A spectral phenotype of oncogenic human papillomavirus-infected exfoliative cervical cytology distinguishes women based on age

Clinica Chimica Acta 411 (2010) 1027–1033

Contents lists available at ScienceDirect

Clinica Chimica Acta j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c l i n c h i m

A spectral phenotype of oncogenic human papillomavirus-infected exfoliative cervical cytology distinguishes women based on age Jemma G. Kelly a, Karen T. Cheung a, Cara Martin b, John J. O'Leary b, Walter Prendiville b, Pierre L. Martin-Hirsch a,c, Francis L. Martin a,⁎ a b c

Centre for Biophotonics, Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK Department of Pathology, Coombe Women and Infants University Hospital, Dolphins Barn Road, Dublin 8, Ireland Lancashire Teaching Hospitals NHS Trust, Preston PR2 9HT, UK

a r t i c l e

i n f o

Article history: Received 9 March 2010 Received in revised form 19 March 2010 Accepted 19 March 2010 Available online 30 March 2010 Keywords: Exfoliative cervical cytology Human papillomavirus Infrared spectroscopy Multivariate analysis Persistent infection

a b s t r a c t Background: Human papillomavirus (HPV) is a sexually-transmitted infection associated with cervical cancer. Of over 100 HPV types identified, 13 are high-risk oncogenic. In unvaccinated women worldwide, the incidence of cervical cancer from HPV16 and HPV18 will remain. Cervical cytology can be graded from normal (atypia-free) to low-grade to high-grade. Infrared (IR) spectroscopy is a non-destructive technique that allows the acquisition of a biochemical-cell fingerprint based on vibrational states of chemical bonds. Methods: Exfoliative cervical cytology specimens (n = 147) were retrieved, graded by a cytologist and HPVtested/genotyped using hybrid capture 2 and the Roche HPV Linear Array. Additionally, the spectral signatures of cervical cell lines C33A, HeLa and SiHa were examined. After washing, cellular material was transferred to low-E glass slides and interrogated using attenuated total reflection Fourier-transform IR (ATR-FTIR) spectroscopy. Given the complex nature of the dataset consisting of thousands of variables (wavenumbers), we used multivariate analysis for data reduction and information retrieval. Principal component analysis coupled with linear discriminant analysis (PCA-LDA) generated a visual representation of the data (scores plot) and, identification of the wavenumbers and consequent biochemical entities responsible for segregation (loadings plot). Results: Immortalised cell lines were readily distinguishable from each other. It was difficult to segregate categories of cytology associated with HPV infection types. However, in low-grade cytology infected with high-risk oncogenic HPV16 or HPV18, it was possible to segregate women based on whether they were aged 20–29 years vs. 30–39 years. Conclusions: Our findings suggest a spectral phenotype in exfoliative cervical cytology associated with transient vs. persistent HPV infection. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Human papillomavirus (HPV) infection is associated with cervical cancer, affecting thousands of women worldwide. It is a sexuallytransmitted infection and risk factors include smoking, number of sexual partners and age of 1st sexual relationship of either female or male partner [1,2]. Approximately 70% of women have a HPV infection during their lifetime but a combination of high-risk (hr) HPV type and a compromised immune system is a fundamental factor in persistent Abbreviations: ATR-FTIR spectroscopy, attenuated total reflection Fourier-transform infrared spectroscopy; IR, infrared; PCA, principal component analysis; LDA, linear discriminant analysis; HPV, human papillomavirus; CIN, cervical intraepithelial − neoplasia; νsPO− 2 , symmetric phosphate stretching vibrations; νasPO2 , asymmetric phosphate stretching vibrations; hrHPV, high-risk HPV; lrHPV, low-risk HPV; 3rd decade women, women aged 20–29 years; 4th decade women, women aged 30–39 years. ⁎ Corresponding author. Tel.: + 44 1524 510206; fax: + 44 1524 593985. E-mail address: [email protected] (F.L. Martin). 0009-8981/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.cca.2010.03.029

