|Home | About | Journals | Submit | Contact Us | Français|
Knowledge of polyomavirus BK (BKV) genomic diversity has greatly expanded. The implications of BKV DNA sequence variation for the performance of molecular diagnostic assays is not well studied. We analyzed 184 publically available VP-1 sequences encompassing the BKV genomic region targeted by an in-house quantitative hydrolysis probe-based PCR assay. A perfect match with the PCR primers and probe was seen in 81 sequences. One Dun and 13 variant prototype oligonucleotides were synthesized as artificial targets to determine how they affected the performance of PCR. The sensitivity of detection of BKV in the PCR assay was a function of the viral genotype. Prototype 1 (BKV Dun) could be reliably detected at concentrations as low as 10 copies/μl. However, consistent detection of all BKV variants was possible only at concentrations of 10,000 copies/μl or higher. For BKV prototypes with 2 or more mismatches (representing genotype IV, genotype II, and genotype 1c strains), the calculated viral loads were 0.57 to 3.26% of the expected values. In conclusion, variant BKV strains lower the sensitivity of detection and may have a substantial effect on quantitation of the viral load. Physicians need to be cognizant of these effects when interpreting the results of quantitative PCR testing in transplant recipients, particularly if there is a discrepancy between the clinical impression and the measured viral load.
Polyomavirus BK (BKV) has become an important pathogen in kidney transplant patients. Immunosuppression given to prevent acute rejection triggers BK viruria in 10 to 60%, viremia in 5 to 30%, and biopsy-proven viral nephropathy in 1 to 10% of patients (8, 10, 11). Initial graft loss rates associated with BKV nephropathy were very high but have now dropped to <25%. This success has been attributed to intensive viral monitoring followed by preemptive reduction in immunosuppression (1, 14). In a recent survey that had an overall response rate of 55.5%, 173 of 200 (86.5%) kidney transplant centers reported screening for BKV in blood by quantitative PCR, while 111 of 202 (55.5%) performed viral screening in urine (2). In the latter category, 90% of respondents preferred PCR screening to urine cytology. While cytology is a useful modality to screen for viral nephropathy in low-resource settings, it is less sensitive than quantitative PCR for detecting viral replication prior to the onset of clinical nephropathy. In addition, it cannot differentiate BKV infection from infection by the related polyomavirus JC, which causes significant graft dysfunction at a substantially lower frequency.
Current quantitative PCR assays were developed several years ago using BKV Dun or similar genotype I strains as reference sequences for the design of primers and probes. However, our knowledge of BKV genomic diversity has expanded enormously (6, 7, 12, 16). PCR-based diagnostic and treatment algorithms must be reevaluated to take into account newly discovered BKV single-nucleotide polymorphisms. One approach to define the potential extent of this problem is to assay the same sample using a panel of different PCR assays (5). This is a labor-intensive method that is not practical for routine application by clinical laboratories. We employed an alternate approach that consists of aligning PCR primer and probe sequences with large data sets of BKV sequences (3). This bioinformatic evaluation was followed by experimental amplification of 14 custom oligonucleotides of extended length designed to comprehensively represent genetic variability in the targeted area of the (VP-1) gene. Our results show that variant BKV strains significantly lower the sensitivity of detecting viral DNA and have a substantial effect on quantitation of the viral load. We recommend that molecular diagnostic laboratories offering BKV testing regularly reevaluate their current assays for the ability to accurately identify and quantitate the majority of viral strains circulating in the communities they serve. This recommendation is particularly applicable to geographic locations with a high incidence of genotypes other than type 1. It is worth recalling that BKV genotype IV has a reported prevalence of 54% in Mongolia, while genotype III accounts for 9% of BKV isolates reported from Africa (17).
A total of 184 BKV VP-1 sequences matching primer and probe sequences of the PCR assay used in our laboratory were retrieved from GenBank. These sequences were aligned by Clustal X with default multialignment parameters (13). The alignments were manually adjusted using BioEdit (T. Hall, Department of Microbiology, North Carolina State University; available at http://www.mbio.ncsu.edu/BioEdit/BioEdit.html).
When not already known, genotype assignment of the sequences was based on phylogenetic clustering using known reference sequences, as previously described (7). Neighbor-joining trees were constructed in Mega 4.1 using Kimura's two-parameter method and the complete deletion option for gaps and missing data. Trees were viewed using the Tree Explorer program. A bootstrap test with 1,000 replicates was used to estimate the confidence of branching patterns in the trees.
