SEARCH

SEARCH BY CITATION

Keywords:

  • DNA methylation;
  • high-grade cervical intraepithelial neoplasia;
  • cervical scrapes, human papillomavirus;
  • tumour suppressor genes

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. References

Combined detection of cell adhesion molecule 1 (CADM1) and T-lymphocyte maturation-associated protein (MAL) promoter methylation in cervical scrapes is a promising triage strategy for high-risk human papillomavirus (hrHPV)-positive women. Here, CADM1 and MAL DNA methylation levels were analysed in cervical scrapes of hrHPV-positive women with no underlying high-grade disease, high-grade cervical intraepithelial neoplasia (CIN) and cervical cancer. CADM1 and MAL methylation levels in scrapes were first related to CIN-grade of the corresponding biopsy and second to CIN-grade stratified by the presence of ‘normal’ or ‘abnormal’ cytology as present in the accompanying scrape preceding the cervical biopsy. The scrapes included 167 women with ≤CIN1, 54 with CIN2/3 and 44 with carcinoma. In a separate series of hrHPV-positive scrapes of women with CIN2/3 (n = 48), methylation levels were related to duration of preceding hrHPV infection (PHI; <5 and ≥5 years). Methylation levels were determined by quantitative methylation-specific PCR and normal cytology scrapes of hrHPV-positive women with histologically ≤CIN1 served as reference. CADM1 and MAL methylation levels increased proportional to severity of the underlying lesion, showing an increase of 5.3- and 6.2-fold in CIN2/3, respectively, and 143.5- and 454.9-fold in carcinomas, respectively, compared to the reference. Methylation levels were also elevated in CIN2/3 with a longer duration of PHI (i.e. 11.5- and 13.6-fold, respectively). Moreover, per histological category, methylation levels were higher in accompanying scrapes with abnormal cytology than in scrapes with normal cytology. Concluding, CADM1 and MAL promoter methylation levels in hrHPV-positive cervical scrapes are related to the degree and duration of underlying cervical disease and markedly increased in cervical cancer.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. References

Persistent infection with high-risk human papillomavirus (hrHPV) types has been causally related to the development of cervical cancer.[1] The majority of cervical cancers are squamous cell carcinoma (SCC), which are preceded by noninvasive lesions referred to as cervical intraepithelial neoplasia (CIN).[2] Histologically, these lesions can be divided into mild (CIN1), moderate (CIN2) or severe (CIN3, including carcinoma in situ) lesions depending on the replacement of the epithelial lining by atypical cells. CIN1 (i.e. low-grade CIN) and part of CIN2 are mostly a consequence of productive hrHPV infections, while CIN3 and the remaining part of CIN2 are associated with transforming hrHPV infections with the potential of malignant progression towards invasive carcinoma.[3] The category of CIN2/3 (i.e. high-grade CIN) reflects a heterogeneous disease, both in terms of progression risk to invasive cancer, a process that generally lasts up to 10–30 years,[3-5] and in terms of the number of chromosomal aberrations.[6] We have recently demonstrated that the heterogeneity of CIN3 at the chromosomal level was related to the duration of preceding hrHPV infection (further referred to as PHI).[7] By comparing the chromosomal profiles of CIN2/3 biopsies of women with a short-term (<5 years PHI) and a long-term (i.e. ≥5 years PHI) hrHPV infection, we found a significantly higher number of chromosomal aberrations in lesions of women with a long-term PHI. These data support the notion that the number of chromosomal aberrations increases with progression of high-grade CIN disease. In addition to these chromosomal aberrations, epigenetic alterations, including DNA methylation of promoter regions of host cell genes involved in cervical carcinogenesis, are frequently detected in cervical carcinoma and a subset of CIN2/3 (reviewed by Wentzensen et al.[8]). We previously have identified CADM1 (cell adhesion molecule 1), originally referred to as tumour suppressor in lung cancer-1 and MAL (T-lymphocyte maturation-associated protein) as novel tumour suppressor genes functionally involved in cervical carcinogenesis.[9, 10] Promoter methylation was shown to be the main mode of inactivation of these genes.[10, 11] Subsequent studies using quantitative methylation-specific PCR (qMSP) analysis on tissue specimens revealed that virtually all CIN3 (97%) and carcinomas (99%) tested positive for CADM1 and/or MAL methylation.[12] Furthermore, combined detection of promoter methylation of these two genes in cervical scrapes of hrHPV-positive women was equally discriminatory for CIN3 or worse (CIN3+) as cytology or cytology combined with HPV16/18 genotyping, dependent on the threshold setting of the qMSP assays.[13] Hence, these methylation markers in combination with primary HPV-testing enable complete cervical screening by objective, nonmorphological molecular methods.

