SEARCH

SEARCH BY CITATION

Keywords:

  • lung cancer;
  • ABCB1;
  • ABCC2;
  • ABCG2;
  • survival;
  • pharmacogenetics

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

ATP-binding cassette (ABC) transporter expression and genetic heterogeneity have been implicated in response to anticancer therapy. This study characterized genetic variability of the ABCB1 (also known as MDR1), ABCC2 (MRP2) and ABCG2 (BCRP) genes, which are key players in the metabolism of many chemotherapeutic agents including those used in the treatment of lung cancer. We genotyped 53 polymorphisms in the candidate genes in genomic DNA samples of 171 cases of small cell lung carcinoma (SCLC) and 206 cases of non-small cell lung carcinoma (NSCLC), and studied their impact on early response to chemotherapy, progression-free survival and overall survival. SNP rs717620 in ABCC2 was moderately associated with a poor response to chemotherapy but strongly with shorter progression-free survival and overall survival in SCLC but not NSCLC patients, indicating that ABCC2 genetic variation is an important factor in SCLC survival after chemotherapy.

Despite various improvements in the treatment of lung cancer, its overall prognosis remains poor. Five-year survival rates are in the range of 5% to 10% (World cancer report 2008).

In general, lung cancer patients can be treated by surgery, radiation, chemotherapy or a combination thereof. Non-small cell lung carcinoma (NSCLC) patients with a localized tumor are mostly treated surgically, while patients with non-localized NSCLC are primarily treated with radio-chemotherapy. Small cell lung carcinoma (SCLC) is mostly treated with chemotherapy, with or without additional radiotherapy. In Europe, between 1980 and 1997, the percentage of patients receiving chemotherapy has increased, especially among patients with later stage disease.1 Therefore, it would be desirable to find markers that are able to predict response to treatment and outcome of the disease for targeted therapy.

Even in patients who are comparable regarding histology, tumor stage and age there is great variation in the response to chemotherapy,2 which is consistent with genetic background as a determinant of drug response. Genetic factors have been shown to play a role in drug disposition and effects.3–9 In particular, polymorphisms in genes involved in metabolism and transport of chemotherapeutic drugs, including ABC transporters, are likely to affect response to therapy. Expression of the ATP-binding cassette (ABC) multidrug transporters has been implicated in tumor cell resistance to anticancer therapy, altered clearance of chemotherapy drugs, and associated toxicity. Genetic heterogeneity has been described in a number of the ABC transporter genes, including ABC transporters that contribute to the pharmacokinetics and/or pharmacodynamics of chemotherapy drugs.7, 10, 11 The members of the ABC transporter family actively transport various substances out of cells. Some members of this family like ABCB1 (also known as multidrug resistance-1, MDR1), ABCG2 (breast cancer resistance protein, BCRP) or ABCC2 (multidrug resistance-associated protein-2, MRP2) are well known for their role in multidrug resistance.12

In normal lung, ABCB1 is expressed on the apical surface of bronchial and bronchiolar epithelium and at the plasma membrane of alveolar macrophages, where it may function to remove external compounds from the lung lumen.13 In lung cancer, ABCB1 expression is initially low, but this may change after exposure to chemotherapy as part of acquired drug resistance.14 ABCB1 confers resistance to cytotoxic drugs, including etoposide and cisplatinum15 and polymorphisms can affect substrate specificity.16 Many reports suggest that polymorphisms in ABCB1 significantly influence the therapeutic response in lung cancer, although these findings are not always concordant.4

ABCG2 transports a variety of chemically unrelated compounds, including drugs such as doxorubicin, daunorubicin, etoposide or methotrexate and its polyglutamates.17 Tumoral ABCG2 protein expression might be relevant for response of advanced NSCLC patients treated with a platinum-based therapy.8 In cell lines, the polymorphic variant C421A, which results in glutamine replacing a lysine at position 141 (rs2231142), is associated with differential protein activity and drug sensitivity.18, 19 Another ABCG2 single nucleotide polymorphism (SNP), G34A (rs2231137) was shown to result in disturbed localization of the ABCG2 transporter in the plasma membrane and decreased drug efflux.20 However, Yanase et al.19 reported the drug resistance profile of the 34A variant to be similar to the wild type.

