Periodic Report Summary - O-PTM-BIOMARKERS (Discovery of novel cancer serum biomarkers based on ... modifications of O-glycoproteins and their application to early detection of cancer
Cancer is a leading cause of death in developed countries, including Europe with one in three people being affected. Early detection, accomplished through an efficient screening programme, remains the most promising approach to improve the long-term survival of cancer patients. Therefore there is a pressing need for the development of biomarkers, which detect the early changes that lead to overt malignancy, particularly for cancers where clinical symptoms only appear when the cancer has progressed and patients have poor survival, e.g. ovarian, pancreatic and lung cancer. This project addressed this problem.
Our proposal combined two unique features:
1. large serum collections having the potential for evaluation of diagnostic serum biomarkers, with high statistical power, and
2. a unique approach involving both highly specific immuno-detection of cancer associated post-translationally modified (PTM) glycoproteins in serum and detection of immunity to these. Our approach to identifying diagnostic biomarkers, useful for early detection of cancer, was based on distinct cancer associated changes in one of the most abundant post-translational modification of proteins namely O-glycosylation (O-PTM). Carcinomas characteristically express O-glycosylated glycoproteins, including membrane mucins such as MUC1 and MUC16, which are aberrantly glycosylated. These glycoproteins are released into serum and are the targets of the current biomarker assays CA15-3 and CA125. The composition and density of attached O-glycans on glycoproteins are changed in carcinomas, and we developed strategies to screen and selectively immuno-probe these changes in a protein and O-glycan specific manner. We demonstrated that composite cancer associated glycopeptide epitopes appeared, being specific for both the core protein and the novel glycan.
These novel epitopes distinguished the cancer glycoprotein from the normal counterpart and induced auto-antibodies which represented potential biomarkers of early stage cancers. The actual cancer glycoproteins could also be shed into serum, but had only been detected in late stage patients so far.
The main objectives of this project were therefore:
1. to identify evaluate and validate an O-PTM auto-antibody signature as an early diagnostic biomarker for breast ovarian, pancreatic and lung cancers using two unique serum banks and novel microarray technology
2. to develop and validate enzyme linked immunosorbent assays (ELISAs) for cancer specific glycoforms of MUC1 and MUC16 for patient management and screening for the above cancers.
The project benefited from having matchless resources, such as access to large serum banks, e.g.
1. the United Kingdom collaborative trial of ovarian cancer screening (UKCTOCS) bank. This was the largest randomised controlled trial of ovarian cancer screening in the world and had recruited 202 638 postmenopausal women who were randomised into a control group or annual screening for six years with ultrasound or serum CA125. A baseline blood sample was available for all women and annual serum samples from the 50 000 women in the CA125 group. There also was a well established and sophisticated set up for systematically collecting, transporting, cataloguing and storing serum samples and over 300 000 samples had already been stored. All women were flagged for 15 years through the Office of National Statistics, which provided information about any cancers they developed and possible deaths. In addition, all women were sent two postal follow up questionnaires 3.5 and 7 years after randomisation. Over 4 000 women were identified to have developed cancer after randomisation into the trial. Cancers were identified based on data provided from the national cancer registry through the office of national statistics.
2. The Guernsey bank. The Imperial Cancer Research Fund (ICRF) began a project in the 1960s to establish a bank of sera to evaluate risk factors and predictive markers for cancer. At the time the primary interest was in breast cancer and the women on the island of Guernsey were selected for this study because the population was very stable and there was only one hospital and six primary care units on the island, making the follow up for detecting the development of cancer relatively easy and complete. Even though studies were initiated in the 1960s, the series available to us were collected from 1977 to 1991. These women were all given mammagraphic screening at the time of blood donation. Sera were collected from 6 500 women between 1977 and 1991 and so far 683 cases of breast, ovarian, lung and pancreatic cancers had been diagnosed. In general, sera from individuals that had not developed cancer provided excellent controls.