Of note, in our previous studies, a small subset of hrHPV-positive scrapes of women with underlying CIN2/3 had no elevated CADM1 or MAL methylation levels. It could be hypothesised that these lesions reflect earlier stages of disease compared to CIN2/3 that are scored methylation positive. Therefore, this study was set out to evaluate promoter methylation levels of the host cell genes CADM1 and MAL in relation to severity and duration of cervical disease. For this purpose, we studied the levels of promoter methylation in scrapes of hrHPV-positive women with no underlying high-grade disease, high-grade CIN and cervical cancer. Methylation levels were first related to CIN-grade and second to CIN-grade further stratified by cytology score, that is normal cytology or abnormal cytology (≥BMD [bordeline mild dyskaryosis] or ≥ASCUS [atypical squamous cells of undetermined significance] in the Bethesda classification) of the preceding cervical scrape.[14] Additionally, CADM1 and MAL methylation levels were examined in hrHPV-positive cervical scrapes of women with CIN2/3 following a short-term (<5 years) and long-term (≥5 years) PHI, respectively.[7]

Material and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. References

Cervical scrapes of women from a cross-sectional screening cohort

qMSP data of 221 consecutive hrHPV-positive women, participating in the population-based screening study Amsterdam (POBASCAM), were used.[13, 15, 16] Included were 54 women with CIN2/3, 19 of whom had scrapes with normal cytology. The median age of this group was 34 years (range 24–49). In addition, 167 women were included without evidence of CIN2/3 or cervical cancer (CIN2+), hereafter referred to as ≤CIN1; 140 of these women had scrapes with normal cytology and their follow-up scrape (approximately 5 years later) was both HPV negative and had normal cytology. The remaining 27 women were diagnosed with CIN1 upon histological assessment of colposcopy-guided biopsies. Their median age was 39 years (range 19–62).

Cervical scrapes of women with < and ≥5 years PHI

To determine the relationship of the duration of PHI, a surrogate for the duration of existence of a lesion,[7] with methylation levels, we analysed hrHPV-positive cervical scrapes of 48 women with CIN2/3 detected in the second screening round (approximately 5 years later). All these hrHPV-positive scrapes preceded the subsequent CIN2/3 biopsy. Duration of the hrHPV infections that preceded the CIN2/3 was defined as shown in Figure 1. CIN2/3 detected in the second screening round and derived from women who had at baseline hrHPV-positive, cytology normal smears, and who were hrHPV-positive for the same hrHPV type at the second screening round, were considered as having a PHI of ≥5 years. These women belonged to the control arm of the POBASCAM study (testing solely via cytomorphological assessment, HPV-status blinded). In the first screening round, these women were only sent for colposcopy when cytology was abnormal (≥BMD or equivalent ≥ASCUS in the Bethesda classification). In the second round, they were managed by the presence of HPV and/or abnormal cytology. CIN2/3 derived from women who were hrHPV negative and had normal cytology at baseline and were hrHPV positive in the second round were considered as having a PHI of <5 years.

image

Figure 1. Schematic overview of the duration of PHI that preceded the CIN2/3. Shown are the characteristics of the cervical scrape at baseline as well as in the second screening round, 5 years later. These latter scrapes preceded the CIN2/3 biopsy. Note: *In the cytology arm of the POBASCAM study, the intervention was based only on the presence of abnormal cytology because HPV status was blinded in the first round; **in the cytology arm of the POBASCAM study intervention in the second round was based on the presence of either HPV and/or abnormal cytology; ***the same hrHPV type as was present at baseline was also detected in the cytological sample 5 years later and in the biopsy.

