JOURNAL OF CLINICAL MICROBIOLOGY, Mar. 1991,
p.
Vol. 29, No. 3
573-577
0095-1137/91/030573-05$02.00/0 Copyright C) 1991, American Society for Microbiology
Comparison of Southern Blot Hybridization and Polymerase Chain Reaction Methods for the Detection of Human Papillomavirus DNA MARK H. SCHIFFMAN,1* HEIDI M. BAUER,2 ATTILA T. LORINCZ,3t M. MICHELE MANOS 2 JANET C. BYRNE,4 ANDREW G. GLASS,S DIANE M. CADELL,6 AND PETER M. HOWLEY4 Environmental Epidemiology Branch, National Cancer Institute, Executive Plaza North,' and Laboratory of Tumor Virus Biology, National Cancer Institute,4 Bethesda, Maryland 20892; Department of Infectious Diseases, Cetus Corporation, Emeryville, California 946082; Corporate Research, Life Technologies, Inc., Gaithersburg, Maryland 208773; Kaiser Permanente Oncology Research, Portland, Oregon 972275; and Westat, Inc., Rockville, Maryland 208506 Received 13 August 1990/Accepted 13 December 1990
A methodologic study was performed to compare the polymerase chain reaction (PCR) and Southern blot hybridization, two commonly used testing strategies for the detection of human papillomavirus (HPV) infection. Three laboratories tested masked aliquots of exfoliated cervical cell specimens obtained from 120 women by cervicovaginal lavage. The study population included 32 women with condylomatous atypia or cervical intraepithelial neoplasia and 88 control women with no known history of cervical neoplasia. Two laboratories used PCR with different sets of consensus primers for HPV detection. The third laboratory used low-stringency Southern blot hybridization to identify all HPV types, followed by high-stringency Southern and/or dot blot hybridization to confirm specific HPV types. One of the PCR primer sets detected HPV types with a differential efficiency that was not predicted by analysis of DNA sequences or direct testing of HPV-containing plasmids. In contrast, the second PCR primer set was shown to be a much broader consensus system, detecting the same HPV types as Southern blotting, though requiring much less clinical specimen. Over 80% of women with cervical intraepithelial neoplasia or condylomatous atypia were found to be HPV infected both by Southern blotting and by the second PCR primer set. Among the control women, 11% were HPV positive by Southern blotting, while 31% were positive with the second set of primers. Most of the HPV infections found only by PCR were not due to HPV type 6, 11, 16, 18, 31, 33, or 45. These known HPV types were uncommon among normal women in the study population, even as determined by the PCR method.
specimen collection and processing could make PCR impractical for routine, widespread clinical use. The methodologic comparison presented here was designed to evaluate two different sets of PCR consensus primers for the detection of genital HPV DNA. In planning this comparison, Southern blot DNA-DNA hybridization was considered to be the current "gold standard" assay because of its well-defined sensitivity and because of the type specificity provided by combined hybridization and restriction enzyme analysis. Each of the three assay methods was performed in a separate laboratory with masked aliquots of the same 120 cervical cell samples.
The association of human papillomavirus (HPV) infection with cervical neoplasia has been firmly established by laboratory and epidemiologic investigations. The growing evidence suggesting that HPV infection is a causal factor in the development of cervical neoplasia provides an impetus to adapt HPV testing for gynecologic screening. Before HPV testing can be considered for routine use, the assays must be well validated, and accurate estimates of type-specific HPV prevalence among the various clinic populations targeted for screening must be obtained. Assay validation and HPV prevalence estimation will require, in turn, careful comparisons of available HPV testing methods. Several HPV testing strategies have been employed over the past few years (4). In particular, much attention has focused on the potential utility of polymerase chain reaction (PCR) methods to amplify HPV-specific DNA sequences (6). PCR promises to be much more sensitive than previously used methods of HPV DNA detection. Moreover, PCR consensus primer sets, which hybridize to highly conserved regions of the HPV genome, have been designed to detect many known HPV types in a single amplification procedure. However, some early reports have suggested that PCR might detect HPV in the majority of women, thereby limiting the predictive value of a positive test (2, 9, 10). Because of the extreme sensitivity of PCR methods, there is some related concern that contamination or carry-over during large-scale
MATERIALS AND METHODS Study population and collection of specimens. As part of an ongoing cohort study of HPV infection and cervical neoplasia, Papanicolaou (Pap) smears and additional cervical cell samples for HPV testing were obtained at enrollment from 23,000 women attending obstetrics or gynecology clinics at Kaiser Permanente in Portland, Oreg. For the present methodologic study, a subgroup of the first 1,000 participants was chosen so as to include two types of patients. A 32-woman case group was composed of all women with cytologic evidence of cervical intraepithelial neoplasia (CIN) or condylomatous atypia first detected at enrollment (n = 21) plus 11 women referred to a colposcopy clinic for previously detected CIN or condylomatous atypia. These 32 case women had a median age of 26.5 years, and most were white (88%) and parous (56%). Twenty-eight percent were married, and 41% were using oral contraceptives. One woman was pregnant.
* Corresponding author. t Present address: Digene Diagnostics, Inc., Silver Spring, MD
20904.
573
574
SCHIFFMAN ET AL.