infection with consequent progression to cervical intraepithelial neoplasia (CIN) and invasive disease [3]. The majority of women with a HPV infection in their 20s (3rd decade) are HPV negative (HPV−) years later and, in these individuals cervical cancer tends not to develop. The time required for HPV clearance is dependent on infection type, i.e. low-risk (lr) (≈5–6 months) or hr (≈8–14 months) HPV [4]. Women in their 30s (4th decade) with HPV are classed as having a persistent infection and at a higher risk of cervical cancer. This suggests that cervical specimens of the same cytological grade from patients in their 30s compared to those in their 20s may have underlying biochemical differences. Over 100 HPV serotypes exist, including 13 hr oncogenic types [1,5]. HPV16 and HPV18 are the most common hr serotypes, together responsible for N70% of cervical cancers worldwide. HPV16 is associated with squamous cell carcinoma, while HPV18 is associated with adenocarcinoma [6]. A cervical smear involves sampling of cells from the transformation zone of the cervix. This zone contains two main cell types: stratified

1028

J.G. Kelly et al. / Clinica Chimica Acta 411 (2010) 1027–1033

squamous epithelial cells (ecto-cervix) and columnar mucous secreting epithelial cells (endo-cervical canal). Cervical cytology is graded as normal (free from atypia), low-grade (borderline nuclear abnormality/cervical intraepithelial neoplasia [CIN]1 or low-grade squamous intraepithelial lesions [LSIL]), high-grade (CIN2/3 or high-grade squamous intraepithelial lesions [HSIL]) and severe dyskaryosis (? carcinoma). Development of a HPV hierarchical model of persistence based on cytology biochemistry to facilitate patient triage would reduce the need for invasive procedures. A bivalent vaccine against HPV16 and HPV18, and a quadrivalent vaccine against HPV6, HPV11, HPV16 and HPV18 have recently been introduced; a vaccination programme is in place for pre-pubescent girls [7]. Although both immunise against the common hrHPV types any knock-on consequences remain unknown, i.e., other hr types becoming more common. Despite the introduction of the cervical cancer vaccination programme, the number of women included in the cervical screening programme will remain the same as the vaccine does not protect against other oncogenic HPV types [3,6,7]. Also, the level and duration of protection following vaccination is largely unknown [8]. HPV classification is currently based on genotypes distinguished by DNA sequence changes to the coding regions E6, E7 and L1 of the HPV genome [1]. There are three genomic regions; the long control region (origin of replication and regulator of gene expression), region of early proteins (E1–E8) and the late protein region (L1, L2). E6 and E7 are potent viral oncogenes expressed in HPV16 and HPV18 as well as frequently associated with invasive disease [9]. E6 can inhibit the function of p53 and E7 is a proliferation inducing oncogene [10]. HPV is a stable virus, resistant to heat and organic solvents [1]. Several cervical cell lines, with different HPV infections, are available and provide in vitro models to study cervical carcinoma; examples include C33A, HeLa, SiHa and CaSki. Current techniques for HPV detection and genotyping include the hybrid capture 2 (hc2) and the Linear Array HPV Genotyping test [11]. The hc2 is used to identify 13 high-risk HPV types and low-risk HPV types, while the Linear Array HPV Genotyping test is capable of distinguishing between 37 HPV subtypes enabling detection of multiple infections [1]. Attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy can be applied to detect the absorption of infrared (IR) radiation by biomolecules. Molecules absorb the mid-IR (λ = 2.5– 25 μm) at specific wavelengths corresponding to energy levels of the chemical bonds present generating a spectrum or biochemical-cell fingerprint (1800–900 cm−1). This region contains spectral peaks associated with lipids (≈1750 cm−1), Amide I (≈1650 cm−1), Amide II (≈1550 cm−1), Amide III (≈1260 cm−1), asymmetric phosphate −1 stretching vibrations (ν as PO − ), carbohydrates 2 ; ≈ 1225 cm (≈1155 cm−1), symmetric phosphate stretching vibrations (νsPO− 2 ; ≈1080 cm−1), glycogen (≈1030 cm−1) and protein phosphorylation (≈970 cm−1) [12]. In ATR-FTIR spectroscopy, the IR radiation is totally internally reflected within a crystal giving rise to an evanescent wave that penetrates the adjacent sample by a few μm. A spot size of ≈250 μm × 250 μm allows for a large sampling size and maximum coverage of the interrogated sample. An absorbance spectrum of the sample is derived following Fourier-transformation of the detected IR beam. ATR-FTIR spectroscopy has previously been used to distinguish normal, low-grade and high-grade cervical cytology specimens [13–15]. ATR-FTIR spectroscopy can generate large, complex datasets with hundred of variables (wavenumbers). Principal component analysis (PCA) is an unsupervised technique that reduces data dimensionality by forming linear combinations [principal components (PCs)] of original variables. Ranking the PCs in order of variance enables the data to be visually presented and subtle differences in the spectra to be exposed. Linear discriminant analysis (LDA) is supervised and maximises inter-category variance while minimising intra-category variance; this allows one to highlight the important discriminating features between categories. LDA can be employed following PCA; resultant scores plots represent the spread of data and loadings plots