Download figure to PowerPoint

For 19 of these women, no hrHPV infection could be detected in the baseline sample (taken 5 years earlier) and these women were classified as having CIN2/3 < 5-year PHI. In the remaining 29 women, a hrHPV infection was detected in the baseline sample, and after 5 years these women were classified as having CIN2/3 ≥ 5-year PHI. It was ensured that in these cases the hrHPV type in the cervical scrape preceding the biopsy matched the type detected at baseline. The median age of women in this study group was 40 years (range 34–56).

Cervical scrapes of women with carcinoma

Cervical scrapes of 44 women with cervical cancer (i.e. 34 SCC and 6 adenocarcinoma, AdCa) from the outpatient clinics of the VU University Medical Center (n = 18) and of the University Medical Center Groningen (n = 26)[17] were included for comparison. The median age of women with cervical carcinoma was 44 years (range 30–61 years).

This study was approved by the Institutional Review Boards of both the VU University Medical Center in Amsterdam, the Ministry of Public Health in The Hague (in case of POBASCAM) and the University Medical Center Groningen.

Histological diagnosis and further stratification by cytology

All biopsies from cervical lesions were graded according to the CIN classification. Preceding cytological scrapes were scored using the CISOE-A classification which is standard in the Netherlands and easy to translate into the Bethesda classification.[18, 19] For this study, we used cytology as either normal (Bethesda classification: negative for intraepithelial lesion or malignancy) or abnormal cytology (≥borderline or mild dyskaryosis, equalling ASCUS or worse according to the Bethesda system[20]).

We related methylation levels in the scrapes with the grade of the underlying CIN and presence of carcinoma (i.e. ≤CIN1, CIN2/3 and carcinoma). Next, we further stratified the grade of CIN according to the presence of normal or abnormal cytology (for instance CIN2/3, normal cytology and CIN2/3, abnormal cytology) and related these categories to methylation levels. The cervical scrapes of the ≤CIN1 group included those of women with a histologically confirmed CIN1 as well as those of women without evidence of clinically relevant disease, as they had an HPV-negative cytological normal cervical scrape at follow-up. Of women with histologically confirmed CIN2/3, the accompanying cervical scrapes were further defined based on < and ≥5-year PHI, which were assigned a lower and higher disease stage, respectively.

As duration of PHI was unknown in the cross-sectional screening cohort, cytological classification was included as an alternative, although less well defined, proxy for disease staging. Cervical scrapes of 140 women with normal cytology and ≤CIN1, obtained from the cross-sectional screening cohort, served as a reference in all comparisons. Accompanying cervical scrapes diagnosed as ‘abnormal cytology’ of women with ≤CIN1 (further referred to as ≤CIN1-abnormal cytology) were considered representative of either (i) relatively new lesions, (ii) regressive lesions or (iii) potential lesions missed by colposcopy. Thus, this group was considered a more severe disease category than ≤CIN1 with normal cytology (the reference group). Irrespective of the cytology status, cervical scrapes of women with a histologically confirmed CIN2/3 were considered a more severe category than those with histology of ≤CIN1. Histologically confirmed CIN2/3 of women with a scrape classified as normal cytology (referred to as CIN2/3-normal cytology) was considered to represent (i) relatively new lesions, (ii) regressive lesions, (iii) absence of intact abnormal indicator cells or (iv) a cytological sampling error. This disease category was classified as less severe than histologically confirmed CIN2/3 and a cytological abnormal scrape (referred to as CIN2/3-abnormal cytology).

The characteristics of the cervical scrapes included in the study, either or not stratified for cytology, are summarised in Table 1.

Table 1. Overview of the hrHPV-positive cervical scrapes analysed in this study
 NAge
MedianRange
  1. a

    Cross-sectional series.

  2. The first-level categorisation was based on histological diagnosis of the biopsy; second-level categorisation was based on the (ab)normality of the cervical scrape itself. Terms are written ‘histology diagnosis-cytology diagnosis’.