3. P series of sera from breast cancer patients. The P series were sera from breast cancer patients collected from consecutive early stage patients attending Guy's Hospital, London, United Kingdom between February 1975 and March 1980. Stage I cancers were restricted to the breast, while Stage II cancers had spread to the lymph nodes, but distant metastases had not been detected. Sera were taken with informed patient consent and were not taken as part of a particular clinical trial. Pairs of sera were taken pre and 10 days post mastectomy from 739 patients. Complete clinical follow up was available for these patients from the Smart data base at Guy's Hospital.
Additional resources included novel technology resulting in the application of micro-arrays carrying glycopeptides with cancer-associated O-glycoproteins (O-PTMs).
Glycopeptides were enzymatically synthesised in vitro using methods developed by beneficiaries in this project. Development of this array technology relied on our data generated in the post genomic era.
The first project objective was to identify, evaluate and validate an O-PTM auto-antibody signature as an early diagnostic biomarker for breast ovarian, pancreatic and lung cancers using two unique serum banks and novel microarray technology.
Screening for auto-antibodies using sera from well women was planned to proceed using two series of sera, a discovery set and a validation set. The discovery set comprised of sera from the UKCTOCS cohort taken from women who went on to develop breast, ovarian, lung or pancreatic cancer within one to three years of donating blood, together with age matched controls. This screen proceeded as planned and a total of 1 125 sera samples were screened on the glycopeptides microarrays. However, because of the results, the content of the validation set was redesigned, consisting of sera taken from women before they developed clinically diagnosed breast cancer with aged matched controls that did not have cancer when the case was diagnosed. Sera were drawn from the UKCTOCs and Guernsey cohorts, with an extra control coming from the Guernsey cohort. A total of 1 822 sera samples were screened in duplicate on the glycopeptides arrays.
Glycopeptides based on MUC1 carrying cancer-associated glycan were synthesised and printed in an array format. The glycopeptides were verified using sera from the P-series on this first generation array format. A second generation array was used to screen the discovery set of sera and a slightly modified version used for the new validation set.
The repertoire was expanded to produce synthetic peptides and recombinant proteins in escherichia coli (e. coli) from mucins MUC2, MUC5AC and MUC6. These were glycosylated to carry the Tn and STn glycans and the glycosylated and unglycosylated peptides were used to screen, in a microarray format, sera from gastric carcinoma patients with the appropriate controls.
To increase the breath of target antigens the project contained a glycoprotemics element. The serum samples were depleted of highly abundant serum proteins and enriched for O-PTM proteins by serial lectin affinity capture. Separation, identification and analysis of novel O-PTM proteins proceeded using different strategies. Glycopeptides from these proteins were synthesised, glycosylated and screened with a small number of commercial sera.
The second objective was to develop and validate enzyme-linked immunosorbent assays (ELISA) for cancer specific glycoforms of MUC1 and MUC16 for patient management and screening for the above cancers. New ELISAs were developed for detection of specific MUC1 glycoforms but had not shown an improvement over the existing commercially available kits. MUC16 was an enormous glycoprotein with two domains, so it presented many problems for the investigator. Glycopeptides based on MUC16 and larger recombinant peptides made in e. coli were used in an effort to develop new O-PTM specific antibodies to MUC16. This work led to the identification of the region of MUC16 that was recognised by one of the antibodies used in the CA125 kit.