An 88-woman randomly selected control group consisted of women with cervical cytologic diagnoses that were normal or showed only reactive atypia. Women with a medical history of cervical neoplasia were excluded from the control group. The current Pap smear from one woman in the control group was subsequently judged to be technically inadequate; her sample was excluded from statistical analyses related to cytologic diagnoses. No attempt was made to match the controls to the cases with regard to demographic or other characteristics. In fact, pregnant women and oral contraceptive users were oversampled in order to study the effects of these variables on HPV detection. However, no statistically significant effects of pregnancy or oral contraceptive use were observed in this small study group; thus, results for the entire control group were pooled for analysis and presentation. The control women had a median age of 30 years, and, like the case women, most were white (89%) and parous (56%). Unlike the case women, most (60%) of the control women were married, and they also tended to be better educated and wealthier than the case women. Because of oversampling, more of the control women were pregnant (22% [all firsttrimester pregnancies]) and more were using oral contraceptives (26%) than in the Portland disease-free cohort from which the control group was chosen (7% first-trimester pregnancies and 19% using oral contraceptives). The cervical cell samples used for HPV testing were obtained as follows. For each woman, a 10-ml saline lavage of the cervix was performed, and the sample obtained was transported to the processing laboratory on wet ice. One milliliter of the lavage specimen was withdrawn and frozen. The remaining 9 ml was centrifuged to produce a dry cell pellet, which was resuspended in saline and divided to form two dry cell pellets, each derived from 4.5 ml of lavage specimen. Southern blot analyses were performed with one of the cell pellets, and the first PCR consensus primer set (PCR 1) was applied to the other identical pellet. Subsequently, to evaluate a second PCR consensus primer set (PCR 2), the reserve 1-ml lavage specimen was used. The PCR methods each used such small fractions of the specimens that the difference in DNA content between the different aliquots (cell pellet versus lavage) was thought not to be an important variable. PCR amplification methods. As shown in Results, PCR 1 amplified different HPV types with differential efficiency; thus, this system was abandoned. In the presentation of methods, therefore, the methodologic details of PCR 2 will be emphasized. The samples were digested with proteinase K prior to amplification. Negative and positive control samples were interspersed and carried through the complete testing process, including PCR amplification and hybridization analyses. For PCR 2, two types of negative controls were used: 1 ml of saline was spiked with 106 K-562 cells (ATCC CCL 243) to mimic a lavage specimen containing human DNA but no HPV, and amplification reactions with no added DNA were also performed. One of each type of negative control was included with each batch of 40 test samples. None of the negative control samples in the present investigation were judged positive. Pre- and postamplification reagents were kept physically separated throughout the experiments to avoid contamination. Positive controls included HPV DNA from SiHa and HeLa cells. The PCR 1 method employed 32 cycles of 1 min at 94°C, 2 min at 48°C, and 2 min at 70°C. PCR 2 used 35 cycles of 50 s at 95°C, 50 s at 55°C, and 1 min at 72°C.
J. CLIN. MICROBIOL.
The alignments of the two PCR primer sets with complementary sequences of several HPV types are shown in Table 1. Both primer sets were designed to amplify fragments of about 450 bp from the Li region of the HPV genome. PCR 1 uses 21-mer primers, while PCR 2, known commonly as the Manos laboratory Li primer set (7), uses 20-mer primers. PCR 2 primers have a larger number of degenerate positions than PCR 1 primers have, and PCR 2 primers are less AT rich. Preliminary experiments with plasmid-derived DNA suggested that both PCR systems could amplify a wide variety of dermal and genital HPV types. PCR 1 successfully amplified all HPV types tested, including types 1A, 2, 6, 11, 13, 16, 18, 31, 33, 35, 39, 42, 43, 44, 45, 51, and 52. PCR 2 similarly amplified types 5, 6, 8, 11, 16, 18, 26, 27, 30, 31, 33, 35, 39, 40, 41, 42, 43, 45, 47, 48, 51, 52, 53, 54, 55, 57, 58, and 59. HPV detection and typing after PCR amplification. For both PCR systems, HPV detection was achieved initially by visualizing the PCR products by gel electrophoresis and ethidium bromide staining. The detection of a band of the expected 450-bp size indicated a positive sample. For PCR 2, a generic HPV probe was synthesized by amplifying Li PCR fragments from HPV types 16, 18, and 31 and an unidentified HPV type which is divergent, by sequence analysis, from the other three types. This generic probe is described in detail elsewhere (1). The samples were further analyzed with type-specific oligonucleotide probes for HPV types 6, 11, 16, 18, 31, 33, and 45. Samples testing positive with the generic probe only were classified as "positive, unidentified type." A few additional samples were classified as "positive, unidentified type" on the basis of convincing ethidium bromide banding, despite negative hybridizations. Further technical details regarding the typing strategy used have been reported
previously (7). Southern blot methods. The Southern blotting was performed by standard methods (5). Briefly, DNA was extracted from the cell pellets by proteinase K-sodium dodecyl sulfate digestion followed by phenol-chloroform purification. Approximately 5 ,ug of DNA per specimen was digested with PstI and electrophoresed on agarose gels. Southern blots were prepared on Hybond nylon filters and probed with HPV types 11, 16, and 18 at low stringency to detect all HPV types. The same filters and/or dot blots were subsequently reprobed under conditions of high stringency with individual HPV probes for types 6/11, 16, 18, 31, 33, 35, 42, 43, 44, 45, 51, 52, and 56 to identify specific HPV types. HPV-positive specimens not reacting with these probes at high stringency were called "positive, unidentified type." Statistical methods. The Southern blot hybridization method was considered the "gold standard"; thus, each PCR method was compared to Southern blotting. McNemar's statistic was calculated for each paired comparison. For unpaired comparisons, the standard chi-square statistic was used. RESULTS The HPV detection results by cytologic diagnosis are shown for each test method in Table 2. By all three methods, samples from women with condylomatous atypia or CIN were significantly more likely to be positive than samples from women with normal cytologic diagnoses or reactive atypia (P < 0.01 for all three methods). This case-control difference could not be explained by taking into account any of the demographic differences between the two groups.