identify the wavenumbers responsible for segregation. Our aim was to determine if ATR-FTIR spectroscopy coupled with PCA-LDA could be employed to discriminate exfoliative cervical cytology based on HPV infection. 2. Participants and methods 2.1. Specimens and preparation Cervical smear specimens (n = 147) collected in PreservCyt liquidbased cytology medium were used in this study. Specimens screened in this study were collected anonymously as part of a large cervical screening HPV study under the umbrella organisation, CERVIVA (The Irish Cervical Screening Research Consortium); ethical approval was obtained from the Research Ethics Committee at the Coombe Women and Infants University Hospital, Dublin, Ireland. An aliquot of 1 ml to 2 ml of PreservCyt smear specimen was centrifuged at 1500 rpm for 5 min, after which the supernatant was then aspirated. The remaining cell pellet was re-suspended in 3 ml autoclaved distilled H2O and centrifuged at 1500 rpm for 5 min, and the supernatant was again removed. This wash step was repeated three times and, the resulting cell pellet was then re-suspended in 0.5 ml autoclaved distilled H2O and transferred to a low-E glass microscope slide (Kevley Technologies, Chesterland, OH, USA). Slides were allowed to air-dry and stored in a desiccator until analysis. Cervical cancer cell lines were obtained from the American Type Culture Collection (ATCC); SiHa (positive for HPV16), HeLa (positive for HPV18) and C33A (HPV−) were cultured until confluent in modified Eagle's medium (MEM) supplemented with 10% foetal calf serum, 1% penicillin and streptomycin, transferred to PreservCyt solution for transportation and the above procedure followed for preparation of slides for ATR-FTIR spectroscopic interrogation. C33A cells are HPV−, extracted from a 66 year Caucasian woman with cervical carcinoma. HeLa cells are from a 33 year woman with cervical adenocarcinoma. SiHa cells were originally isolated from a 55 year woman with grade II cervical squamous cell carcinoma. 2.2. HPV detection and typing The initial HPV screen on all specimens was performed using the hc2 test (Qiagen, UK). Briefly, DNA was extracted from an aliquot of 4 ml PreservCyt sample using the Sample Conversion Kit (Qiagen) for HPV testing by hc2. The HPV DNA status of the specimens were assessed using the hc2 HPV kit for hrHPV detection of types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, according to manufacturer's guidelines. Specimens that tested positive for HPV by hc2 test were then further genotyped using the Linear Array HPV test (Roche Diagnostics, Germany). The kit can detect 37 different hrHPV and lrHPV genotypes. DNA was extracted from a 250 μl aliquot of PreservCyt sample using the Ampilute Media Extraction Kit (Roche Diagnostics, Germany) along with a supplied positive (HPV16 plasmid) and a negative control. DNA was amplified in a multiplex PCR which detects 37 HPV genotypes along with β-globin controls. Following amplification, the DNA sample is denatured, hybridised in a reverse hybridisation reaction to a strip containing multiple oligonucleotide probes captured onto a test strip supplied with the kit, and detection performed in a colorimetric reaction as per manufacturer's instructions. 2.3. ATR-FTIR spectroscopy Spectroscopic interrogation was performed using a Bruker Vector 22 FTIR spectrometer with Helios attachment containing a diamond crystal (Bruker Optics Ltd, Coventry, UK). IR spectra were acquired from 10 independent locations per slide using a CCTV camera. Spectra (8 cm−1 spectral resolution giving 4 cm−1 data spacing equivalent to 235 wavenumbers, co-added for 32 scans) were converted into