Fourteen chromatographically purified synthetic oligonucleotides, each 135 nucleotides in length and representing all known BKV genetic variation in the VP-1 gene region targeted by the PCR assay, were purchased for use as artificial targets to compare amplification efficiencies (Integrated DNA Technologies, Coralville, IA). Nucleotides AGGG were incorporated at one end of these synthetic nucleotides, and AAAT at the other end. Oligonucleotide solutions in EB buffer (Qiagen) were standardized by measuring the absorbance at 260 nm and represented target sequence concentrations ranging from 1E8 to 1E0 copies/μl.
The assay targeted the BKV VP-1 gene as follows: forward primer, 5′-GCAGCTCCCAAAAAGCCAAA-3′ (1600 to 1619; Dun numbering); reverse primer, 5′-CTGGGTTTAGGAAGCATTCTA-3′ (1726 to 1706; Dun numbering); probe, 5′-ACCCGTGCAAGTGCCAAAACTACTAATAAAAGG-3′ (1623 to 1655; Dun numbering).
The real-time PCR was performed in a total volume of 20 μl and contained the following components: 10 μl TaqMan Fast Universal PCR Master Mix (2×; catalogue number 4352042), 1.5 μl of each primer (prepared at 1 μM), 1 μl of probe (prepared at 10 μM), and 6 μl oligonucleotides at specified concentrations. The PCR cycling program consisted of the following steps: 95°C for 4 min and then 95°C for 10 s, 60°C for 30 s, and 72°C for 10 s for a total of 40 cycles. Thermal cycling was performed using an Applied Biosystems 7500 apparatus. Standard precautions were employed to prevent PCR contamination. Pre- and postamplification steps were done in separate laboratories. The quantitation of target copy numbers used a standard curve with the pBKV(34-2) plasmid, which contains the BKV Dun genome (ATCC 45025).
Analysis of the real-time PCR assay was performed using SDS software (Applied Biosystems). Unknown target concentrations were determined by linear regression using threshold cycles (CT) plotted against the log10 copy number of the standard BKV plasmid. Corrections for sample dilution and descriptive statistics were performed in Microsoft Excel 2007.
The breakdown of genotypes for the 184 sequences retrieved for the study was as follows: 16 Ia, 30 Ib1, 45 Ib2, 30 Ic, 53 IV, 7 II, and 3 type III. An alignment of these sequences indicated a perfect match with the PCR primers and probe for 81 sequences (13 genotype Ia, 29 genotype Ib1, 35 genotype Ib2, 3 genotype 1c, and 1 genotype II). All genomic variability in the BKV VP-1 region targeted by our quantitative PCR assay could be represented by 14 unique prototype sequences (Table 1). Prototype 1 matches the BKV Dun reference sequence, as well as several other sequences, all except one of which are genotype 1. Prototypes 2 thru 14 correspond to a broad spectrum of sequences that includes genotypes II, III, and IV at coverage frequencies that are summarized in Table 2. It is apparent that prototypes 1, 2, 6, 7, 8, 9, 10, 12, 13, and 14 cover primarily genotype I strains. Genotype II is represented by prototypes 5 and 11, genotype III by prototype 5, and genotype IV by prototypes 3 and 4. The locations of nucleotide mismatches between the viral prototype and PCR primer/probe sequences are depicted in Fig. 1 and enumerated in Table 3, which shows the relative amplification efficiencies of the different prototype sequences.
The results indicate that the sensitivity of detection of BKV is a function of the viral genotype. Thus, prototype 1, which represents BKV Dun and similar strains, could be consistently detected at concentrations as low as 10 copies/μl (10,000 copies/ml). This is clinically relevant, since plasma BKV loads of this magnitude have been used as a trigger to lower immunosuppression and initiate antiviral therapy. Notably, at lower concentrations, namely, 1 copy/μl, detection of prototype 1 was not possible in 1 of 3 replicates (Table 3). Prototype 3, which covered the majority of genotype IV strains, behaved essentially similarly to prototype 1 in terms of the sensitivity of detection. However, prototype 4, which corresponds to 6/53 (11%) known genotype IV strains, could not be detected in 2 of 3 replicates set up at concentrations of 1E2, i.e., 100 copies/μl and 1E1, i.e., 10 copies/μl. Consistent detection (defined as detection in 3 of 3 replicates) of all 14 BKV prototypes was possible only at concentrations of 1E4, i.e., 10,000 copies/μl or higher. At concentrations of 1E1, i.e., 10 copies/μl, the assay was able to detect only 6 of 14 prototype sequences in all three replicates, while lowering of the prototype concentration to 1E0, i.e., 1 copy/μl, reduced the performance of the assay to reproducible detection of only 3 prototypes.
