subsets.

CONCLUSIONS

This data also supports the concept of heterogeneity of HCC.

Patterns of p73 N-terminal isoform expression and p53 status have prognostic value in gynecological cancers

The goal of this study was to determine whether patterns of expression profiles of p73 isoforms and of p53 mutational status are useful combinatorial biomarkers for predicting outcome in a gynecological cancer cohort. This is the first such study using matched tumor/normal tissue pairs from each patient. The median follow-up was over two years. The expression of all 5 N-terminal isoforms (TAp73, DeltaNp73, DeltaN'p73, Ex2p73 and Ex2/3p73) was measured by real-time RT-PCR and p53 status was analyzed by immunohistochemistry. TAp73, DeltaNp73 and DeltaN'p73 were significantly upregulated in tumors. Surprisingly, their range of overexpression was age-dependent, with the highest differences delta (tumor-normal) in the youngest age group. Correction of this age effect was important in further survival correlations. We used all 6 variables (five p73 isoform levels plus p53 status) as input into a principal component analysis with Varimax rotation (VrPCA) to filter out noise from non-disease related individual variability of p73 levels. Rationally selected and individually weighted principal components from each patient were then used to train a support vector machine (SVM) algorithm to predict clinical outcome. This SVM algorithm was able to predict correct outcome in 30 of the 35 patients. We use here a mathematical tool for pattern recognition that has been commonly used in e.g. microarray data mining and apply it for the first time in a prognostic model. We find that PCA/SVM is able to test a clinical hypothesis with robust statistics and show that p73 expression profiles and p53 status are useful prognostic biomarkers that differentiate patients with good vs. poor prognosis with gynecological cancers.

Application of graph-based analysis of genomic sequence context for characterization of drug targets

We show that analyzing individual genes of target proteins in terms of multiplicities of possible realizations of position-dependent thermodynamic states of their DNA molecules constitutes a new bioinformatics paradigm. It provides information that is unique and complementary to results of existing methods of sequence analysis. Using this graph-theory based approach, we developed informative and computationally immensely tractable tool to gain insight into intricate details of properties of drug targets. We present validation of our method by processing seventeen target genes for approved drugs in complexes with known 3D structures. Our novel method can identify coding segments that form important parts of the active and binding sites (individual significance estimated by p-values < or = 0.001). We discuss limitations and advantages of the methodology. Because of its generality, this approach can be used for novel quantitative target-drug assessment and it is applicable to analysis of coding as well as non-coding regions. We also propose the application of this method in quantitative sequence-activity models.

WT p53, but not tumor-derived mutants, bind to Bcl2 via the DNA binding domain and induce mitochondrial permeabilization

The induction of apoptosis by p53 in response to cellular stress is its most conserved function and crucial for p53 tumor suppression. We recently reported that p53 directly induces oligomerization of the BH1,2,3 effector protein Bak, leading to outer mitochondrial membrane permeabilization (OMMP) with release of apoptotic activator proteins. One important mechanism by which p53 achieves OMMP is by forming an inhibitory complex with the anti-apoptotic BclXL protein. In contrast, the p53 complex with the Bcl2 homolog has not been interrogated. Here we have undertaken a detailed characterization of the p53-Bcl2 interaction using structural, biophysical, and mutational analyses. We have identified the p53 DNA binding domain as the binding interface for Bcl2 using solution NMR. The affinity of the p53-Bcl2 complex was determined by surface plasmon resonance analysis (BIAcore) to have a dominant component KD 535 +/- 24 nm. Moreover, in contrast to wild type p53, endogenous missense mutants of p53 are unable to form complexes with endogenous Bcl2 in human cancer cells. Functionally, these mutants are all completely or strongly compromised in mediating OMMP, as measured by cytochrome c release from isolated mitochondria. These data implicate p53-Bcl2 complexes in contributing to the direct mitochondrial p53 pathway of apoptosis and further support the notion that the DNA binding domain of p53 is a dual function domain, mediating both its transactivation function and its direct mitochondrial apoptotic function.