J.G. Kelly et al. / Clinica Chimica Acta 411 (2010) 1027–1033

absorbance by Bruker OPUS software. The crystal was washed with distilled water prior to use and in between specimens. Spectra were cut and baseline-corrected over the 1800 cm−1–900 cm−1 region and normalised to 1650 cm−1, using Bruker OPUS software. 2.4. Computational analysis PCA is an unsupervised technique that reduces data dimensionality by forming linear combinations of variables, in this case wavenumbers, using the Pirouette software package (Infometrix Inc, Woodinville, WA, USA). LDA is a supervised technique employed after PCA to facilitate minimisation of intra-category variance and maximisation of inter-category variance. Scores plots provide visualisation of the data, whereby the spread of data is an indicator of biochemical similarity. Loadings plots allow for identification of wavenumbers and corresponding biochemicals responsible for any segregation. 3. Results 3.1. Exfoliative cervical cytology categorised by HPV type Initial observations of the spectra suggested that the 1155 cm−1– 960 cm−1 region most discriminated the different cytology grades (Fig. 1A). However, more subtle differences may exist and often computational analysis is required to extract these. Several PCA-LDA analyses were performed using the following as categories: grade of cytology, multiple infections, HPV type, HPV16 ± HPV18 infection and patient's age. Fig. 2 shows PCA-LDA scores plots categorising cytology based on infection with HPV type and presence/absence of infection. When cytology specimens positive for HPV16 and/or HPV18 were compared, some overlap between the HPV16 and HPV18 clusters was observed but, they also appear reasonably distinct (Fig. 2A). The cluster representing both HPV types overlaps more with the HPV16

1029

spectra. In atypia-free specimens, there appears to be little biochemical difference in the presence or absence of HPV (Fig. 2B). However, the number of HPV16-infected specimens with low-grade abnormalities facilitated a comparison based on patients' ages: 3rd decade vs. 4th decade women (Fig. 2C). Good separation of clusters was observed and loadings curves revealed Amide II (1543 cm−1) N carbohydrates (1153 cm − 1 ) N ν s PO − (1069 cm − 1 ) N Amide I 2 −1 −1 (1620 cm and 1655 cm ) N protein phosphorylation (972 cm−1) as the most important biochemical differences between the two categories (Fig. S1). 3.2. Comparing specimens infected with one HPV type Specimens infected with one HPV type only were classed into particular categories. In total this gave seven categories with more than one individual specimen per category: HPV16, HPV18, HPV51, HPV52, HPV56, HPV59 and HPV68. A PCA-LDA scores plot containing all seven clusters revealed some inter-category variation; of note, HPV18-, HPV52- and HPV68-infected categories appeared to segregate away from the rest (Fig. 2D). That these were the most distinct categories in this scores plot was confirmed from the loadings curves (Fig. S2). PCA-LDA was repeated with HPV16-, HPV51-, HPV56- and HPV59-infected cytology specimens in order to determine if intercategory distinguishing features had been diminished by larger variance associated with the HPV18, HPV52 or HPV68 categories (Fig. 2E). In this case, HPV59-infected specimens give rise to the most distinct cluster whereas clusters associated with HPV16, HPV51 and HPV56 overlap (Fig. 2E; Fig. S3). The loadings curves in this latter comparison highlighted in order of decreasing ascendancy protein phosphorylation (984 cm−1 ) N Amide III (1261 cm −1 ) N Amide I (1628 cm−1) N carbohydrate (1153 cm−1) N Amide II (1539 cm−1) as discriminating features for HPV59; protein phosphorylation (984 cm−1) N Amide II (1543 cm−1 and 1524 cm−1) N carbohydrate (1153 cm−1) N Amide I/Amide II trough (1594 cm−1) for HPV18; −1 carbohydrate (1153 cm−1) N νsPO− ) N 1191 cm−1 Npro2 (1069 cm −1 −1 tein phosphorylation (972 cm ) N 1042 cm N Amide I (1624 cm−1) for HPV52; and, Amide II (1543 cm−1) N Amide I (1620 cm−1) N protein phosphorylation (968 cm−1) N carbohydrate (1153 cm−1) for HPV68. 3.3. Low-grade cervical cytology

Fig. 1. (A) Average spectra from each cytology category of the specimens interrogated; normal HPV+, low-grade, high-grade and normal HPV−, as indicated. (B) Average spectra from each cell line; C33A, HeLa and SiHa, as indicated.