The major project findings were that:
1. the presence and level of auto-antibodies to MUC1 glycopeptides was significantly higher in the sera from breast cancer patients compared with the controls and a highly significant correlation with age was observed. No such auto-antibodies were found in patients with metastatic breast cancer. High levels of a subset of auto-antibodies to the core3MUC1 and STnMUC1 glycoforms were significantly associated with reduced incidence and increased time to metastasis
2. in order to extend our relatively simple combination of threshold decision rules to something that could represent more complex patterns in the data, we applied boosted decision trees to the P series data. We also used conformal predictors to further enhance this method by allowing us to preselect a level of error and meet this error by allowing our rule to be 'unsure' about predictions
3. small numbers of sera were screened from controls without gastric disease, patients with chronic gastritis or with intestinal metaplasia and patients with gastric carcinoma on microarrays of glycopeptides of MUC5AC, MUC2 and MUC6. Reactivity to MUC2 naked/glycoforms and MUC5AC naked/glycoforms was more pronounced in intestinal metaplasia and carcinomas
4. discovery set of sera showed no increase in auto-antibodies above the controls. Moreover, some of the controls in this cohort gave relatively high levels of signals and often showed a higher rate of positivity than the cancer cases but this was not statistically significant
5. the results of the discovery set led to the redesign of the validation set, concentrating on breast cancers cases as the largest number of serum samples were available for this cancer type, and including age matched controls and an additional control group from the Guernsey cohort who never developed cancer. However, the results were again disappointing with no differences in antibody levels being observed between the cancers and controls or between the two control groups
6. glycoproteomics revealed potential new targets carrying tumour associated glycans
7. analysis of the specificity of commercially available CA15.3 assays using recombinant MUC1 carrying defined O-linked glycans showed differences in the specificity of the kit dependent on the manufacturer
8. glycoprofiling of MUC1 and MUC16 found in sera showed that these mucins only carried sialylated glycans when present in serum.
The programme would likely have a positive impact on the quality of life of European Union citizens and the strength of the European science and technology base. Auto-antibodies in the sera of breast cancer patients binding to MUC1 were associated with fewer metastases and a longer time to the development of metastasis. These therefore had the potential to predict the chances of developing metastasis within a certain number of years, which was important to know for the management of breast cancer patients. Further research needed to be done to confirm the results in an independent cohort and identify the epitope recognised by the auto-antibodies in the sera of breast cancer patients that bound to MUC1-core three. Published data and our own results showed that the enzyme responsible for synthesising the core three glycan was not expressed by the breast or by breast cancers, therefore the auto-antibodies might be recognising a mimic and the true epitope needed to be defined.
This project also established the differences in performance of the different CA15.3 assays. This was not generally appreciated by the medical and scientific community. One of the problems with the assays was the lack of a consistent and definable MUC1 standard.
Furthermore, groups with complementary expertise were integrated to form a strong internationally competitive consortium, thus the European scientific potential was promoted. A unique technology platform for a broadly based biomarker discovery programme only available in Europe was also established, along with the supreme importance of using large numbers of samples to evaluate potential biomarkers.
Our proposal combined two unique features:
1. large serum collections having the potential for evaluation of diagnostic serum biomarkers, with high statistical power, and
2. a unique approach involving both highly specific immuno-detection of cancer associated post-translationally modified (PTM) glycoproteins in serum and detection of immunity to these. Our approach to identifying diagnostic biomarkers, useful for early detection of cancer, was based on distinct cancer associated changes in one of the most abundant post-translational modification of proteins namely O-glycosylation (O-PTM). Carcinomas characteristically express O-glycosylated glycoproteins, including membrane mucins such as MUC1 and MUC16, which are aberrantly glycosylated. These glycoproteins are released into serum and are the targets of the current biomarker assays CA15-3 and CA125. The composition and density of attached O-glycans on glycoproteins are changed in carcinomas, and we developed strategies to screen and selectively immuno-probe these changes in a protein and O-glycan specific manner. We demonstrated that composite cancer associated glycopeptide epitopes appeared, being specific for both the core protein and the novel glycan.
These novel epitopes distinguished the cancer glycoprotein from the normal counterpart and induced auto-antibodies which represented potential biomarkers of early stage cancers. The actual cancer glycoproteins could also be shed into serum, but had only been detected in late stage patients so far.
The main objectives of this project were therefore:
1. to identify evaluate and validate an O-PTM auto-antibody signature as an early diagnostic biomarker for breast ovarian, pancreatic and lung cancers using two unique serum banks and novel microarray technology
2. to develop and validate enzyme linked immunosorbent assays (ELISAs) for cancer specific glycoforms of MUC1 and MUC16 for patient management and screening for the above cancers.
The project benefited from having matchless resources, such as access to large serum banks, e.g.