VOL. 29,
PCR VERSUS SOUTHERN BLOT FOR HPV TESTING
1991
575
TABLE 1. PCR primer sets used in the comparison, aligned with known HPV types Primer
Sequence'
5' nucleotide
JCB1 (21-mer) HPV 6b HPV 11 HPV 16 HPV 18 HPV 31 HPV 33
5'CTAYWTGCAAATATCCAGATT ---CA-T -------------GTC---------T------TT ----------------TT-T-------T------TT-T -----------G--CA--------------
6456 6441 6316 6292 6234 6273
JCB2 (21-mer) HPV 6b
5'TGAAAAAYAAA CTGTARAT CA
Ll consensus primer method
PCR 1 Positive strand
Negative strand
HPV HPV HPV HPV HPV
PCR 2 Positive strand
Negative strand
11
------T---T--- A---------T-------A----
16 18 31 33
------T-------A---------T-----C-A------T--T---T---A---------C-------G----
MY11 (20-mer) HPV 6b HPV 11 HPV 16 HPV 18 HPV 31 HPV 33
5'GCMCAGGGWCATAAYAATGG
MY09 (20-mer) HPV 6b HPV 11 HPV 16 HPV 18 HPV 31 HPV 33
5'CGTCCMARRGGAWACTGATC ----C-AA---T---------A-GG---A---------T-AA---A---------A-GG---T-T-----A-C-GT---A------
6903 6888 6766 6745 6684 6720
6722 6707 6584 6558 6500 6539
--C----A----C--T----A----C--A----C- T---T --A----T----C- -C --T----A--C- T ------A--A-T---- T---
7171 7155 7035 7012 6951 6988
----C-AA---A------
Degenerate code: M, A or C; R, A or G; W, A or T; Y, C or T. Degenerate positions in primers and mismatches between primers and HPV sequence are indicated in boldface type. Dashes indicate exact nucleotide matches.
PCR 2 found more samples positive than Southern blotting, while PCR 1 found the fewest samples positive. The larger number of samples that were HPV positive by PCR 2 compared with Southern blot hybridization was most evident among control women, because both methods detected HPV in the majority of women with condylomatous atypia or CIN. (One of the four samples from women with CIN 2/3 was judged inadequate for Southern blotting but, for this comparison, was grouped as HPV negative.) In contrast, PCR 1 detected fewer HPV infections in women with or without cytologic abnormalities than either of the other two methods. Table 3 highlights the difference in HPV detection between PCR 1 and Southern blotting. All samples considered TABLE 2. HPV detection by cytologic diagnosis using three different methods No. (%) of specimens positive for
Cytologic diagnosis
Southern blot
Normal (n = 68) Reactive atypia (n = 19) Condylomatous atypia (n = 20) CIN 1 (n = 8) CIN 2/3 (n = 4)
8 2 17 7 2
(12) (11) (85) (88) (50)
HPV by: PCR 1
3 (4) 0 (0) 7 (35) 6 (75) 2 (50)
HPV positive by PCR 1 were confirmed by Southern blotting, but PCR 1 detected HPV in only 18 (50%) of the 36 samples that were considered positive by Southern blotting (P < 0.001 by McNemar's test). The HPV types missed by PCR 1 were not random (Table 4). PCR 1 detected HPV in all of the 15 samples classified by Southern blotting as HPV type 16, 31, 33, or 35. In contrast, PCR 1 failed to detect HPV in 18 (86%) of the 21 samples with other HPV types identified by Southern blotting. This selectivity in detection was highly significant (P < 0.001). PCR 2 detected HPV in all 36 samples considered positive by Southern blotting and in an additional 20 samples as well (P < 0.001 by McNemar's test; Table 5). Agreement in HPV typing was good for those 36 samples considered positive by both methods. In the subset of 24 samples found by both
TABLE 3. HPV detection by PCR 1 and Southern blottinga No. of PCR 1 results
Southern blot result PCR 2
20 7 17 8 4
(29) (37) (85) (100) (100)
+
Total
84 18
102
a Statistical significance: McNemar's chi-square
=
+
Total
0 18
84 36
18
120
18.0 (1 df), P < 0.001.
576
J. CLIN. MICROBIOL.
SCHIFFMAN ET AL.
TABLE 4. HPV types detected or missed by PCR 1 for the 36 samples classified as HPV positive by Southern blot hybridizationa
TABLE 6. HPV types in samples considered positive by PCR 2, according to Southern blot result' No. of samples:
No. of samples:
HPV
HPV typeb
16, 31, 33, All other'
or
35
Detected by PCR 1
Missed by PCR 1
15 3
0 18
18 18
4 16
=
b
Southern blotting and PCR 2 to contain a single HPV type, there was only one definite typing disagreement (taking into account the different sets of type-specific probes available in the two laboratories). Among the 12 samples found by at least one method to contain multiple HPV types, the two techniques tended to agree on at least one of the types present but to disagree on the total number of types (data not shown). The 20 additional positive samples detected by PCR 2 alone were confirmed by repeat testing. As shown in Table 6, only 4 (20%) of the 20 samples contained HPV type 6/11, 16, 18, 31, 33, or 45. Most of the additional positives were HPV types for which type-specific oligonucleotides were not available. As mentioned above with reference to Table 2, these unidentified types were almost all from women with normal cytologic diagnoses or reactive atypia. HPV types commonly associated with high-grade cervical neoplasia (types 16, 18, 31, and 33, in particular) were uncommon among the 87 control women, according to the results of either Southern blotting (prevalence, 1%) or PCR 2 (prevalence, 3%). Among the 32 case women with condylomatous atypia or CIN, these types were found in 44% by Southern blot and in 50% by PCR 2.
DISCUSSION The results of this intermethod comparison demonstrate clearly that PCR primer performance cannot be predicted adequately by sequence homology or plasmid DNA experiments. PCR 1 appeared, by such preliminary criteria, to employ a consensus primer set capable of amplifying many known HPV types. In practice, however, the primer set amplified selected HPV types only. The reasons for this selective performance are unknown but may include the high AT nucleotide content of the primers. We conclude that optimization with clinical specimens is required before a primer set can be termed consensus, i.e., capable of detecting a broad range of HPV types. TABLE 5. HPV detection by PCR 2 and Southern blot' No. of PCR 2 results Southern blot result
a
andeStheted
6/11, 16, 18, 31, 33, or 45 Other
Statistical significance: chi-square 25.7 (1 df), P < 0.001. Typed according to Southern blot result. There were seven multiple infections, five of which were classified as "16, 31, 33, or 35" because at least one type was in that group. c Types 42, 43, 45, 51, 52, and 56 and unknown types. By Southern blotting there were no infections with type 6/11, 18, or 44. a
Positive by PCR 2 Detected by PCR 2 Pand negative blot blot and Southern by Southern blot
typeb
+
64 0
Total
64
Statistical significance: McNemar's chi-square
=
+
Total
20 36
84 36
56
120
20.0 (1 df), P < 0.001.
a
Statistical significance:
chi-square = 4.9 (1 df), P = 0.03.
b Typed according to PCR 2. There were nine
multiple infections, seven of which were called "type 6/11, 16, 18, 31, 33, or 45" because at least one type was in that group.