Rational design of DNA sequences for nanotechnology, microarrays and molecular computers using Eulerian graphs

Nucleic acids are molecules of choice for both established and emerging nanoscale technologies. These technologies benefit from large functional densities of 'DNA processing elements' that can be readily manufactured. To achieve the desired functionality, polynucleotide sequences are currently designed by a process that involves tedious and laborious filtering of potential candidates against a series of requirements and parameters. Here, we present a complete novel methodology for the rapid rational design of large sets of DNA sequences. This method allows for the direct implementation of very complex and detailed requirements for the generated sequences, thus avoiding 'brute force' filtering. At the same time, these sequences have narrow distributions of melting temperatures. The molecular part of the design process can be done without computer assistance, using an efficient 'human engineering' approach by drawing a single blueprint graph that represents all generated sequences. Moreover, the method eliminates the necessity for extensive thermodynamic calculations. Melting temperature can be calculated only once (or not at all). In addition, the isostability of the sequences is independent of the selection of a particular set of thermodynamic parameters. Applications are presented for DNA sequence designs for microarrays, universal microarray zip sequences and electron transfer experiments.

Efficient RNA interference depends on global context of the target sequence: quantitative analysis of silencing efficiency using Eulerian graph representation of siRNA

Several aspects of gene silencing by small interfering RNA duplexes (siRNA) influence the efficiency of the silencing. They can be divided into two categories, one covering the cell-specific factors and the other covering molecular factors of the RNA interference (RNAi). A prerequisite for sequence-based siRNA design is that hybridization thermodynamics is the dominant factor. Our assumption is that cell-specific parameters (cell line, degradation, cross-hybridization, target conformation, etc.) can be pooled into an average cellular factor. Our hypothesis is that the molecular basis of the positional dependence of siRNA-induced gene silencing is the uniqueness of context of a corresponding target sequence segment relative to all other such segments along the attacked RNA. We encode this context into descriptors derived from Eulerian graph representation of siRNAs and show that the descriptor based upon the contextual similarity and predicted thermodynamic stability correlates with the experimentally observed silencing efficiency of human lamin A/C gene. We further show that information encoded in this regression function is generalizable and can be used as a predictor of siRNA efficiency in unrelated genes (CD54 and PTEN). In summary, our method represents an evolution of siRNA design from the currently used algorithms which are only qualitative in nature.

p53 has a direct apoptogenic role at the mitochondria

p53 induces apoptosis by target gene regulation and transcription-independent signaling. However, a mechanism for the latter was unknown. We recently reported that a fraction of induced p53 translocates to the mitochondria of apoptosing tumor cells. Targeting p53 to mitochondria is sufficient to launch apoptosis. Here, we provide evidence that p53 translocation to the mitochondria occurs in vivo in irradiated thymocytes. Further, we show that the p53 protein can directly induce permeabilization of the outer mitochondrial membrane by forming complexes with the protective BclXL and Bcl2 proteins, resulting in cytochrome c release. p53 binds to BclXL via its DNA binding domain. We probe the significance of mitochondrial p53 and show that tumor-derived transactivation-deficient mutants of p53 concomitantly lose the ability to interact with BclXL and promote cytochrome c release. This opens the possibility that mutations might represent "double-hits" by abrogating the transcriptional and mitochondrial apoptotic activity of p53.

DNA sequence context and multiplex hybridization reactions: melting studies of heteromorphic duplex DNA complexes

Heteromorphic hybrid duplex DNA complexes are duplex states, other than perfectly matched duplexes, that can form when single strands comprising several different perfectly matched duplexes are simultaneously present in solution. Such cross-hybridization "side reactions" are of particular nuisance in multiplex reaction schemes, where many strands are designed to hybridize in parallel fashion with only their corresponding perfect complement strand. Relative to the perfect match duplexes, the sequence dependent features of these heteromorphic duplex states and their thermodynamic stability are an important consideration for multiplex hybridization reaction design. We have measured absorbance versus temperature melting curves and performed differential scanning calorimetry measurements on various mixtures of eight different 24 base single strands. When perfect complementary pairs of strands are mixed in single reactions, four perfectly matched duplexes form. When mixtures of strands that are not perfectly matched are prepared and analyzed, melting transitions for cross-hybridization are observed along with significant hyperchromicity changes. This is indicative of a melting hybrid, heteromorphic duplex states formed from two nonperfectly matched strands. In addition, when both the perfectly matched and noncomplementary strands are mixed together (in multiplex hybridization reactions) at molar ratios of 1:1, 3:1, and 1:3, evidence of perfect duplex and heteromorphic duplex complexes is found in all cases. A new analytical tool for considering heterogeneous, duplex complexes in multiplex hybridization mixtures is presented and employed to interpret the acquired melting data.