1. the United Kingdom collaborative trial of ovarian cancer screening (UKCTOCS) bank. This was the largest randomised controlled trial of ovarian cancer screening in the world and had recruited 202 638 postmenopausal women who were randomised into a control group or annual screening for six years with ultrasound or serum CA125. A baseline blood sample was available for all women and annual serum samples from the 50 000 women in the CA125 group. There also was a well established and sophisticated set up for systematically collecting, transporting, cataloguing and storing serum samples and over 300 000 samples had already been stored. All women were flagged for 15 years through the Office of National Statistics, which provided information about any cancers they developed and possible deaths. In addition, all women were sent two postal follow up questionnaires 3.5 and 7 years after randomisation. Over 4 000 women were identified to have developed cancer after randomisation into the trial. Cancers were identified based on data provided from the national cancer registry through the office of national statistics.
2. The Guernsey bank. The Imperial Cancer Research Fund (ICRF) began a project in the 1960s to establish a bank of sera to evaluate risk factors and predictive markers for cancer. At the time the primary interest was in breast cancer and the women on the island of Guernsey were selected for this study because the population was very stable and there was only one hospital and six primary care units on the island, making the follow up for detecting the development of cancer relatively easy and complete. Even though studies were initiated in the 1960s, the series available to us were collected from 1977 to 1991. These women were all given mammagraphic screening at the time of blood donation. Sera were collected from 6 500 women between 1977 and 1991 and so far 683 cases of breast, ovarian, lung and pancreatic cancers had been diagnosed. In general, sera from individuals that had not developed cancer provided excellent controls.
3. P series of sera from breast cancer patients. The P series were sera from breast cancer patients collected from consecutive early stage patients attending Guy's Hospital, London, United Kingdom between February 1975 and March 1980. Stage I cancers were restricted to the breast, while Stage II cancers had spread to the lymph nodes, but distant metastases had not been detected. Sera were taken with informed patient consent and were not taken as part of a particular clinical trial. Pairs of sera were taken pre and 10 days post mastectomy from 739 patients. Complete clinical follow up was available for these patients from the Smart data base at Guy's Hospital.
Additional resources included novel technology resulting in the application of micro-arrays carrying glycopeptides with cancer-associated O-glycoproteins (O-PTMs).
Glycopeptides were enzymatically synthesised in vitro using methods developed by beneficiaries in this project. Development of this array technology relied on our data generated in the post genomic era.
The first project objective was to identify, evaluate and validate an O-PTM auto-antibody signature as an early diagnostic biomarker for breast ovarian, pancreatic and lung cancers using two unique serum banks and novel microarray technology.
Screening for auto-antibodies using sera from well women was planned to proceed using two series of sera, a discovery set and a validation set. The discovery set comprised of sera from the UKCTOCS cohort taken from women who went on to develop breast, ovarian, lung or pancreatic cancer within one to three years of donating blood, together with age matched controls. This screen proceeded as planned and a total of 1 125 sera samples were screened on the glycopeptides microarrays. However, because of the results, the content of the validation set was redesigned, consisting of sera taken from women before they developed clinically diagnosed breast cancer with aged matched controls that did not have cancer when the case was diagnosed. Sera were drawn from the UKCTOCs and Guernsey cohorts, with an extra control coming from the Guernsey cohort. A total of 1 822 sera samples were screened in duplicate on the glycopeptides arrays.
Glycopeptides based on MUC1 carrying cancer-associated glycan were synthesised and printed in an array format. The glycopeptides were verified using sera from the P-series on this first generation array format. A second generation array was used to screen the discovery set of sera and a slightly modified version used for the new validation set.
The repertoire was expanded to produce synthetic peptides and recombinant proteins in escherichia coli (e. coli) from mucins MUC2, MUC5AC and MUC6. These were glycosylated to carry the Tn and STn glycans and the glycosylated and unglycosylated peptides were used to screen, in a microarray format, sera from gastric carcinoma patients with the appropriate controls.