Another observation of interest is that, compared to the reference prototype (number 1), the measured viral load is variably underestimated for all other prototypes (numbers 2 to 14). This effect is most pronounced at low concentrations. In general, quantitation of the BKV load by real-time PCR depends principally on two factors: (i) the linearity of the standard curve (prepared using the BKV Dun strain plasmid in our assay) and (ii) the amplification efficiency of the target sequence. Both these factors appear to contribute to the prototype-specific variations seen in our experimental system. The linear part of a real-time PCR standard curve is characterized by a CT difference of approximately 3.3 between serial 10-fold dilutions of the plasmid standard. Using this yardstick, CT measurements were in the linear range for BKV Dun prototype 1 at all concentrations up to 1E2, i.e., 100 copies/μl. All values outside the linear range are indicated in boldface in Table 3. For 13 variant BKV strains (prototypes 2 through 14, taken together), linearity was observed only at concentrations up to 1E5, i.e., 100,000 copies/μl. The linear range of the PCR assay for variant strains fell sharply at lower concentrations: linear measurements were obtained for only 6/13 prototypes at concentrations ≥1E3, i.e., 1,000 copies/μl; 3/13 prototypes at concentrations ≥1E2, i.e., 100 copies/μl; and 0/13 prototypes at concentrations ≤1E1, i.e., 10 copies/μl. Amplification efficiency was also significantly reduced for BKV variants (prototypes 2 to 14) expressed as a percentage of the BKV Dun reference (prototype 1), which was assumed to represent 100% efficiency (Table 3). Thus, for prototype 2, the viral-DNA yield was 13.31% of the expected value at 1E7copies/μl and 15.51% of the expected value at 1E6, i.e., 1,000,000 copies/μl. For measurements in the lower nonlinear part of the standard curve (boldface in Table 3), the calculated yields were quite variable and inaccurate, which explains why the mean of several calculated results was >100% of input DNA.
Finally, the experiments conducted illustrate that the amplification efficiency is a function of the number of sequence mismatches between the viral target and PCR assay primer/probe sequences. A sequence alignment of all prototype sequences is shown in Fig. 1, including the locations of all variant nucleotides in relation to the PCR primers and probe. Table 3 enumerates the mismatches between the target sequence and the forward primer (F), reverse primer (R), and probe (P) sequences. Prototypes 3, 4, 10, and 11 had 2 or more mismatches. Prototypes 2, 5, 12, 13, and 14 had 1 mismatch with the primer sequences, while prototypes 6, 7, 8, and 9 had 1 mismatch with the probe. For 4/13 BKV variants with 2 or more primer/probe mismatches, the calculated viral loads were 0.57 to 3.26% of the expected values (i.e., approximately 100-fold lower).Considering 5/13 BKV variant prototypes with only 1 primer mismatch, 2/5 and 3/5, respectively, yielded calculated target copy numbers approximately 10- and 20-fold lower than the expected values. Finally, for 4/13 BKV variants with only 1 probe mismatch, underestimation of the target copy number was less pronounced and observed deviation was within 5-fold of the expected value.
Quantitative PCR is now widely used to monitor BKV infection after kidney transplantation. Different laboratories employ different viral targets and primer sequences for amplifying viral DNA. Assays are typically based on the BKV Dun strain or similar reference sequences of genotype I. In the assay evaluated here, primers and probes designed in this manner showed a perfect match with only 87/184 (47.3%) and one or more mismatches with 97/184 (52.7%) publicly available sequences. Mismatches were seen most often with genotype IV, followed by genotypes 1c, 1b, II, and III. This rank order is consistent with the known phylogenetic distances between different viral strains.
To study the impact of genetic variability on BKV PCR, 14 prototype sequences incorporating representative single-nucleotide polymorphisms in 184 viral strains were synthesized. These prototypes generally corresponded to specific genotypes, but with occasional exceptions, which may represent more recent mutational events occurring after the divergence of major genotypes in the course of viral evolution. The ability of the quantitative real-time PCR assay to accurately detect virus was maximal for BKV genotypes 1a, 1b1, and 1b2. Sequences represented by prototype I could be consistently detected at all concentrations up to 1E1, i.e., 10 copies/μl (i.e., 1E4, or 10,000 copies/ml), which is a threshold that has been frequently used for reducing immunosuppression in patients with BK viremia or viral nephropathy (4). For genotype IV, comparable detection was observed for sequences corresponding to prototype 3, but those represented by prototype 4 were consistently amplified only at a concentration that was 2 log units higher. Genotype II/III sequences corresponding to prototype 5 could be detected in all replicates only if the concentration was 1E2, i.e., 100 copies/μl or higher. For genotype III sequences (prototype sequence 11), the threshold for consistent detection was 1E4, i.e., 10,000 copies/μl. The thresholds described are specific for the assay used and could be altered by modifying the amplification conditions. Nonetheless, the comparative data dramatically illustrate the effect of the viral genotype on detection of viral DNA.