Calculating sequence-dependent melting stability of duplex DNA oligomers and multiplex sequence analysis by graphs

The analytical methods for characterizing DNA sequence-dependent thermodynamic stability have been reviewed. A set of n-n sequence stability parameters is presented. Examples in which these values are used to calculate the thermodynamic stability of short duplex DNA oligomers are presented. The problem of determining sets of isothermal sequences is addressed by representing DNA sequences as graphs. Representing DNA sequences by a graph descriptor with special mathematical properties minimizes the computational difficulty of determining the number of DNA sequences with identical predicted thermodynamic stability. This is achieved by replacement of a whole set of sequences by a single representative. Applications of this concept were demonstrated for sequences assembled from individual bases and sequences assembled from oligomeric blocks.

Thermal unfolding of ribonuclease A in phosphate at neutral pH: deviations from the two-state model

The thermal denaturation of ribonuclease A (RNase A) in the presence of phosphate at neutral pH was studied by differential scanning calorimetry (DSC) and a combination of optical spectroscopic techniques to probe the existence of intermediate states. Fourier transform infrared (FTIR) spectra of the amide I' band and far-uv circular dichroism (CD) spectra were used to monitor changes in the secondary structure. Changes in the tertiary structure were monitored by near-uv CD. Spectral bandshape changes with change in temperature were analyzed using factor analysis. The global unfolding curves obtained from DSC confirmed that structural changes occur in the molecule before the main thermal denaturation transition. The analysis of the far-uv CD and FTIR spectra showed that these lower temperature-induced modifications occur in the secondary structure. No pretransition changes in the tertiary structure (near-uv CD) were observed. The initial changes observed in far-uv CD were attributed to the fraying of the helical segments, which would explain the loss of spectral intensity with almost no modification of spectral bandshape. Separate analyses of different regions of the FTIR amide I' band indicate that, in addition to alpha-helix, part of the pretransitional change also occurs in the beta-strands.

Enhanced prediction accuracy of protein secondary structure using hydrogen exchange Fourier transform infrared spectroscopy

A novel equilibrium hydrogen exchange Fourier transform IR (HX-FTIR) spectroscopy method for predicting secondary structure content was employed using spectra obtained for a training set of 23 globular proteins. The IR bandshape and frequency changes resulting from controlled levels of H-D exchange were observed to be protein-dependent. Their analysis revealed these variations to be partly correlated to secondary structure. For each protein, a set of 6 spectra was measured with a systematic variation of the solvent H-D ratio and was subjected to factor analysis. The most significant component spectra for each protein, representing independent aspects of the spectral response to deuteration, were each subjected to a second factor analysis over the entire training set. Restricted multiple regression (RMR) analysis using the loadings of the principal components from 19 of these H-D analyses revealed an improvement in prediction accuracy compared with conventional bandshape-based analyses of FTIR data. Nearly a factor of 2 reduction in error for prediction of helix fractions was found using s1, the average spectral response for the H-D set. In some cases, significant error reduction for prediction of minor components was found using higher factors. Using the same analytical methods, prediction errors with this new deuteration-response-FTIR method were shown to be even better than those obtained by use of electronic circular dichroism (ECD) data for helix predictions and to be significantly lower for ECD-based sheet prediction, making these the best secondary structure predictions obtained with the RMR method. Tests of a limited variable selection scheme showed further improvements, consistent with previous results of this approach using ECD data.