The comparison of PCR 2 with Southern blotting showed that PCR 2 used a true consensus primer set, apparently amplifying all HPV types detectable by low-stringency Southern blotting (at least those types represented in the study population). In addition, PCR 2 detected HPV in many women whose cervical cell specimens were negative by Southern blotting. This increase in sensitivity resulted mainly from the additional detection of unidentified HPV types (other than types 6, 11, 16, 18, 31, 33, and 45) among women without apparent cervical neoplasia. In contrast, the increased sensitivity of PCR 2 compared with Southern blot hybridization was hardly evident among women with condylomatous atypia or CIN because both methods detected HPV in virtually all cases. The cervicovaginal lavage used in this study to collect cervical cell specimens provides ample DNA, permitting optimal sensitivity of the Southern blot method (8). The additional sensitivity of PCR might be more apparent in studies relying on less ample cervical scrapes or swabs because the PCR method required a specimen less than 1/100 the size of the clinical specimen used by Southern blot to obtain comparable results. A remaining question is the clinical significance of the HPV infections of unidentified type found by PCR 2 alone. The HPV-positive women with such infections had mainly normal diagnoses or had reactive atypia. Some of the types considered unidentified by PCR 2 could be known or notyet-typed genital HPVs for which oligonucleotide probes were not available, and, alternatively, some of the apparent infections could be false-positive results. Further hybridization analyses with new probes, as well as cloning and sequencing efforts under way, should clarify this issue. When either Southern blotting or PCR 2 was used, HPV types commonly associated with high-grade cervical neoplasia (types 16, 18, 31, and 33, in particular) were found in less than 10% of our control population but in nearly 50% of the case population. The low prevalence of these particular HPV types among cytologically normal women in our study group suggests that type-specific HPV screening might serve as a useful predictor of undetected or incipient cervical neoplasia in this population. An ongoing cohort study of the same Kaiser Permanente patient population is designed to investigate the utility of such screening. In this context, it is important to note that the control women in our study can be considered a population at low risk for HPV infection. Their median age was 30, most were married, and we excluded by chart review all women with a medical history of cervical neoplasia. From other studies, it appears that HPV prevalence peaks in women with multiple sexual partners in their early to mid-twenties and then declines with increasing age (3). Thus, studies of younger, unmarried female populations
VOL. 29, 1991
(1) might be expected to yield higher HPV prevalences among cytologically normal women than we observed, and the predictive value of a positive HPV test would decline as a result. In conclusion, we have shown that a carefully optimized PCR method compares favorably with Southern blot hybridization and shows promise for HPV DNA detection in clinical specimens. As sensitive PCR assays are compared and validated, it will be important to reach a scientific consensus regarding two important issues that remain controversial: the prevalence of different types of HPV in women with normal and abnormal cytologic diagnoses and the corresponding prognostic importance of HPV DNA detection. REFERENCES 1. Bauer, H. M., Y. Ting, C. E. Greer, J. C. Chambers, C. J. Tashiro, J. Chimera, A. Reingold, and M. M. Manos. 1991. Genital human papillomavirus infection in female university students as determined by a PCR-based method. JAMA 265: 472-477. 2. Chow, V., K. M. Tham, M. Yeo-Gloss, S. K. Lim-Tam, I. Sng, T. Thirumoorthy, and H. Bernard. 1990. Molecular diagnosis of genital HPV DNA types by polymerase chain reaction and sensitivity-standardized filter in situ hybridization in randomly sampled cohorts of Singapore women. Mol. Cell. Probes 4:121131. 3. Ley, C., H. M. Bauer, A. Reingold, M. H. Schiffman, J. C.
PCR VERSUS SOUTHERN BLOT FOR HPV TESTING
4. 5.
6.
7.
8.
9.
10.
577
Chambers, C. J. Tashiro, and M. M. Manos. Submitted for publication. Lorincz, A. T. 1989. Human papillomavirus detection tests, p. 953-959. In K. K. Holmes et al. (ed.), Sexually transmitted diseases, 2nd ed. McGraw-Hill Book Co., New York. Lorincz, A. T., W. D. Lancaster, and G. F. Temple. 1986. Cloning and characterization of the DNA of a new human papillomavirus from a woman with dysplasia of the uterine cervix. J. Virol. 58:225-229. Manos, M. M., Y. Ting, D. K. Wright, A. J. Lewis, and T. R. Broker. 1989. Use of polymerase chain reaction amplification for the detection of genital human papillomaviruses. Cancer Cells 7:209-214. Ting, Y., and M. M. Manos. 1990. Detection and typing of genital human papillomaviruses, p. 356-367. In M. Innis et al. (ed.), PCR protocols: a guide to methods and applications. Academic Press, Inc., San Diego, Calif. Vermund, S. H., M. H. Schiffman, G. Goldberg, D. B. Ritter, A. Weltman, and R. D. Burk. 1989. Molecular diagnosis of genital human papillomavirus infection: comparison of two methods used to collect exfoliated cervical cells in the genital tract. Am. J. Obstet. Gynecol. 160:304-308. Ward, P., G. N. Parry, R. Yule, D. V. Coleman, and A. D. B. Malcolm. 1989. Human papillomavirus subtype 16a. Lancet ii:170. Young, L. S., I. S. Bevan, M. A. Johnson, P. I. Blomfield, T. Bromidge, N. J. Maitland, and C. B. J. Woodman. 1989. The polymerase chain reaction: a new epidemiological tool for investigating cervical human papillomavirus infection. Br. Med. J. 298:14-17.
p.