ABCC2 is expressed at low levels in normal lung tissue and at high levels in lung cancer.21 Interestingly, ABCC2 has a higher expression in tumors of higher histological grade and poor differentiation compared to well-differentiated tumors.21 This suggests that these tumors may quickly develop resistance to anti-cancer agents. In tumor cell lines ABCC2 mRNA overexpression was associated with resistance to etoposide, vincristine, cisplatin, doxorubicin and epirubicin.15, 22–24 Moreover, ABCC2 effluxes a wider range of drugs than ABCC1, including cisplatin,25 the drug most widely employed in the treatment of lung cancer patients. Allelic variants of the ABCC2 gene have also previously been found to be associated with toxicity induced by chemotherapy agents26 and lung cancer survival.10, 27 Moreover, the ABCC2 SNP-24CT (rs717620), located in the promoter, has been reported to lower the expression of the protein.28–30

Using a tagging approach, in our study we have taken into account all the common genetic variability of the ABCB1, ABCC2 and ABCG2 genes. We analyzed the relation of 53 SNPs in these three genes with three different endpoints, namely early response to therapy, progression-free survival and overall survival, separately for SCLC and NSCLC in patients treated with chemotherapy. Given that combination treatment is a standard in lung cancer chemotherapy, different treatment groups were also considered. This study represents, to the best of our knowledge, the most comprehensive analysis to date of these three key genes polymorphisms and lung cancer chemotherapy.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Study subjects

A total of 377 patients (206 with NSCLC and 171 with SCLC) of Caucasian origin with histologically confirmed primary lung cancer eligible for first-line chemotherapy were recruited between March 1999 and February 2007 at the Thoraxklinik (university hospital for the treatment of thoracic diseases) in Heidelberg, Germany. Patients had never received antineoplastic chemotherapy before nor had they previously been diagnosed with other malignancies. Tumor stage at the time of diagnosis was determined retrospectively by one of two experienced chest radiologists according to the cTNM classification of the Union Internationale contre le Cancer (UICC)31 for all NSCLC patients and for a subset (24%) of SCLC patients. The remaining SCLC cases were classified using the Veterans Administration Lung Cancer Study Group (VALG) criteria for limited and extensive disease which then were considered as equivalent with UICC Stages 2–3 and 4, respectively.5

All patients included in this study received first line chemotherapy, NSCLC patients were treated with gemcitabine and/or platinum-based drugs and/or other drugs, whereas SCLC patients received etoposide and/or platinum-based drugs and/or other drugs (detailed in Table 1).

Table 1. Main clinical characteristics for the NSCLC and SCLC patient groups
inline image

Tumor response was assessed after the second cycle of first line chemotherapy as complete remission (CR), partial remission (PR), stable disease (SD) or progressive disease (PD) according to the RECIST (1.0) criteria for solid tumors.32

Progression-free survival (PFS) was defined as the time interval between start of chemotherapy until documented progression (uncensored observation) or until the last progression-free examination (censored observation) irrespective of whether that patient was lost to follow up or whether death occurred later. OS was defined as the time interval between start of chemotherapy and death (uncensored observation) or the last date when the patient was still alive (censored observation).

All participants gave their informed consent. The study was approved by the ethical committee of the Medical Faculty of the University of Heidelberg (Nr. 201/98 and 199/2001). All blood samples used in this study had been collected prior to start of chemotherapy. The study subjects were asked to complete a self-administered questionnaire with detailed information on pre-treatment, smoking habits and possible occupational exposure.

Selection of tagging SNPs

We surveyed the entire set of common genetic variants in ABCB1, ABCG2 and ABCC2. All polymorphisms in the candidate gene regions (including 5 kb upstream of the first exon and 5 kb downstream of the last exon of each gene), with minor allele frequency (MAF) ≥5% in Caucasians from the International HapMap Project (version 22; http://www.hapmap.org), were included. Tagging SNPs were selected with the use of the Tagger program within Haploview (http://www.broad.mit.edu/mpg/haploview/; http://www.broad.mit.edu/mpg/tagger/),33, 34 using pairwise tagging with a minimum r2 of 0.8.