Novel matrix descriptor for secondary structure segments in proteins: demonstration of predictability from circular dichroism spectra

An extension to standard protein secondary structure predictions using optical spectra that encompasses the number and average lengths of segments of uniform secondary structure in the sequence is demonstrated. The connectivity and numbers of segments can be described by a matrix descriptor [sij] (i, j representing segment types such as helix and beta-sheet strands). Independent knowledge of the fractional concentration of each secondary structure type and of the total number of residues in the protein then with [sij] yields the average segment length of each type. The physical background for prediction of this extended structural descriptor from spectral data is summarized, rules for its generation from reference X-ray structures are defined, and formal variants of its form are discussed. Using a novel neural network approach to analyze a training set of electronic circular dichroism (ECD) and vibrational circular dichroism (VCD) spectra for 23 proteins, matrix descriptors encompassing helix, sheet, and other forms are predicted. The results show that the matrix descriptor can be predicted to an accuracy comparable to that of conventionally predicted average fractional secondary structures. In this respect the ECD predictions of [sij] were significantly more accurate than the VCD ones, which may result from the longer range length dependence of the ECD bandshape and intensity. Summary results for a parallel analysis using Fourier transform infrared spectra indicate somewhat lower reliability than those for VCD.

Predictions of protein secondary structures using factor analysis on Fourier transform infrared spectra: effect of Fourier self-deconvolution of the amide I and amide II bands

Fourier self-deconvolution (FSD) was performed on protein amide I and II Fourier transform infrared (FTIR) spectra to test if the resultant increased band shape variation would lead to improvements in protein secondary structure prediction with our factor analysis based restricted multiple regression (RMR) methods. FTIR spectra of 23 proteins dissolved in H2O were measured and normalized to a constant amide I peak absorbance. The deconvolved spectra were renormalized by area so that the deconvolved spectra sets had the same area as before. Principal component analysis of the deconvolved spectra sets was carried out, which was followed by a selective multiple linear regression (RMR) analysis of the principal component loadings with regard to the fractional components (FC) of secondary structure. As compared to analyses based on the original spectra set, helix and sheet predictions were not noticeably improved by FSD; but, if a very large number of component spectra (16) were retained in the pool to select which loadings to be used in the RMR optimization, better predictions of turn and "other" resulted. The prediction quality varied depending on the deconvolution parameters used.

Characterization of alanine-rich peptides, Ac-(AAKAA)n-GY-NH2 (n = 1-4), using vibrational circular dichroism and Fourier transform infrared. Conformational determination and thermal unfolding

Vibrational circular dichroism (VCD) and Fourier transform IR (FTIR) were measured for a series of short alanine-based peptides having the general formula Ac-(AAKAA)n-GY-NH2 (n = 1-4) from 5 to 50 degrees C in D2O and at room temperature in both TFE and H2O. In both of these latter solvents, the dominant structural form at the lowest temperature for the longest oligomers is alpha-helical. The same is true for the n = 4 peptide in D2O, but under these more dilute aqueous conditions, the shorter (n = 3) peptides have mixed helix-coil structures and the n = 1 and 2 peptides are random coils. The VCD data do not support the 310-helix as a dominant contributor to the conformation of these oligomers in any of these solvents. These vibrational spectral data are consistent with lower-concentration electronic CD results and additionally indicate increased helical stability at higher concentrations. VCD amide I data for the 22mer (n = 4) in D2O indicate that the peptide undergoes a transition from a highly helical conformation at 5 degrees C to a dominant random coil structure at approximately 45 degrees C with a Tm of approximately 25 degrees C (effective midpoint). Factor analysis of the thermal data showed that three principal components were required to describe both the VCD and FTIR data for the n = 4 peptide in D2O. The transition is characterized by a gradual loss of contribution from a spectral component representing the alpha-helical fraction. The third component is evidence of an optically detected intermediate conformation best viewed as a mixed coil-helix structure resulting from end fraying of the helical peptide as the temperature is increased. The nature of the junction between the interior helix and frayed ends is not determined by these data and could involve local (phi and psi) angles mimicking a 310-helix that would provide consistency with ESR and NMR results from Millhauser and co-workers.