Vol. 29, No. 3
573-577
0095-1137/91/030573-05$02.00/0 Copyright C) 1991, American Society for Microbiology
Comparison of Southern Blot Hybridization and Polymerase Chain Reaction Methods for the Detection of Human Papillomavirus DNA MARK H. SCHIFFMAN,1* HEIDI M. BAUER,2 ATTILA T. LORINCZ,3t M. MICHELE MANOS 2 JANET C. BYRNE,4 ANDREW G. GLASS,S DIANE M. CADELL,6 AND PETER M. HOWLEY4 Environmental Epidemiology Branch, National Cancer Institute, Executive Plaza North,' and Laboratory of Tumor Virus Biology, National Cancer Institute,4 Bethesda, Maryland 20892; Department of Infectious Diseases, Cetus Corporation, Emeryville, California 946082; Corporate Research, Life Technologies, Inc., Gaithersburg, Maryland 208773; Kaiser Permanente Oncology Research, Portland, Oregon 972275; and Westat, Inc., Rockville, Maryland 208506 Received 13 August 1990/Accepted 13 December 1990
A methodologic study was performed to compare the polymerase chain reaction (PCR) and Southern blot hybridization, two commonly used testing strategies for the detection of human papillomavirus (HPV) infection. Three laboratories tested masked aliquots of exfoliated cervical cell specimens obtained from 120 women by cervicovaginal lavage. The study population included 32 women with condylomatous atypia or cervical intraepithelial neoplasia and 88 control women with no known history of cervical neoplasia. Two laboratories used PCR with different sets of consensus primers for HPV detection. The third laboratory used low-stringency Southern blot hybridization to identify all HPV types, followed by high-stringency Southern and/or dot blot hybridization to confirm specific HPV types. One of the PCR primer sets detected HPV types with a differential efficiency that was not predicted by analysis of DNA sequences or direct testing of HPV-containing plasmids. In contrast, the second PCR primer set was shown to be a much broader consensus system, detecting the same HPV types as Southern blotting, though requiring much less clinical specimen. Over 80% of women with cervical intraepithelial neoplasia or condylomatous atypia were found to be HPV infected both by Southern blotting and by the second PCR primer set. Among the control women, 11% were HPV positive by Southern blotting, while 31% were positive with the second set of primers. Most of the HPV infections found only by PCR were not due to HPV type 6, 11, 16, 18, 31, 33, or 45. These known HPV types were uncommon among normal women in the study population, even as determined by the PCR method.
specimen collection and processing could make PCR impractical for routine, widespread clinical use. The methodologic comparison presented here was designed to evaluate two different sets of PCR consensus primers for the detection of genital HPV DNA. In planning this comparison, Southern blot DNA-DNA hybridization was considered to be the current "gold standard" assay because of its well-defined sensitivity and because of the type specificity provided by combined hybridization and restriction enzyme analysis. Each of the three assay methods was performed in a separate laboratory with masked aliquots of the same 120 cervical cell samples.
The association of human papillomavirus (HPV) infection with cervical neoplasia has been firmly established by laboratory and epidemiologic investigations. The growing evidence suggesting that HPV infection is a causal factor in the development of cervical neoplasia provides an impetus to adapt HPV testing for gynecologic screening. Before HPV testing can be considered for routine use, the assays must be well validated, and accurate estimates of type-specific HPV prevalence among the various clinic populations targeted for screening must be obtained. Assay validation and HPV prevalence estimation will require, in turn, careful comparisons of available HPV testing methods. Several HPV testing strategies have been employed over the past few years (4). In particular, much attention has focused on the potential utility of polymerase chain reaction (PCR) methods to amplify HPV-specific DNA sequences (6). PCR promises to be much more sensitive than previously used methods of HPV DNA detection. Moreover, PCR consensus primer sets, which hybridize to highly conserved regions of the HPV genome, have been designed to detect many known HPV types in a single amplification procedure. However, some early reports have suggested that PCR might detect HPV in the majority of women, thereby limiting the predictive value of a positive test (2, 9, 10). Because of the extreme sensitivity of PCR methods, there is some related concern that contamination or carry-over during large-scale
MATERIALS AND METHODS Study population and collection of specimens. As part of an ongoing cohort study of HPV infection and cervical neoplasia, Papanicolaou (Pap) smears and additional cervical cell samples for HPV testing were obtained at enrollment from 23,000 women attending obstetrics or gynecology clinics at Kaiser Permanente in Portland, Oreg. For the present methodologic study, a subgroup of the first 1,000 participants was chosen so as to include two types of patients. A 32-woman case group was composed of all women with cytologic evidence of cervical intraepithelial neoplasia (CIN) or condylomatous atypia first detected at enrollment (n = 21) plus 11 women referred to a colposcopy clinic for previously detected CIN or condylomatous atypia. These 32 case women had a median age of 26.5 years, and most were white (88%) and parous (56%). Twenty-eight percent were married, and 41% were using oral contraceptives. One woman was pregnant.
* Corresponding author. t Present address: Digene Diagnostics, Inc., Silver Spring, MD
20904.
573
574
SCHIFFMAN ET AL.