Categories
≤CIN11673919–62
CIN2/3a543424–49
CIN2/3 < 5-year PHI194039–50
CIN2/3 ≥ 5-year PHI294034–56
Carcinoma444430–61
Subcategories stratified by preceding cytology[2]
≤CIN1-normal cytology (reference group)1403919–62
≤CIN1-abnormal cytology274019–51
CIN2/3a-normal cytology193424–46
CIN2/3a-abnormal cytology353425–49

DNA isolation, HPV testing, bisulphite treatment and qMSP

DNA was isolated as described previously.[13, 17] The presence of hrHPV DNA was determined using the GP5+/6+-PCR-EIA detection and typing method.[21] For scrapes derived from the University Medical Center Groningen, DNA was isolated and HPV detection and typing was performed as described previously.[22-24]

Isolated DNA was subjected to bisulphite treatment and qMSP analysis for CADM1 M18 and MAL M1 as described before (primer and probe sequences are shown in Table 2).[11, 12] The β-actin gene was chosen as an internal reference.[25] Cycle threshold (Ct) values were measured at a fixed fluorescence threshold of 0.01 and Ct-ratios were determined as described in Hesselink et al.[13] Raw qMSP values were already available for the screening cohort.[13]

Table 2. Primers and probe sequences used for qMSP analysis
Gene promoter regionForward primer 5′-3′6-FAM 5′-3′ TAMRA probeReverse primer 5′-3′Tm (°C)Amplicon size (bp)
  1. Tm: annealing temperature.

CADM1 M18ATTTTATTAGTTGTTCGTTCGGGTACCTACCTCAAACTAACGACGTTAACTACCTCCGACTCGACAACACTACACTCGCC6090
MAL M1GCGTAGTATTAAGTAGAGAGGTTCGACTAAACCGACGCTAATTCGACGACGCTAATAAAAAATAAAACCGACCGC59107
β-actinTGGTGATGGAGGAGGTTTAGTAAGTACCACCACCCAACACACAATAACAAACACAAACCAATAAAACCTACTCCTCCCTTAA60133

Statistical analysis

Overall methylation levels per disease category were examined and fold changes over the reference (i.e. cervical scrapes of 140 women with normal cytology and ≤CIN1, obtained from the cross-sectional screening cohort), were calculated using the median methylation levels. The Mann–Whitney U test was used to determine whether any differences between the various groups were significant. Analyses were performed in SPSS (version 17). p-Values below 0.05 were considered significant.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. References

Methylation levels in hrHPV-positive cervical scrapes increase proportional to degree of underlying cervical disease and peak in women with cervical cancer

To test CADM1 and MAL methylation levels in hrHPV-positive cervical scrapes in relation to histological classification of underlying disease, hrHPV-positive scrapes of 54 women with CIN2/3 and 44 women with cervical cancer were analysed and resulting methylation levels compared to those of 140 hrHPV-positive women with ≤CIN1-normal cytology (the reference group). CADM1 and MAL methylation levels were 5.3- and 6.2-fold increased, respectively, in cervical scrapes of women with CIN2/3 compared to the reference group (both p < 0.0005; Table 3A). Cervical scrapes of women with carcinoma displayed the highest methylation levels, that is 143.5- and 453.9-fold increase for CADM1 and MAL, respectively.

Table 3. Overview of the median methylation levels and fold changes per study group compared to reference group
  CADM1 M18MAL M1
  Ct ratio Ct ratio 
  MedianRangeFold changeMedianRangeFold change
  1. a

    Cross-sectional series.

  2. The first-level categorisation was based on histological diagnosis of the biopsy; second-level categorisation was based on the (ab)normality of the cervical scrape itself. Terms are written ‘histology diagnosis-cytology diagnosis’.

ACategories
 ≤CIN10.0110–3.421.30.0380–9.8071.3
 CIN2/3a0.0450–6.3325.30.1820–6.8716.2
 CIN2/3 < 5-year PHI0.0250–0.2153.00.1060–0.3263.6
 CIN2/3 ≥ 5-year PHI0.0970–6.79711.50.40–8.41813.6
 Carcinoma1.220–171.204143.513.3970.109–471.208453.9
BSubcategories stratified by cytology[2]
 ≤CIN1-normal cytology0.0090–3.4201 (ref)0.030–9.8071 (ref)
 ≤CIN1-abnormal cytology0.0260–2.4513.00.0920–4.7523.1
 CIN2/3a-normal cytology0.0270–0.6383.10.0740–6.8712.5
 CIN2/3a-abnormal cytology0.0810–6.3329.50.2290–5.1837.8