Vibrational circular dichroism spectra of proteins in the amide III region: measurement and correlation of bandshape to secondary structure

Vibrational circular dichroism (VCD) spectra have been measured for 23 globular proteins dissolved in H2O/phosphate buffer over the 1400 to 1100 cm(-1) region which encompasses the amide III mode. Spectral responses characteristic of the dominant secondary structure type were found as broad features at approximately 1300 cm(-1), with the extreme forms having positive VCD for highly helical proteins and negative VCD for highly sheet-containing proteins. Quantitative correlation with secondary structure was carried out using previously developed factor analysis and restricted multiple regression (FA/RMR) techniques. Since the absorbance intensity of the amide III mode is difficult to determine due to overlap with other transitions, an alternative, absolute intensity-independent, simple structural analysis method was used. A linear regression was developed between the fractional components of secondary structure for the protein set and the overlap integrals of the normalized spectra from the set with that of a selected protein. The results of this simple method are quite comparable to those of the FA/ RMR approach for analysis with amide III VCD. On the other hand, test calculations with the new method when used with electronic CD spectra are not as good as FA/RMR due to its more intensity-dependent relationship with secondary structure.

Secondary structures comparison of aquaporin-1 and bacteriorhodopsin: a Fourier transform infrared spectroscopy study of two-dimensional membrane crystals

Aquaporins are integral membrane proteins found in diverse animal and plant tissues that mediate the permeability of plasma membranes to water molecules. Projection maps of two-dimensional crystals of aquaporin-1 (AQP1) reconstituted in lipid membranes suggested the presence of six to eight transmembrane helices in the protein. However, data from other sequence and spectroscopic analyses indicate that this protein may adopt a porin-like beta-barrel fold. In this paper, we use Fourier transform infrared spectroscopy to characterize the secondary structure of highly purified native and proteolyzed AQP1 reconstituted in membrane crystalline arrays and compare it to bacteriorhodopsin. For this analysis the fractional secondary structure contents have been determined by using several different algorithms. In addition, a neural network-based evaluation of the Fourier transform infrared spectra in terms of numbers of secondary structure segments and their interconnections [sij] has been performed. The following conclusions were reached: 1) AQP1 is a highly helical protein (42-48% alpha-helix) with little or no beta-sheet content. 2) The alpha-helices have a transmembrane orientation, but are more tilted (21 degrees or 27 degrees, depending on the considered refractive index) than the bacteriorhodopsin helices. 3) The helices in AQP1 undergo limited hydrogen/deuterium exchange and thus are not readily accessible to solvent. Our data support the AQP1 structural model derived from sequence prediction and epitope insertion experiments: AQP1 is a protein with at least six closely associated alpha-helices that span the lipid membrane.

Protein structural segments and their interconnections derived from optical spectra. Thermal unfolding of ribonuclease T1 as an example

A novel descriptor for protein structure is examined here that goes beyond predictions of the average fractional components (FC) of a few conformational types and represents the number and interconnection of segments of continuous, well-defined secondary structural elements such as alpha-helices and beta-sheets. This matrix descriptor can be predicted from optical spectra using neural network methods. The new matrix plus traditional FC descriptors can be quickly and generally obtained to provide a level of detail not previously derived from optical spectra and a discrimination between proteins that might otherwise be viewed as being very similar using just the FC descriptor. As an example of its potential utilization, this matrix descriptor approach was applied to an analysis of both the native state and the reversible thermal denaturation of ribonuclease T1 in H2O. Analyses of the FTIR spectral data indicate initial loss of the major helical segment at 50-55 degrees C but with little accompanying change in the number of sheet segments or the sheet FC values. Circular dichroism (CD) and vibrational CD data are also used to support this interpretation based on FC changes with temperature. Parallel analysis of the corresponding data for this protein in D2O demonstrates that the method is sensitive to the match between the degree of H-D exchange used to prepare samples for the unknown and the reference data set.