An 88-woman randomly selected control group consisted of women with cervical cytologic diagnoses that were normal or showed only reactive atypia. Women with a medical history of cervical neoplasia were excluded from the control group. The current Pap smear from one woman in the control group was subsequently judged to be technically inadequate; her sample was excluded from statistical analyses related to cytologic diagnoses. No attempt was made to match the controls to the cases with regard to demographic or other characteristics. In fact, pregnant women and oral contraceptive users were oversampled in order to study the effects of these variables on HPV detection. However, no statistically significant effects of pregnancy or oral contraceptive use were observed in this small study group; thus, results for the entire control group were pooled for analysis and presentation. The control women had a median age of 30 years, and, like the case women, most were white (89%) and parous (56%). Unlike the case women, most (60%) of the control women were married, and they also tended to be better educated and wealthier than the case women. Because of oversampling, more of the control women were pregnant (22% [all firsttrimester pregnancies]) and more were using oral contraceptives (26%) than in the Portland disease-free cohort from which the control group was chosen (7% first-trimester pregnancies and 19% using oral contraceptives). The cervical cell samples used for HPV testing were obtained as follows. For each woman, a 10-ml saline lavage of the cervix was performed, and the sample obtained was transported to the processing laboratory on wet ice. One milliliter of the lavage specimen was withdrawn and frozen. The remaining 9 ml was centrifuged to produce a dry cell pellet, which was resuspended in saline and divided to form two dry cell pellets, each derived from 4.5 ml of lavage specimen. Southern blot analyses were performed with one of the cell pellets, and the first PCR consensus primer set (PCR 1) was applied to the other identical pellet. Subsequently, to evaluate a second PCR consensus primer set (PCR 2), the reserve 1-ml lavage specimen was used. The PCR methods each used such small fractions of the specimens that the difference in DNA content between the different aliquots (cell pellet versus lavage) was thought not to be an important variable. PCR amplification methods. As shown in Results, PCR 1 amplified different HPV types with differential efficiency; thus, this system was abandoned. In the presentation of methods, therefore, the methodologic details of PCR 2 will be emphasized. The samples were digested with proteinase K prior to amplification. Negative and positive control samples were interspersed and carried through the complete testing process, including PCR amplification and hybridization analyses. For PCR 2, two types of negative controls were used: 1 ml of saline was spiked with 106 K-562 cells (ATCC CCL 243) to mimic a lavage specimen containing human DNA but no HPV, and amplification reactions with no added DNA were also performed. One of each type of negative control was included with each batch of 40 test samples. None of the negative control samples in the present investigation were judged positive. Pre- and postamplification reagents were kept physically separated throughout the experiments to avoid contamination. Positive controls included HPV DNA from SiHa and HeLa cells. The PCR 1 method employed 32 cycles of 1 min at 94°C, 2 min at 48°C, and 2 min at 70°C. PCR 2 used 35 cycles of 50 s at 95°C, 50 s at 55°C, and 1 min at 72°C.
J. CLIN. MICROBIOL.
The alignments of the two PCR primer sets with complementary sequences of several HPV types are shown in Table 1. Both primer sets were designed to amplify fragments of about 450 bp from the Li region of the HPV genome. PCR 1 uses 21-mer primers, while PCR 2, known commonly as the Manos laboratory Li primer set (7), uses 20-mer primers. PCR 2 primers have a larger number of degenerate positions than PCR 1 primers have, and PCR 2 primers are less AT rich. Preliminary experiments with plasmid-derived DNA suggested that both PCR systems could amplify a wide variety of dermal and genital HPV types. PCR 1 successfully amplified all HPV types tested, including types 1A, 2, 6, 11, 13, 16, 18, 31, 33, 35, 39, 42, 43, 44, 45, 51, and 52. PCR 2 similarly amplified types 5, 6, 8, 11, 16, 18, 26, 27, 30, 31, 33, 35, 39, 40, 41, 42, 43, 45, 47, 48, 51, 52, 53, 54, 55, 57, 58, and 59. HPV detection and typing after PCR amplification. For both PCR systems, HPV detection was achieved initially by visualizing the PCR products by gel electrophoresis and ethidium bromide staining. The detection of a band of the expected 450-bp size indicated a positive sample. For PCR 2, a generic HPV probe was synthesized by amplifying Li PCR fragments from HPV types 16, 18, and 31 and an unidentified HPV type which is divergent, by sequence analysis, from the other three types. This generic probe is described in detail elsewhere (1). The samples were further analyzed with type-specific oligonucleotide probes for HPV types 6, 11, 16, 18, 31, 33, and 45. Samples testing positive with the generic probe only were classified as "positive, unidentified type." A few additional samples were classified as "positive, unidentified type" on the basis of convincing ethidium bromide banding, despite negative hybridizations. Further technical details regarding the typing strategy used have been reported
previously (7). Southern blot methods. The Southern blotting was performed by standard methods (5). Briefly, DNA was extracted from the cell pellets by proteinase K-sodium dodecyl sulfate digestion followed by phenol-chloroform purification. Approximately 5 ,ug of DNA per specimen was digested with PstI and electrophoresed on agarose gels. Southern blots were prepared on Hybond nylon filters and probed with HPV types 11, 16, and 18 at low stringency to detect all HPV types. The same filters and/or dot blots were subsequently reprobed under conditions of high stringency with individual HPV probes for types 6/11, 16, 18, 31, 33, 35, 42, 43, 44, 45, 51, 52, and 56 to identify specific HPV types. HPV-positive specimens not reacting with these probes at high stringency were called "positive, unidentified type." Statistical methods. The Southern blot hybridization method was considered the "gold standard"; thus, each PCR method was compared to Southern blotting. McNemar's statistic was calculated for each paired comparison. For unpaired comparisons, the standard chi-square statistic was used. RESULTS The HPV detection results by cytologic diagnosis are shown for each test method in Table 2. By all three methods, samples from women with condylomatous atypia or CIN were significantly more likely to be positive than samples from women with normal cytologic diagnoses or reactive atypia (P < 0.01 for all three methods). This case-control difference could not be explained by taking into account any of the demographic differences between the two groups.