Using only specimens exhibiting low-grade abnormalities the data was split into three age categories: “3rd decade women”, “4th decade women” and “3rd + 4th decade women”. Separate PCA-LDA comparisons were made for each category based on number of HPV infections (one, two or three); good cluster segregation was noted for these three categories (Fig. 3A–C). However, cluster segregation improves when data for 3rd decade women alone is used and, even more so (albeit with smaller numbers) in 4th decade women compared to that noted in the scores plot for 3rd + 4th decade women. Loadings curves revealed that the important biochemical entities consistently responsible for segregation of 3rd + 4th decade women for 1HPV vs. 2HPV vs. 3HPV are protein phosphorylation and Amide III. For 3rd decade women, segregation was consistently associated with Amide III, carbohydrates, protein phosphorylation and Amide II. Distinguishing biochemical entities for 4th decade women were protein phosphorylation and proteins (1404 cm−1). Further comparisons are made using these age groups as categories. The PCA-LDA scores plots of HPV16 vs. HPV18 suggest that these do not modify the biochemistry of cytology from 3rd + 4th decade women; some cluster separation with 3rd decade women is noted but very distinct segregation with 4th decade women occurs (Fig. 3D,F,H). From the loadings plot, a comparison may be made between the biochemical entities important for each age group; while Amide III and carbohydrates distinguish these categories of 3rd

1030

J.G. Kelly et al. / Clinica Chimica Acta 411 (2010) 1027–1033

Fig. 2. PCA-LDA scores plots analysing specimens based on HPV status. A shows HPV16 vs. HPV18 vs. HPV16 + 18, from low– and high-grade specimens only. B is a comparison of specimens of normal cytology for HPV+ vs. HPV−. C shows low-grade patients infected with HPV16 only; the two categories are divided by age: 3rd (20–29 years) and 4th (30– 39 years) decade women. D and E incorporate specimens from all cytology grades that have tested positive for one type of HPV: HPV16, HPV18, HPV51, HPV52, HPV56, HPV59 and HPV 68. D has 7 categories, E has four.

decade women, they appear to be less relevant for 4th decade women. A similar pattern is revealed with 0 hrHPV vs. 1 hrHPV (Fig. 3E,G,I). The loadings curves show that Amide II (1545 cm−1) is important for both categories in 3rd + 4th decade women. Biochemical entities most distinguishing 3rd decade women include Amide III, lipids, Amide I (1632 cm−1) and νsPO− 2 while for 4th decade women they are Amide II (1505 cm−1), νasPO− 2 , protein phosphorylation and Amide I (1679 cm−1). 3.4. A biochemical fingerprint of cervical cell lines Average IR spectra from the cell lines examined were compared; SiHa appears to be the most distinct, especially in the Amide I and Amide II regions (Fig. 1B). This distinction is highlighted in the PCALDA analysis of C33A vs. HeLa vs. SiHa where the scores plot shows that LD1 separates SiHa from C33A and HeLa (Fig. 4A). Loadings from inter-category comparisons (C33A vs. HeLa, C33A vs. SiHa and HeLa vs. SiHa) highlight the features most important. These reveal that −1 carbohydrates (1155 cm−1) N νsPO− and 1092 cm−1) N 2 (1069 cm −1 −1 Amide II (1505 cm ) N glycogen (1022 cm ) N Amide I (1678 cm−1) in order of ranked importance segregate C33A from HeLa. For C33A vs. SiHa, lipids (1705 cm−1) N Amide II (1516 cm−1) N Amide I (1613 cm−1) N protein (1458 cm−1 and 1312 cm−1) N protein phosphorylation are the most distinguishing. Finally, a comparison of −1 HeLa vs. SiHa shows Amide I (1674 cm−1) N νsPO− )N 2 (1099 cm −1 protein (1497 cm−1 and 1469 cm−1) N νasPO− (1219 cm ) N Amide 2 II (1551 cm−1) segregate these cells lines. 4. Discussion ATR-FTIR spectroscopy has been explored as a novel methodology to biochemically signature exfoliative cervical cytology by HPV type; such spectral signatures have potential as a diagnostic tool [16]. A persistent HPV infection causes biological changes, which may