This resulted in a selection of 16 tagging SNPs for ABCG2, with a mean r2 of the selected SNPs with the SNPs they tag of 0.963, 25 tagging SNPs for ABCB1, with a mean r2 of 0.956, and 12 tagging SNPs for ABCC2, with a mean r2 of 0.981. Our selection therefore captures to a very high degree the known common variability in these genes. The genes and the linkage disquilibrium (LD) blocks which they include are shown in Figure 1. Considering that the genomic regions of the three genes are characterized by high levels of LD, we postulate that such SNPs are also likely to tag any hitherto unidentified common SNPs in the gene.

thumbnail image

Figure 1. Tagging SNPs selected for investigation. (a) ABCB1 gene and its tagging SNPs. From top to bottom, schematic representation of the gene (boxes represent exons, lines represent introns), position of the SNPs in the region, LD structure of the gene region (color intensity is proportional to r2, whereas numbers represent D′; squares where numbers are not reported have D′ = 100%). (b) ABCC2 gene and its tagging SNPs. (c) ABCG2 gene and its tagging SNPs. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

DNA extraction

Buffy coat from 5 ml venous blood in EDTA was archived at −80°C. Genomic DNA (gDNA) was isolated using either the QIAamp DNA blood midi kit (Qiagen, Hilden, Germany), or by an automated DNA extraction protocol on the MagNA Pure LC System (Roche Diagnostics - Applied Science, Mannheim, Germany) according to the manufacturer's instructions.

Genotyping

Genotyping was carried out using the Taqman assay. The MGB Taqman probes were synthesized by Applied Biosystems (Applied Biosystems, Foster City, CA). The reaction mix included 10 ng gDNA, 10 pmol for each primer, 2 pmol for each probe and 5 μl of 2x master mix (Applied Biosystems), in a final volume of 5 μl. Thermocycling involved 40 cycles with 30 s at 95°C followed by 60 s at 60°C. PCR plates were read on an ABI PRISM 7900HT instrument (Applied Biosystems). Genotype discrimination was performed using SDS software (Applied Biosystems), version 2.2. All samples that did not give a reliable result in the first round of genotyping were resubmitted for up to two additional rounds of genotyping. Data points that still remained unfilled after this procedure were left blank.

All genotyping was conducted by personnel blinded to sample identity. Eight percent of genotypes were repeated for quality control, and yielded a concordance rate of 100%.

Haplotype reconstruction

Haplotype blocks were identified from the genotyping data using SNPtool (http://www.dkfz.de/de/molgen_epidemiology/tools/SNPtool.html)35 and the Haploview v4.2 software. A MAF > 0.05, HWE p > 0.001 and a call rate >75% were used as cut-off values. Individual haplotypes were then statistically inferred using the PHASE v.2.1.1 algorithm, based on a Bayesian approach (http://www.stat.Washington.edu/stephens/).36

Statistical analysis

Allele frequencies of all SNPs were calculated and their genotype distributions were tested for deviations from Hardy-Weinberg Equilibrium (HWE) using the chi-square test.

Univariate analysis exhibited tumor stage as statistically significant factor of influence on response in NSCLC patients only (p = 0.01) and for all lung cancer patients (p = 0.0007), but not in SCLC (p = 0.3). It was also a statistically significant factor for the PFS and the OS endpoints in all patients (p < 0.0004). Gender had no significant influence. Detailed results of univariate analyses are showed in Supporting Information Table 1.

Response to chemotherapy was assessed using odds ratios (OR) with 95% confidence intervals (CI) obtained from multivariable logistic regression, by comparing genotype frequencies in responders (CR and PR) and non-responders (SD and PD). Hazard ratio estimates (HR) with 95% CI were calculated using Cox proportional hazard models, both for PFS and OS. All analyses were adjusted for tumor stage, gender and age.

PFS and OS were evaluated using methods for censored survival time, in particular, the Kaplan–Meier estimate for graphical presentation and the logrank test for statistical testing.

For all analyses the homozygotes for the common allele were considered as reference group. For haplotype analysis the most frequent was set as the reference, whereas haplotypes with a frequency below 0.05 were declared as rare haplotypes and combined.