VOL. 29,
PCR VERSUS SOUTHERN BLOT FOR HPV TESTING
1991
575
TABLE 1. PCR primer sets used in the comparison, aligned with known HPV types Primer
Sequence'
5' nucleotide
JCB1 (21-mer) HPV 6b HPV 11 HPV 16 HPV 18 HPV 31 HPV 33
5'CTAYWTGCAAATATCCAGATT ---CA-T -------------GTC---------T------TT ----------------TT-T-------T------TT-T -----------G--CA--------------
6456 6441 6316 6292 6234 6273
JCB2 (21-mer) HPV 6b
5'TGAAAAAYAAA CTGTARAT CA
Ll consensus primer method
PCR 1 Positive strand
Negative strand
HPV HPV HPV HPV HPV
PCR 2 Positive strand
Negative strand
11
------T---T--- A---------T-------A----
16 18 31 33
------T-------A---------T-----C-A------T--T---T---A---------C-------G----
MY11 (20-mer) HPV 6b HPV 11 HPV 16 HPV 18 HPV 31 HPV 33
5'GCMCAGGGWCATAAYAATGG
MY09 (20-mer) HPV 6b HPV 11 HPV 16 HPV 18 HPV 31 HPV 33
5'CGTCCMARRGGAWACTGATC ----C-AA---T---------A-GG---A---------T-AA---A---------A-GG---T-T-----A-C-GT---A------
6903 6888 6766 6745 6684 6720
6722 6707 6584 6558 6500 6539
--C----A----C--T----A----C--A----C- T---T --A----T----C- -C --T----A--C- T ------A--A-T---- T---
7171 7155 7035 7012 6951 6988
----C-AA---A------
Degenerate code: M, A or C; R, A or G; W, A or T; Y, C or T. Degenerate positions in primers and mismatches between primers and HPV sequence are indicated in boldface type. Dashes indicate exact nucleotide matches.
PCR 2 found more samples positive than Southern blotting, while PCR 1 found the fewest samples positive. The larger number of samples that were HPV positive by PCR 2 compared with Southern blot hybridization was most evident among control women, because both methods detected HPV in the majority of women with condylomatous atypia or CIN. (One of the four samples from women with CIN 2/3 was judged inadequate for Southern blotting but, for this comparison, was grouped as HPV negative.) In contrast, PCR 1 detected fewer HPV infections in women with or without cytologic abnormalities than either of the other two methods. Table 3 highlights the difference in HPV detection between PCR 1 and Southern blotting. All samples considered TABLE 2. HPV detection by cytologic diagnosis using three different methods No. (%) of specimens positive for
Cytologic diagnosis
Southern blot
Normal (n = 68) Reactive atypia (n = 19) Condylomatous atypia (n = 20) CIN 1 (n = 8) CIN 2/3 (n = 4)
8 2 17 7 2
(12) (11) (85) (88) (50)
HPV by: PCR 1
3 (4) 0 (0) 7 (35) 6 (75) 2 (50)
HPV positive by PCR 1 were confirmed by Southern blotting, but PCR 1 detected HPV in only 18 (50%) of the 36 samples that were considered positive by Southern blotting (P < 0.001 by McNemar's test). The HPV types missed by PCR 1 were not random (Table 4). PCR 1 detected HPV in all of the 15 samples classified by Southern blotting as HPV type 16, 31, 33, or 35. In contrast, PCR 1 failed to detect HPV in 18 (86%) of the 21 samples with other HPV types identified by Southern blotting. This selectivity in detection was highly significant (P < 0.001). PCR 2 detected HPV in all 36 samples considered positive by Southern blotting and in an additional 20 samples as well (P < 0.001 by McNemar's test; Table 5). Agreement in HPV typing was good for those 36 samples considered positive by both methods. In the subset of 24 samples found by both
TABLE 3. HPV detection by PCR 1 and Southern blottinga No. of PCR 1 results
Southern blot result PCR 2
20 7 17 8 4
(29) (37) (85) (100) (100)
+
Total
84 18
102
a Statistical significance: McNemar's chi-square
=
+
Total
0 18
84 36
18
120
18.0 (1 df), P < 0.001.
576
J. CLIN. MICROBIOL.
SCHIFFMAN ET AL.
TABLE 4. HPV types detected or missed by PCR 1 for the 36 samples classified as HPV positive by Southern blot hybridizationa
TABLE 6. HPV types in samples considered positive by PCR 2, according to Southern blot result' No. of samples:
No. of samples:
HPV
HPV typeb
16, 31, 33, All other'
or
35
Detected by PCR 1
Missed by PCR 1
15 3
0 18
18 18
4 16
=
b
Southern blotting and PCR 2 to contain a single HPV type, there was only one definite typing disagreement (taking into account the different sets of type-specific probes available in the two laboratories). Among the 12 samples found by at least one method to contain multiple HPV types, the two techniques tended to agree on at least one of the types present but to disagree on the total number of types (data not shown). The 20 additional positive samples detected by PCR 2 alone were confirmed by repeat testing. As shown in Table 6, only 4 (20%) of the 20 samples contained HPV type 6/11, 16, 18, 31, 33, or 45. Most of the additional positives were HPV types for which type-specific oligonucleotides were not available. As mentioned above with reference to Table 2, these unidentified types were almost all from women with normal cytologic diagnoses or reactive atypia. HPV types commonly associated with high-grade cervical neoplasia (types 16, 18, 31, and 33, in particular) were uncommon among the 87 control women, according to the results of either Southern blotting (prevalence, 1%) or PCR 2 (prevalence, 3%). Among the 32 case women with condylomatous atypia or CIN, these types were found in 44% by Southern blot and in 50% by PCR 2.
DISCUSSION The results of this intermethod comparison demonstrate clearly that PCR primer performance cannot be predicted adequately by sequence homology or plasmid DNA experiments. PCR 1 appeared, by such preliminary criteria, to employ a consensus primer set capable of amplifying many known HPV types. In practice, however, the primer set amplified selected HPV types only. The reasons for this selective performance are unknown but may include the high AT nucleotide content of the primers. We conclude that optimization with clinical specimens is required before a primer set can be termed consensus, i.e., capable of detecting a broad range of HPV types. TABLE 5. HPV detection by PCR 2 and Southern blot' No. of PCR 2 results Southern blot result
a
andeStheted
6/11, 16, 18, 31, 33, or 45 Other
Statistical significance: chi-square 25.7 (1 df), P < 0.001. Typed according to Southern blot result. There were seven multiple infections, five of which were classified as "16, 31, 33, or 35" because at least one type was in that group. c Types 42, 43, 45, 51, 52, and 56 and unknown types. By Southern blotting there were no infections with type 6/11, 18, or 44. a
Positive by PCR 2 Detected by PCR 2 Pand negative blot blot and Southern by Southern blot
typeb
+
64 0
Total
64
Statistical significance: McNemar's chi-square
=
+
Total
20 36
84 36
56
120
20.0 (1 df), P < 0.001.
a
Statistical significance:
chi-square = 4.9 (1 df), P = 0.03.
b Typed according to PCR 2. There were nine
multiple infections, seven of which were called "type 6/11, 16, 18, 31, 33, or 45" because at least one type was in that group.