Methylation levels in hrHPV-positive women with CIN2/3 increase with duration of PHI

In the CIN2/3 described in the previous paragraph, no information on duration of PHI and therefore proxy of duration of lesion existence was known as these were obtained from the first screening round of POBASCAM. Therefore, cervical scrapes of the second screening round were also examined and the duration of PHI (<5 years and ≥5 years) was used as a surrogate marker for the duration of existence of the lesion.[7] The methylation levels and fold changes detected in cervical scrapes of women with CIN2/3 < 5-year PHI and CIN2/3 ≥ 5-year PHI compared to the reference group are given in Table 3A. Methylation levels increased with duration of PHI; whereas methylation levels of CADM1 and MAL were 3.0- and 3.6-fold increased in the cervical scrapes of the CIN2/3 < 5-year PHI group compared to the reference group, respectively, cervical scrapes of the CIN2/3 ≥ 5-year PHI group had further elevated methylation levels, being 11.5- and 13.6-fold increased compared to the reference group. The elevation of methylation levels for both CADM1 and MAL was significantly higher in cervical scrapes of CIN2/3 ≥ 5-year PHI compared to CIN2/3 < 5-year PHI (p = 0.023 and p = 0.005, respectively; Fig. 2). Still, CADM1 and MAL methylation levels in the cervical scrapes of the carcinoma group were significantly higher than both CIN2/3 < 5-year PHI (both p < 0.0005) and CIN2/3 ≥ 5-year PHI groups (p = 0.002 and p < 0.0005, respectively).

Methylation levels increase in relation to both histology and cytology

To further test the hypothesis that methylation levels increase with progression of CIN2/3, women from the cross-sectional cohort were further stratified by baseline cytology results, as shown in Table 3B. In case of cytological abnormal scrapes of women with ≤CIN1 (≤CIN1-abnormal cytology), CADM1 and MAL methylation levels were increased 3.0- and 3.1-fold, respectively, compared to the scrapes classified as normal cytology of women with ≤CIN1 (≤CIN1-normal cytology), the reference group (p < 0.0005 and p = 0.049, respectively). Methylation levels for CADM1 and MAL in cervical scrapes classified as normal cytology of women with histologically confirmed CIN2/3 (CIN2/3-normal cytology) were similar to levels of ≤CIN1-abnormal cytology, that is 3.1- and 2.5-fold higher, respectively, compared to the reference group (p = 0.001 and p = 0.154, respectively; Fig. 3). Abnormal cytology scrapes of women with histologically confirmed CIN2/3 (CIN2/3-abnormal cytology) showed even further elevated methylation levels for both markers, that is 9.5- and 7.8-fold increase, respectively. The increase in methylation levels in cervical scrapes with CIN2/3-abnormal cytology compared to ≤CIN1-abnormal cytology nearly reached significance for CADM1 and was significant for MAL (p = 0.067 and p = 0.031, respectively).

image

Figure 2. Methylation levels, in log10 scale, relative to β-actin for (a) CADM1 and (b) MAL in hrHPV-positive cervical scrapes of women with underlying CIN2/3 with < and ≥5-year PHI.

Download figure to PowerPoint

image

Figure 3. Methylation levels, in log10 scale, relative to β-actin for (a) CADM1 and (b) MAL in cervical scrapes stratified for normal and abnormal cytology of women with either no underlying disease (≤CIN1) or with CIN2/3.