Predictions of secondary structure using statistical analyses of electronic and vibrational circular dichroism and Fourier transform infrared spectra of proteins in H2O

Vibrational circular dichroism (VCD) and Fourier transform IR (FTIR) methods for prediction of protein secondary structure are systematically compared using selective regression analysis. VCD and FTIR spectra over the amide I and II bands of 23 proteins dissolved in H2O were analyzed using the principal component method of factor analysis (PC/FA) and regression fits to fractional components (FC) of secondary structure. Predictive capability was determined by computing structures for proteins sequentially left out of the regression. All possible combinations of PC/FA spectral parameters (coefficients) were used to form a full set of restricted multiple regressions (RMR) of PC/FA coefficients with FC values, both independently for each spectral data set as well as for the VCD and FTIR sets grouped together and with similarly obtained electronic CD (ECD) data. The distribution of predictive error for a set of the best RMR relationships that use a given number of spectral coefficients was used to select the optimal prediction algorithm. Minimum predictive error resulted for a small subset (three to six) of spectral coefficients, which is consistent with our earlier findings using VCD measured for proteins in 2H2O and ECD data. Subtracting the average absorption spectrum from all the training set FTIR spectra before analysis yields more variance in the FTIR band shape and improves the predictive ability of the best PC/FA RMR to near that for the VCD. Both methods (FTIR and VCD) using data for proteins in H2O are somewhat better predictors than amide I' (in 2H2O) VCD alone and, for helix, worse than ECD alone. Combining FTIR and VCD data did not dramatically change the prediction results. Predictions are improved by combining both with ECD data, indicating that the improvement is due to using their very different structural sensitivities. The coupled H2O-based spectral analyses and the mixed amide I' + II VCD plus ECD analysis are comparable for the helix and sheet components, indicating that partial deuteration is not a major source of prediction error.

Comparison of and limits of accuracy for statistical analyses of vibrational and electronic circular dichroism spectra in terms of correlations to and predictions of protein secondary structure

This work provides a systematic comparison of vibrational CD (VCD) and electronic CD (ECD) methods for spectral prediction of secondary structure. The VCD and ECD data are simplified to a small set of spectral parameters using the principal component method of factor analysis (PC/FA). Regression fits of these parameters are made to the X-ray-determined fractional components (FC) of secondary structure. Predictive capability is determined by computing structures for proteins sequentially left out of the regression. All possible combinations of PC/FA spectral parameters (coefficients) were used to form a full set of restricted multiple regressions with the FC values, both independently for each spectral data set as well as for the two VCD sets and all the data grouped together. The complete search over all possible combinations of spectral parameters for different types of spectral data is a new feature of this study, and the focus on prediction is the strength of this approach. The PC/FA method was found to be stable in detail to expansion of the training set. Coupling amide II to amide I' parameters reduced the standard deviations of the VCD regression relationships, and combining VCD and ECD data led to the best fits. Prediction results had a minimum error when dependent on relatively few spectral coefficients. Such a limited dependence on spectral variation is the key finding of this work, which has ramifications for previous studies as well as suggests future directions for spectral analysis of structure. The best ECD prediction for helix and sheet uses only one parameter, the coefficient of the first subspectrum. With VCD, the best predictions sample coefficients of both the amide I' and II bands, but error is optimized using only a few coefficients. In this respect, ECD is more accurate than VCD for alpha-helix, and the combined VCD (amide I' + II) predicts the beta-sheet component better than does ECD. Combining VCD and ECD data sets yields exceptionally good predictions by utilizing the strengths of each. However, the residual error, its distribution, and, most importantly, the lack of dependence of the method on many of the significant components derived from the spectra leads to the conclusion that the heterogeneity of protein structure is a fundamental limitation to the use of such spectral analysis methods. The underutilization of these data for prediction of secondary structure suggests spectral data could predict a more detailed descriptor.