We performed also exploratory analyses of four therapy-based groups: 157 NSCLC patients receiving gemcitabine, 167 SCLC patients receiving etoposide, 127 SCLC and 160 NSCLC patients treated with platinum-based therapy.

All calculations were done using the statistical software package SAS (SAS Institute, Cary, NC) version 9.1.3. Because of the large number of SNPs analyzed, we applied a conservative threshold for statistical significance, based on a revised version of the Bonferroni method where we calculated for each gene the number of effective independent variables (alleles), Meff, by use of the SNP Spectral Decomposition approach37, 38 (http://gump.qimr.edu.au/general/daleN/SNPSpDsuperlite/).

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Patient characteristics and treatment information are given in Table 1 separately for the histological sub-populations. Three different endpoints were analyzed in this study: response after the second cycle of treatment, progression-free survival (PFS), and overall survival (OS).

We analyzed all the common genetic variability of the three genes of interest (53 SNPs tagging 210 common SNPs in the three candidate gene regions). The distribution of the genotypes of all the SNPs analyzed is given in Supporting Information Tables 2–4. For all the SNPs the distribution of genotypes was found to be within HWE and there were no significant differences in the allele frequencies when comparing SCLC with NSCLC patients.

In Table 2, we showed the results of the three SNPs most significantly associated with the three considered endpoints: rs717620 of ABCC2, rs6979885 which is situated in the 1 intron of ABCB1, and rs3109823 that belongs to the 11 intron of ABCG2.

Table 2. Statistically significant associations between SNPs in transporter genes and prognostic parameters
inline image

We analyzed subjects according to six different groups: G1 = all SCLC patients, G3 = SCLC patients who received platinum-based drugs, G5 = SCLC who received etoposide. G3 and G5 are partially overlapping subsets of G1. G2 = all NSCLC patients, G4 = NSCLC patients who received platinum-based drugs, G6=NSCLC patients who received gemcitabine. G4 and G6 are partially overlapping subsets of G2. The results are shown in Table 2.

Carriers of the T allele of ABCC2 rs717620 had a worse response after the second cycle (p = 0.0467 for C/T vs. C/C in group G1); shorter PFS (p = 0.0015 for C/T vs. C/C in group G3) and also shorter OS (p = 0.0023 for C/T+T/T vs. C/C in group G3 as well as in group G5 p = 0.0098 for T/T vs. C/C). Carriers of the A allele of ABCB1 rs6979885 showed a better PFS (p = 0.0047 for A/G+A/A vs. GG in group G1) and OS (p = 0.0049 for A/G+A/A vs. GG in group G1). Finally, individuals carrying the C allele of ABCG2 rs3109823 showed a better response to therapy (p = 0.0287 for C/T+C/C vs. T/T in group G3), a better PFS (p = 0.0322 for C/T+C/C vs. T/T in group G3) and also prolonged OS (p = 0.0062 for C/T+C/C vs. T/T in group G3).

For NSCLC, we did not observe any statistically significant associations between the three genotypes and the three clinical outcomes. All significant associations between ABCB1, ABCC2 and ABCG2 SNPs and the three clinical endpoints (overall survival, progression-free survival and response after 2nd cycle of chemotherapy) and the distribution of the genotypes of all the SNPs analyzed are presented in Supporting Information Tables 2–4. Haplotype analysis did not reveal any statistically significant associations at study-wide level in any of the different outcome or subgroups. The frequencies of the haplotypes and the results of the analysis are reported in Supporting Information Tables 5–16.