ultimately result in cervical cancer. Our hypothesis is that these changes can be detected by IR spectroscopy and identified by PCALDA. The processed spectra suggest some differences between cytology grades (Fig. 1A). However, more subtle differences may need to be extracted by multivariate analysis [17]. It was found that differences due to HPV infection type could only be detected in cases with verified CIN (low-grade and/or high-grade). Specifically, a distinction can be made between specimens infected with HPV16 or HPV18. Normal cytology for HPV− vs. HPV+ showed no separation which suggests they are biochemically similar (Fig. 2B). Initial investigations focused on cytology infected with individual HPV types; this revealed some differences between these two categories. Among specimens with low-grade abnormalities it was possible to distinguish those with one, two or three HPV types (Fig. 3A). A number of specimens infected with only one HPV type were compared (Fig. 2D,E). Although categories representing cytology infected with HPV18, HPV52 or HPV68 were distinct, overlap was observed particularly for clusters representing the HPV16, HPV51, HPV56 or HPV59 categories. Subsequent analysis using only these latter categories pointed to overlap remaining between clusters for HPV16, HPV52 and HPV56. The cluster for HPV59 was distinct, suggesting that some individual HPV types are detectable (Fig. 2E). However, these specimens are representative of a range of cytological grades and any of the differences detected could be due to cancer progression, as well as HPV-induced cancerous changes. All specimens that tested positive for HPV were hr, the genotype analysis then was a 2nd round of testing which identified other types including lrHPV (Fig. 2C,D). Focussing on categories such as age, number of HPV types and cytology enabled the analysis to be more precise. In particular, dividing low-grade cervical cytology by age produced the most distinguishing findings. The scores plots show an increased segregation based on age, particularly so for 0 hrHPV vs. 1 hrHPV; with the same cytology grading, these two groups could have very different

J.G. Kelly et al. / Clinica Chimica Acta 411 (2010) 1027–1033

1031

Fig. 3. All low-grade specimens classed by age groups 3rd + 4th decade women (A, D, and E), 3rd decade women (B, F, and G) and 4th decade women (C, H, and I). The presence of different numbers of HPV infection types was compared (A–C): 1 HPV vs. 2 HPV vs. 3 HPV. In D, F and H the affect of HPV16 is compared with HPV18. Finally, E, G and I show 0 hrHPV compared to 1 hrHPV.

prognoses. Distinction here between and within the age groups supports the concept that there is a fundamental biological difference between 3rd decade and 4th decade women with low-grade cervical cytology. In addition to the spread of data, a comparison of the biochemical entities that discriminate the age groups can be made. The important entities for HPV16 vs. HPV18 for 3rd decade women are Amide I, carbohydrates and Amide III and ∼ glycogen; while, for 4th decade women Amide II, protein phosphorylation, Amide I, glycogen and lipids appear important. A similar comparison for 0 hrHPV vs. 1 hrHPV shows Amide III, lipids, Amide II, Amide I and νsPO− 2 to be important for 3rd decade women, and Amide II, νasPO− 2 , protein phosphorylation and Amide I for 4th decade women. A direct comparison of the two age groups using low-grade patients with HPV16 indicates Amide II, carbohydrates, νsPO− 2 , Amide I and protein phosphorylation are the biochemical entities significantly discriminating the two categories. Distinctions between HPV16 and HPV18 infection could facilitate an extrapolation of this technique to identify other HPV types. Interrogation of a larger number of patients infected with one HPV type would first be required in order to identify a biochemical-cell fingerprint related to each type. Consistently, biochemicals specific to 3rd decade women are Amide I and Amide III, while protein phosphorylation is important for 4th decade women. Numerous stages are involved in the integration and replication of HPV DNA into cervical cells, leading to a persistent infection and possibly invasive disease. An important step is the phosphorylation of the E7 protein [18]. This phosphorylated E7 protein