Clinical management of BKV infection depends not only on the detection of virus, but also on estimation of the viral load in body fluids. By definition, the amplification efficiency was 100% for prototype 1, which represents 81 genotype 1a, 1b1, and 1b2 sequences showing a perfect match with PCR primer/probe sequences. However, as shown in one prior investigation (5), quantitation of the viral load was markedly compromised for variant BKV strains. Thus, at a DNA input of 1E7, or 10,000,000 copies/μl, the calculated viral load for genotype IV sequences was 1.06% of the expected value for sequences represented by prototype 3 and 2.47% for prototype 4 (Table 3). For genotype II and III sequences represented by prototype 5, amplification efficiencies as low as 9.36% were observed. Quantitation of genotype 1c sequences (prototype 6) was generally not as severely affected, with the exception of one sequence from East Asia (AB245326, prototype 14). Two genotype 1a sequences, AB276228 from South Africa (prototype 13) and AB268405 from Japan (prototype 14), also showed poor amplification. In clinical practice, changes in the viral load of 10-fold or higher are often considered to be meaningful. Using this criterion, only the variants corresponding to prototypes 6, 7, 8, and 9 gave acceptable results, since the calculated copy numbers for other prototypes were less than 10% of the expected values. Underestimation of the target copy number was not markedly dependent on the target concentration per se, but marked assay variability was observed for samples tested at the lower end of the standard curve.
It is notable that in several instances a single-nucleotide primer mismatch with the target sequence substantially compromised detection by PCR. Mismatches at the level of the probe sequence were less critical. Analogous findings have been reported in the literature. Thus, single-nucleotide polymorphisms have been reported to not always be detrimental to assay performance (5), while 2 mismatches at the 3′ end of a PCR primer can result in up to 2-log-unit differences in the calculated cytomegalovirus load (9). Large numbers of mismatches result in complete lack of amplification of the viral target sequence (a false-negative result) (15).
Two previously published studies on the effect of BKV genetic variation on assay performance deserve mention. Hoffman et al. used seven different primer/probe sets to perform PCR on urine specimens and found substantial interassay disagreements that were most striking for genotype III and IV strains (5). Expected and observed DNA copy numbers varied by as much as 4.2 log10 templates per reaction. Notably, significant assay variation was seen with primers directed at the BKV large T antigen, as well as VP-1, while the agnogene was not evaluated. Dumoulin and Hirsch found gene polymorphisms in the target sequence of their assay in 32%, 23%, and 82% of sequences corresponding to the forward primer, reverse primer, and probe, respectively (3). The effects of these polymorphisms on amplification of genotype-specific sequences were not evaluated. Modification of the PCR assay using primer or probe sequences with degeneracy at 4 nucleotide positions was able to correct for the majority of the observed sequence mismatches. However, no attempt was made to provide coverage for more divergent variant BKV strains, which constituted approximately 10% of published sequences. The modified PCR assay was found to be comparable to the original PCR assay, but its performance with respect to detection and accurate quantitation or uncommon BKV strains other than genotype I was not evaluated.
In conclusion, our studies indicate that BKV quantitative PCR assays designed using genotype 1 reference sequences, such as BKV Dun, do not perform satisfactorily for mutant viral strains. Detection sensitivity and amplification efficiencies are particularly compromised for genotypes Ic, II, III, and IV. While these genotypes are relatively uncommon in the United States, genotype IV strains occur more frequently in the Far East. Transplant physicians should suspect variant viral strains when unexplained graft dysfunction occurs in the setting of low-level BK viruria and absence of viremia. A high index of suspicion for the presence of a variant strain should also arise in patients with rejection-like infiltrates at biopsy that do not respond to steroid treatment. Sensitive virus detection and accurate quantitation of the viral load in such patients may be achieved by using alternate PCR assays targeting a different viral gene, employing degenerate primers to account for variant sequences (3), or by using a multiplex approach that allows simultaneous amplification of common variant strains (5).
This study was supported by NIH grant RO1 AI 51227 to P.R.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the National Institute of Allergy and Infectious Disease.
Jill March provided excellent secretarial assistance.
We have no conflicts of interest.
Published ahead of print on 28 September 2011.