To take into account the problem of multiple testing, we calculated for each gene the number of effective independent variables (Meff). The Meff values for ABCB1, ABCC2 and ABCG2 are 16, 8 and 10, respectively. The experiment-wise significance threshold is therefore 0.05/(16+8+10) = 0.0015. Using this threshold, only one association remained significant, namely that between SNP rs717620 and PFS, calculated with Cox regression. Figure 2 shows the Kaplan–Meier curve for rs717620 and PFS in group G3. Median survival for patients carrying the minor allele (T) was 247 days, compared to 329 days for homozygotes for the common allele (C) (log-rank p = 0.0156).

thumbnail image

Figure 2. Kaplan–Meier curve showing progression-free survival for the SCLC patient subgroup receiving platinum-based drugs in relation to ABCC2 genotype at rs717620.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Currently, for lung cancer the chemotherapy regime is predominantly chosen on the basis of the histological type, tumor stage and the performance status of the patient. Genotyping of DNA from peripheral blood could also provide valuable information to tailor therapy and improve efficacy of treatment. For example, drug transporters are important in determining drug disposition.11 Therefore, we investigated all the common genetic variability of the ABCB1, ABCC2 and ABCG2 in relation to clinical parameters of response to therapy and survival of SCLC and NSCLC patients. In our study we found several associations between polymorphisms in ABCs genes and the three considered endpoints in lung cancer treatment. Because of the relatively small sample size and the large number of SNPs studied, we chose to be conservative from a statistical point of view in order to avoid false positive findings, and applied a strict multiple testing correction as described in the methods of this work. Only the association between rs717620 (situated in the promoter of ABCC2) and PFS appeared to be statistically significant. Although it reached statistical significance (after multiple testing correction) only for PFS, the association between the carriers of the T allele and a worse response to the therapy was consistent (at a level of significance of p < 0.05) for all the three endpoints in SCLC patients, which makes this finding unlikely to be a false positive.

Given that most of lung cancer patients enrolled in this study were treated with combination therapy, according to the current clinical practice, within the two histology groups we considered treatment subgroups. The subgroup where we observed the strongest association was G3, which includes SCLC patients who received platinum-based drugs. This finding is intriguing, because human ABCC2 cDNA was originally isolated from a cisplatin-resistant human cancer cell line, thus suggesting that it can provide resistance to cisplatin.24

This polymorphism is thought to be associated with altered gene expression, with posttranscriptional reduction of protein expression for the haplotypes containing the T variant.28–30 Many studies have been performed on this polymorphism to elucidate its possible functional role although none of them has been carried out on lung tissue or in lung cancer tissue.28–30 We can only hypothesize that the polymorphism could alter the protein expression or function in SCLC, and that this could lead to a poor prognosis.

It is interesting to note that in a previous report Han and collaborators found that this polymorphism had favorable impact on PFS in Korean NSCLC patients (stage IIIB or IV) treated with irinotecan and cisplatin.10 In our study, with almost twice the sample size, we observed a similar effect for this polymorphism in our NSCLC patients (treated mostly with gemcitabine and platinum-based drugs) with an OR of 0.391 (CI 95% 0.157–0.970; p value = 0.0429). Thus, our findings reinforce the ones from Han et al. relating to NSCLC. It remains to be determined whether the polymorphism affects the therapy in NSCLC and in SCLC in opposite ways. Such a difference may either be therapy- or tumor-related. These results are not completely unexpected since SCLC and NSCLC are different entities from various points of view, for example they do not share various molecular genetic abnormalities: SCLC has a higher p53 mutation rate,39 while NSCLC harbors mutations in the RAS gene40 and often presents over-expression of COX-2.41 Moreover, while loss of cell cycle controls is a hallmark of cancers, the mechanisms by which the two major type of lung cancer reach this status are distinct.42–45 Therefore they are, from a molecular point of view, very different. Clinically the two entities differ in aggressiveness and treatment response. They are therefore treated with different drugs. For example etoposide, which is a substrate for ABCC2, is used in the first-line treatment of SCLC but not NSCLC patients. However the effects observed in our study are stronger in the G3 (platinum-treated) group than in G5 (etoposide treated) a drug (platinum) which both group of patients is treated with. The different effect visible in SCLC and in NSCLC reinforces the idea that the two cancer types are, from a genetic point of view, two different entities and therefore explains, at least in part, why also genetic prognostic markers could be different. It is worth noting that allelic variation in the ABCC2 gene has also been found to be associated with severe toxicity induced by chemotherapy agents,26 consistent with the idea that a polymorphism which impairs the pump efficiency might increase the toxic effect of metabolized drugs and in this way in the long term negatively affect survival. Moreover, given that two independent effects of −24C>T on transcriptional and posttranscriptional levels are possible,30 larger epidemiological studies are needed to understand the clinical relevance of this polymorphism in SCLC and NSCLC therapy and survival. Recently, in a study on 248 patients with SCLC performed at the Mayo clinic27 the potential impact of ABCC2 polymorphism on survival was tested. The authors found rs11597282 genotype to be associated with worse survival and several haplotypes to be associated with a better survival. In our study, we did not genotype rs11597282 because its MAF is lower (0.02) than the threshold we selected (0.05) and therefore we believed that we would have had only limited power to uncover a possible association. Sun and colleagues genotyped rs717620 but did not report any statistically significant association for this SNP. The possible explanation for this discrepancy in the findings could be a partially different chemotherapy/radiotherapy regimen. Sun et al report virtually all SCLC patients to have been treated with platinum-containing compounds (cisplatin and carboplatin) while in our study most patients were treated with a combination of etoposide and platinum. Ethnicity could also be an important factor which could at least in part explain the differences.