The comparison of PCR 2 with Southern blotting showed that PCR 2 used a true consensus primer set, apparently amplifying all HPV types detectable by low-stringency Southern blotting (at least those types represented in the study population). In addition, PCR 2 detected HPV in many women whose cervical cell specimens were negative by Southern blotting. This increase in sensitivity resulted mainly from the additional detection of unidentified HPV types (other than types 6, 11, 16, 18, 31, 33, and 45) among women without apparent cervical neoplasia. In contrast, the increased sensitivity of PCR 2 compared with Southern blot hybridization was hardly evident among women with condylomatous atypia or CIN because both methods detected HPV in virtually all cases. The cervicovaginal lavage used in this study to collect cervical cell specimens provides ample DNA, permitting optimal sensitivity of the Southern blot method (8). The additional sensitivity of PCR might be more apparent in studies relying on less ample cervical scrapes or swabs because the PCR method required a specimen less than 1/100 the size of the clinical specimen used by Southern blot to obtain comparable results. A remaining question is the clinical significance of the HPV infections of unidentified type found by PCR 2 alone. The HPV-positive women with such infections had mainly normal diagnoses or had reactive atypia. Some of the types considered unidentified by PCR 2 could be known or notyet-typed genital HPVs for which oligonucleotide probes were not available, and, alternatively, some of the apparent infections could be false-positive results. Further hybridization analyses with new probes, as well as cloning and sequencing efforts under way, should clarify this issue. When either Southern blotting or PCR 2 was used, HPV types commonly associated with high-grade cervical neoplasia (types 16, 18, 31, and 33, in particular) were found in less than 10% of our control population but in nearly 50% of the case population. The low prevalence of these particular HPV types among cytologically normal women in our study group suggests that type-specific HPV screening might serve as a useful predictor of undetected or incipient cervical neoplasia in this population. An ongoing cohort study of the same Kaiser Permanente patient population is designed to investigate the utility of such screening. In this context, it is important to note that the control women in our study can be considered a population at low risk for HPV infection. Their median age was 30, most were married, and we excluded by chart review all women with a medical history of cervical neoplasia. From other studies, it appears that HPV prevalence peaks in women with multiple sexual partners in their early to mid-twenties and then declines with increasing age (3). Thus, studies of younger, unmarried female populations
VOL. 29, 1991
(1) might be expected to yield higher HPV prevalences among cytologically normal women than we observed, and the predictive value of a positive HPV test would decline as a result. In conclusion, we have shown that a carefully optimized PCR method compares favorably with Southern blot hybridization and shows promise for HPV DNA detection in clinical specimens. As sensitive PCR assays are compared and validated, it will be important to reach a scientific consensus regarding two important issues that remain controversial: the prevalence of different types of HPV in women with normal and abnormal cytologic diagnoses and the corresponding prognostic importance of HPV DNA detection. REFERENCES 1. Bauer, H. M., Y. Ting, C. E. Greer, J. C. Chambers, C. J. Tashiro, J. Chimera, A. Reingold, and M. M. Manos. 1991. Genital human papillomavirus infection in female university students as determined by a PCR-based method. JAMA 265: 472-477. 2. Chow, V., K. M. Tham, M. Yeo-Gloss, S. K. Lim-Tam, I. Sng, T. Thirumoorthy, and H. Bernard. 1990. Molecular diagnosis of genital HPV DNA types by polymerase chain reaction and sensitivity-standardized filter in situ hybridization in randomly sampled cohorts of Singapore women. Mol. Cell. Probes 4:121131. 3. Ley, C., H. M. Bauer, A. Reingold, M. H. Schiffman, J. C.
PCR VERSUS SOUTHERN BLOT FOR HPV TESTING
4. 5.
6.
7.
8.
9.
10.
577
Chambers, C. J. Tashiro, and M. M. Manos. Submitted for publication. Lorincz, A. T. 1989. Human papillomavirus detection tests, p. 953-959. In K. K. Holmes et al. (ed.), Sexually transmitted diseases, 2nd ed. McGraw-Hill Book Co., New York. Lorincz, A. T., W. D. Lancaster, and G. F. Temple. 1986. Cloning and characterization of the DNA of a new human papillomavirus from a woman with dysplasia of the uterine cervix. J. Virol. 58:225-229. Manos, M. M., Y. Ting, D. K. Wright, A. J. Lewis, and T. R. Broker. 1989. Use of polymerase chain reaction amplification for the detection of genital human papillomaviruses. Cancer Cells 7:209-214. Ting, Y., and M. M. Manos. 1990. Detection and typing of genital human papillomaviruses, p. 356-367. In M. Innis et al. (ed.), PCR protocols: a guide to methods and applications. Academic Press, Inc., San Diego, Calif. Vermund, S. H., M. H. Schiffman, G. Goldberg, D. B. Ritter, A. Weltman, and R. D. Burk. 1989. Molecular diagnosis of genital human papillomavirus infection: comparison of two methods used to collect exfoliated cervical cells in the genital tract. Am. J. Obstet. Gynecol. 160:304-308. Ward, P., G. N. Parry, R. Yule, D. V. Coleman, and A. D. B. Malcolm. 1989. Human papillomavirus subtype 16a. Lancet ii:170. Young, L. S., I. S. Bevan, M. A. Johnson, P. I. Blomfield, T. Bromidge, N. J. Maitland, and C. B. J. Woodman. 1989. The polymerase chain reaction: a new epidemiological tool for investigating cervical human papillomavirus infection. Br. Med. J. 298:14-17.