has an affinity for the hypophosphorylated (active form) retinoblastoma tumour suppressor gene product (pRB; cell cycle control); once bound, proliferation is induced and cell cycle control is lost [10,19]. Additionally, E7 from low-risk HPV has a lower affinity for pRB than high-risk [19]. Although there are many processes which occur through the integration of HPV DNA, this particular step involving phosphorylation of proteins could be associated with a marker of persistent infection. Upon close inspection of the average spectra from each cytological grade, there is a decrease in absorbance at 968 cm−1 (protein phosphorylation) associated with cervical cancer progression (Fig. 1A). This confirms previous work indicating protein phosphorylation to be an important marker of disease progression [14]. The three cervical cell lines were very dissimilar from each other, with SiHa cells (HPV16-infected) being the most distinctive. This is not surprising as SiHa cells have fewer copies of integrated virus and are squamous cell-derived whereas HeLa cells are adenocarcinomaderived. This is observed in the scores plots of C33A vs. HeLa vs. SiHa, the two-category comparisons and, to some extent, the average spectra (Fig. 1B; Fig. 4). The PCA-LDA loadings curves of the three categories show νasPO− 2 N Amide I/Amide II trough N Amide II N carbohydrate distinguish C33A; Amide II N νsPO− 2 N glycogen N Amide I distinguish HeLa and SiHa cells. HeLa and SiHa cells had symmetrical loadings curves, implying they are very dissimilar. Important biochemical differences between the three cell lines can alternatively be sought by −1 pair-wise comparisons. Carbohydrates, νsPO− ), 2 , Amide II (1505 cm glycogen and Amide I (1678 cm−1) are responsible for the differences

1032

J.G. Kelly et al. / Clinica Chimica Acta 411 (2010) 1027–1033

Fig. 4. PCA-LDA scores and loadings plots of cell line comparisons. (A) Scores plot of C33A vs. HeLa vs. SiHa; B and C show the scores and loadings plots of C33A vs. HeLa; D and E show the scores and loadings plots of HeLa vs. SiHa and F and G show the scores and loadings plots of C33A vs. SiHa.

between C33A and HeLa; lipids, Amide II (1516 cm−1), Amide I (1613 cm−1), proteins and protein phosphorylation distinguish C33A from SiHa; and, comparison of HeLa vs. SiHa revealed Amide I −1 (1674 cm−1), 1099 cm−1, proteins, νasPO− ) 2 and Amide II (1551 cm are segregating. Vaccine efficacy is also dependent on the recipient being HPV negative at the time of vaccination [8]. Although vaccination offers protection against the more common oncogenic HPV types, 11 others remain. Furthermore, the level and duration of protection is largely unknown, necessitating a continuation of cervical smear tests. Biospectroscopy is increasingly being recognised for its potential to fingerprint different cell types [20,21], grade different cancers and have a role in the pathology laboratory of the future [22,23]. Application of ATR-FTIR spectroscopy as part of a cervical screening programme could enable earlier diagnosis and identification of patients likely to progress to high-grade lesions and cervical carcinoma. In conclusion, we found ATR-FTIR spectroscopy to be capable of distinguishing between hrHPV infections in women exhibiting pre-cancerous cervical cellular atypia. An underlying biochemical difference between 3rd and 4th decade women was confirmed and identified with specimens of the same cytological grade and HPV type. ATR-FTIR spectroscopy has the potential to identify patients likely to progress to invasive disease based on a persistent HPV infection.

Acknowledgement This project was sponsored by the Rosemere Cancer Foundation.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.cca.2010.03.029.