Download figure to PowerPoint

Methylation levels in the CIN2/3-normal cytology groups and the ≤CIN1-abnormal cytology group were comparable to those of detected in cervical scrapes of women with CIN2/3 < 5-year PHI (Table 3). Because of the low numbers of normal cytology samples no cytology subcategorisation could be made for cervical scrapes of CIN2/3 with a known duration of PHI and for carcinoma.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. References

Quantitative measurement of DNA methylation within the promoter regions of tumour suppressor genes CADM1 and MAL in hrHPV-positive cervical scrapes showed increased methylation levels proportional to degree and duration of underlying cervical disease. Both cervical scrapes of women with CIN2/3 and cervical cancer revealed significantly elevated methylation compared to the reference group (i.e. normal cytology scrapes of women with ≤CIN1). Moreover, methylation levels in cervical scrapes of women with CIN2/3 ≥ 5-year PHI were significantly higher compared to those with CIN2/3 < 5-year PHI. As duration of PHI can be considered as a surrogate for lesion age,[7] this suggests that methylation levels increase with lesion duration. With cytology added to further refine disease stage, methylation levels were consistently lower for samples with normal cytology compared to abnormal cytology in each histology class. Moreover, methylation levels of both CADM1 and MAL peak in cervical scrapes of women with cervical cancer.

These data indicate that elevated methylation levels are suggestive of progressive CIN disease and may be a reflection of the size of the underlying CIN. More advanced and larger lesions are likely to exfoliate more aberrant cells, facilitating the detection of promoter methylation in cervical scrapes. Accordingly, Wentzensen et al.[26] and Yang et al.[27] recently demonstrated that high-grade squamous intraepithelial lesion cytology and CIN3 colposcopy and biopsy results indicate larger CIN3. The observation that scrapes of women with CIN2/3 and a short-term (<5 years) PHI revealed lower methylation levels compared to lesions with long-term infections indicates that part of the scrapes of women with a short-term hrHPV infection may be scored as ‘methylation-negative’ when choosing certain thresholds of the respective qMSPs for scoring methylation. Although not significant (Fisher-exact test, p = 0.18), the proportion of scrapes with normal cytology was higher in women with CIN2/3 and a short-term hrHPV infection, that is 37%, compared to 17% in women with CIN2/3 and a long-term hrHPV infection. This supports the notion that CIN2/3 with short-term hrHPV infection are early onset CIN2/3 having a low progression risk for carcinoma. This is in line with our recent finding that CIN2/3 with a long-term and persistent hrHPV infection (≥5 years) had significantly more chromosomal aberrations compared to CIN2/3 of women with a short-term hrHPV infection.[7]

Interestingly, methylation levels in normal cytology scrapes of women with CIN2/3 and cytological abnormal scrapes of women with ≤CIN1 were comparable to those detected in cervical scrapes of women with CIN2/3 < 5-year PHI. This is in agreement with low to intermediate levels of methylation being indicative of a relatively new CIN2/3, or a potentially regressive lesion, and do not point to a high short-term progression risk. In fact, methylation levels were higher in cervical scrapes of more advanced CIN2/3 (≥5-year PHI) and dramatically increased in case of cervical cancer, indicating that cancers are unlikely to be missed by these methylation markers.

Remarkably, within the group of hrHPV-positive women with normal cytology, a significant increase in CADM1 methylation and an increase in MAL methylation were seen in the cervical scrapes preceding a histological diagnosis of CIN2/3 compared to ≤CIN1. This implies that methylation analysis is capable of accurately detecting CIN2/3 that are missed by cytology, potentially due to absence of intact, abnormal indicator cells, or due to cytological sampling error. In fact, upon setting a certain threshold for methylation-positivity, as described in our previous study,[13] 75% (15/20) of CIN2+ and 85% (11/13) of CIN3+ could be detected at a specificity of 66 and 65%, respectively, in hrHPV-positive women with normal cytology in our population-based screening cohort. In comparison, HPV16/18 genotyping, a currently proposed triage strategy for hrHPV-positive women with normal cytology,[28] only reached a sensitivity of 50% for CIN2+ and 54% for CIN3+ at a similar specificity (69%) (data not shown).

These data are in line with previous studies showing an increase in the number of methylation-positive samples, or an increase in methylation levels with either the degree of histological or cytological abnormality.[8, 12, 29-36] However, to the best of our knowledge, this is the first study to demonstrate a relationship to duration of high-grade CIN disease.

Based on these findings, it may be suggested that undetectable or very low methylation levels in cervical scrapes point to a low or negligible risk of clinically meaningful cervical disease. Intermediate levels indicate potential risk of underlying disease, warranting close surveillance. High methylation levels indicate the presence of clinically relevant disease requiring immediate referral to the gynaecologist.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. References