The fact that there is a considerable overlap in the substrates of most ABC transporters and therefore a built-in redundancy of the cellular export systems, could explain why we found no effects of large magnitude of the other single polymorphisms on response and prognosis.46 However, we cannot completely exclude the possibility that other SNPs in ABC transporter could be involved, at least in part, in the prognosis of lung cancer and in the response to treatment. In fact a great limitation in pharmacogenetics studies in general is the fact that a polymorphism that has no effect on a substrate drug in a particular study does not preclude a potential gene–drug interaction with another substrate drug. Moreover, these relationships are often dependent on route of administration, drug dosage and schedule, or can also be largely dependent on unknown genetic factors that could interact with those taken into consideration. For these reasons, studies that comprehensively take into consideration related genes in the same setting might offer more possibilities in discovering possible relationship between genetics and prognosis.

This is the first comprehensive study of all the common genetic variability in three key ABC genes, and highlights the effect of ABCC2 rs717620 on SCLC therapy. Our study is one of the largest of its kind, and investigated survival in both NSCLC and SCLC.

In conclusion, there are many indications that suggest ABCC2 genetic variation to be an important factor in lung cancer survival, although a very large, comprehensive study is warranted to further our understanding of how ABCC2 SNPs may impact lung cancer patient's survival.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The authors thank the patients and staff at the Thoraxklinik Heidelberg, for enabling this study. They thank Dr. Wolfgang Hagmann for critical reading of the manuscript, Dr. Heike Dally, Dr. Gisela Werle-Schneider, Birgit Jäger, Lin Zielske, Katharina Rauschenbach and Ingrid Heinzmann-Groth for their help with sample and/or data collection and Renate Rausch for helping with statistical analysis. The sample collection was in part funded by the Deutsche Krebshilfe (70-2919).

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
IJC_27567_sm_SuppTab1.doc36KSupporting Information Table 1.
IJC_27567_sm_SuppTab2_ABCB1C.xls45KSupporting Information Table 2.
IJC_27567_sm_SuppTab3_ABCC2C.xls31KSupporting Information Table 3.
IJC_27567_sm_SuppTab4_ABCG2.xls34KSupporting Information Table 4.
IJC_27567_sm_SuppTab5.doc60KSupporting Information Table 5.
IJC_27567_sm_SuppTab6.doc39KSupporting Information Table 6.
IJC_27567_sm_SuppTab7.doc64KSupporting Information Table 7.
IJC_27567_sm_SuppTab8.doc176KSupporting Information Table 8.
IJC_27567_sm_SuppTab9.doc191KSupporting Information Table 9.
IJC_27567_sm_SuppTab10.doc187KSupporting Information Table 10.
IJC_27567_sm_SuppTab11.doc97KSupporting Information Table 11.
IJC_27567_sm_SuppTab12.doc101KSupporting Information Table 12.
IJC_27567_sm_SuppTab13.doc101KSupporting Information Table 13.
IJC_27567_sm_SuppTab14.doc122KSupporting Information Table 14.
IJC_27567_sm_SuppTab15.doc130KSupporting Information Table 15.
IJC_27567_sm_SuppTab16.doc130KSupporting Information Table 16.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.