References [1] Bosch FX, Iftner T. The aetiology of cervical cancer. NHSCSH publication no. 22; 2005. [2] Collins S, Mazloomzadeh S, Winter H, et al. High incidence of cervical human papillomavirus infection in women during their first sexual relationship. BJOG 2002;109:96–8. [3] Wentzensen N, Klug SJ. Early detection of cervical carcinomas: finding an overall approach. Dtsch Arztebl Int 2008;105:617–22. [4] Stanley M. Immune responses to human papillomavirus. Vaccine 2006;24:S16–22. [5] van Hamont D, van Ham MAPC, Bakkers JMJE, Massuger LFAG, Melchers WJG. Evaluation of the SPF10-INNO LiPA human papillomavirus (HPV) genotyping test and the roche linear array HPV genotyping test. J Clin Microbiol 2006;44:3122–9. [6] Woodman CBJ, Collins SI, Young LS. The natural history of cervical HPV infection: unresolved issues. Nat Rev Cancer 2002;7:11–22. [7] Kyrgiou M, Shafi M. HPV vaccine. Obstet Gynaecol Reprod Med 2008;19:26–8. [8] Lepique AP, Rabachini T, Villa LL. HPV vaccination: the beginning of the end of cervical cancer? Mem Inst Oswaldo Cruz 2009;104:1–10. [9] Zur Hausen H. Papillomaviruses and cancer: from basic studies to clinical application. Nat Rev Cancer 2002;2:342–50. [10] Snijders PJF, Steenbergen RDM, Heideman DAM, Meijer CJLM. HPV-mediated cervical carcinogenesis: concepts and clinical implications. J Pathol 2006;208:152–64. [11] Stevens MP, Garland SM, Rudland E, Tan J, Quinn MA, Tabrizi SN. Comparison of the digene hybrid capture 2 assay and roche amplicor and linear array human papillomavirus (HPV) tests in detecting high-risk HPV genotypes in specimens from women with previous abnormal Pap smear results. J Clin Microbiol 2007;45:2130–7. [12] Stuart B. Infrared spectroscopy: fundamentals and applications1st ed. . Chichester, England: John Wiley and Sons Inc; 2005. [13] Walsh MJ, German MJ, Singh M, et al. IR microspectroscopy: potential applications in cervical cancer screening. Cancer Lett 2007;246:1–11. [14] Walsh MJ, Singh MN, Pollock HM, et al. ATR microspectroscopy with multivariate analysis segregates grades of exfoliative cervical cytology. Biochem Biophys Res Commun 2007;352:213–9. [15] Kelly JG, Angelov PP, Walsh MJ, et al. Intelligent interrogation of mid-IR spectroscopy data from exfoliative cervical cytology using self-learning classifier eClass. Int J Comput Intel Res 2008;4:392–401. [16] Wood BR, Chiriboga L, Yee H, Quinn MA, McNaughton D, Diem M. Fourier transform infrared (FTIR) spectral mapping of the cervical transformation zone, and dysplastic squamous epithelium. Gynecol Oncol 2004;93:59–68. [17] Martin FL, German MJ, Wit E, Fearn T, Ragavan N, Pollock HM. Identifying variables responsible for clustering in discriminant analysis of data from infrared microspectroscopy of a biological sample. J Comput Biol 2007;14:1176–84. [18] Smotkin D, Wettstein FO. The major human papillomavirus protein in cervical cancers is a cytoplasmic phosphoprotein. J Virol 1987;61:1686–9. [19] Münger K, Howley PM. Human papillomavirus immortalization and transformation functions. Virus Res 2002;89:213–28.

J.G. Kelly et al. / Clinica Chimica Acta 411 (2010) 1027–1033 [20] Walsh MJ, Fellous TG, Hammiche A, et al. Fourier transform infrared microspectroscopy identifies symmetric PO− 2 modifications as a marker of the putative stem cell region of human intestinal crypts. Stem Cells 2008;26:108–18. [21] Walsh MJ, Hammiche A, Fellous TG, et al. Tracking the cell hierarchy in the human intestine using biochemical signatures derived by mid-infrared microspectroscopy. Stem Cell Res 2009;3:15–27.

1033

[22] Baker MJ, Gazi E, Brown MD, Shanks JH, Clarke NW, Gardner P. Investigating FTIR based histopathology for the diagnosis of prostate cancer. J Biophotonics 2009;2: 104–13. [23] Pijanka J, Sockalingum GD, Kohler A, et al. Synchrotron-based FTIR spectra of stained single cells. Towards a clinical application in pathology. Lab Invest 2010;90:787–807; doi:10.1038/labinvest.2010.8.