Detection and characterization of triple-helical pyrimidine-purine-pyrimidine nucleic acids with vibrational circular dichroism

Vibrational circular dichroism (VCD) spectra were measured in the C = O stretching region for poly(U)*poly(A).poly(U), poly(dT)*poly(dA).poly(dT), and poly(U)*poly(dA). poly(dT). These VCD spectra of the triple-helical structure were dramatically different from those of the corresponding duplexes. The VCD indicates that a very similar base-pair structure is present in these triplexes. The same sign pattern was found for poly(C+)*poly(I).poly(C), which implies a generality of structure than can result from the steric constraint of the triple helix conformation. By contrast, the corresponding duplexes are quite different in terms of their VCD. The transitions between triplex, duplex, and single-stranded forms were studied as a function of temperature and interpreted using factor analysis. The relative stabilities of the triplexes lie in the order RNA > DNA > hybrid. Nondegenerate dipole-coupling calculations for a U*A.U oligomer were carried out for the C = O stretching modes to model the spectral changes observed. The experimental absorbance spectra indicate that the bases have nonequivalent H-bonds which can be achieved if a reverse Hoogsteen base-pairing scheme is assumed. The computational VCD results with such a scheme were in better qualitative agreement with experiment than those using the expected Hoogsteen base-pairing scheme.

The dynamic behavior of annexin V as a function of calcium ion binding: a circular dichroism, UV absorption, and steady-state and time-resolved fluorescence study

The binding of calcium ions to annexin V in the absence of phospholipids has been studied by UV-difference spectroscopy, circular dichroism, and steady-state and time-resolved fluorescence. In the absence of calcium, the unique tryptophan 187, located in domain III of annexin V, is surrounded by a strongly hydrophobic environment, as indicated by its "blue" fluorescence emission maximum (325 nm). This corresponds well with the description of the structure determined by X-ray crystallography of several crystal forms. The Trp187 time-resolved fluorescence decay shows the existence of a fast (picosecond) excited-state reaction which can involve the formation of an H-bond between the indole NH group and the proximate epsilon-OH and/or alpha-carbonyl groups of Thr224. Titration with calcium tends to stabilize the overall structure, as shown by circular dichroism, while leading to large modifications of the local structure around Trp187 making it accessible to the solvent as shown by UV-difference spectra, circular dichroism spectra, and the displacement of its fluorescence emission maximum at saturating concentrations of calcium (350 nm). A rapid (picosecond) formation of an excited-state complex, probably involving one or a few water molecules of the solvation shell, is observed. These observations correlate well with the conformational change observed in crystal structures obtained in high calcium concentrations, involving the removal of Trp187 from the buried position to the surface of the molecule [Sopkova, J., Renouard, M., & Lewit-Bentley, A. (1993) J. Mol. Biol. 234, 816-825; Concha, N. O., Head, J. F., Kaetzel, M. A., Dedman, J. R., & Seaton, B. A. (1993) Science 261, 1321-1324]. In the solvent-exposed conformation, the indole ring becomes mobile in the subnanosecond and nanosecond time range. This conformational change and the increase in local flexibility can be important for the accommodation of the protein on the surface of phospholipid membranes.

Quantitative analysis of vibrational circular dichroism spectra of proteins. Problems and perspectives

Experimental and computational aspects of the quantitative analysis of vibrational circular dichroism (VCD) of proteins are discussed. Experimentally, the effect of spectral resolution, sample concentration, cell selection and spectral normalization effects are considered. The influence of random intensity variations on the results of quantitative analysis of amide I' VCD are shown to be minor up to a 15% variation in spectral intensity. A computational algorithm, based on factor analysis of the spectra and multiple linear regression calculation of fractions of secondary structures (FC), was designed to analyse quantitatively the details of the VCD spectra-structure relationship. It also enabled the results of VCD measured independently for the amide I' and amide II regions to be combined. Our study is based primarily on the optimization of the calculation to predict FC values for proteins not included in the reference data set used for regression. The best prediction is obtained with the function using only part of the observable independent VCD spectral components. Inclusion of all components actually reduces the prediction accuracy of the analysis. Spectroscopic reasons for such behaviour and the consequences of the interdependence of the crystallographic FC values on the spectra-structure analysis are discussed. Finally, the possibility of utilizing VCD spectra to obtain quantitative structural information about the protein beyond the conventional secondary structure composition is explored. A matrix descriptor of super-secondary structure features for proteins is designed, and preliminary results for prediction of this descriptor from amide I' VCD spectra are presented. These latter calculations use a novel design of the back-propagation neural network.