To increase the breath of target antigens the project contained a glycoprotemics element. The serum samples were depleted of highly abundant serum proteins and enriched for O-PTM proteins by serial lectin affinity capture. Separation, identification and analysis of novel O-PTM proteins proceeded using different strategies. Glycopeptides from these proteins were synthesised, glycosylated and screened with a small number of commercial sera.
The second objective was to develop and validate enzyme-linked immunosorbent assays (ELISA) for cancer specific glycoforms of MUC1 and MUC16 for patient management and screening for the above cancers. New ELISAs were developed for detection of specific MUC1 glycoforms but had not shown an improvement over the existing commercially available kits. MUC16 was an enormous glycoprotein with two domains, so it presented many problems for the investigator. Glycopeptides based on MUC16 and larger recombinant peptides made in e. coli were used in an effort to develop new O-PTM specific antibodies to MUC16. This work led to the identification of the region of MUC16 that was recognised by one of the antibodies used in the CA125 kit.
The major project findings were that:
1. the presence and level of auto-antibodies to MUC1 glycopeptides was significantly higher in the sera from breast cancer patients compared with the controls and a highly significant correlation with age was observed. No such auto-antibodies were found in patients with metastatic breast cancer. High levels of a subset of auto-antibodies to the core3MUC1 and STnMUC1 glycoforms were significantly associated with reduced incidence and increased time to metastasis
2. in order to extend our relatively simple combination of threshold decision rules to something that could represent more complex patterns in the data, we applied boosted decision trees to the P series data. We also used conformal predictors to further enhance this method by allowing us to preselect a level of error and meet this error by allowing our rule to be 'unsure' about predictions
3. small numbers of sera were screened from controls without gastric disease, patients with chronic gastritis or with intestinal metaplasia and patients with gastric carcinoma on microarrays of glycopeptides of MUC5AC, MUC2 and MUC6. Reactivity to MUC2 naked/glycoforms and MUC5AC naked/glycoforms was more pronounced in intestinal metaplasia and carcinomas
4. discovery set of sera showed no increase in auto-antibodies above the controls. Moreover, some of the controls in this cohort gave relatively high levels of signals and often showed a higher rate of positivity than the cancer cases but this was not statistically significant
5. the results of the discovery set led to the redesign of the validation set, concentrating on breast cancers cases as the largest number of serum samples were available for this cancer type, and including age matched controls and an additional control group from the Guernsey cohort who never developed cancer. However, the results were again disappointing with no differences in antibody levels being observed between the cancers and controls or between the two control groups
6. glycoproteomics revealed potential new targets carrying tumour associated glycans
7. analysis of the specificity of commercially available CA15.3 assays using recombinant MUC1 carrying defined O-linked glycans showed differences in the specificity of the kit dependent on the manufacturer
8. glycoprofiling of MUC1 and MUC16 found in sera showed that these mucins only carried sialylated glycans when present in serum.
The programme would likely have a positive impact on the quality of life of European Union citizens and the strength of the European science and technology base. Auto-antibodies in the sera of breast cancer patients binding to MUC1 were associated with fewer metastases and a longer time to the development of metastasis. These therefore had the potential to predict the chances of developing metastasis within a certain number of years, which was important to know for the management of breast cancer patients. Further research needed to be done to confirm the results in an independent cohort and identify the epitope recognised by the auto-antibodies in the sera of breast cancer patients that bound to MUC1-core three. Published data and our own results showed that the enzyme responsible for synthesising the core three glycan was not expressed by the breast or by breast cancers, therefore the auto-antibodies might be recognising a mimic and the true epitope needed to be defined.
This project also established the differences in performance of the different CA15.3 assays. This was not generally appreciated by the medical and scientific community. One of the problems with the assays was the lack of a consistent and definable MUC1 standard.
Furthermore, groups with complementary expertise were integrated to form a strong internationally competitive consortium, thus the European scientific potential was promoted. A unique technology platform for a broadly based biomarker discovery programme only available in Europe was also established, along with the supreme importance of using large numbers of samples to evaluate potential biomarkers.