Genome-wide transcriptional analyses provide valuable insights into the biology of KSHV and human pathogens in general. The first latent KSHV mRNA (kaposin) and the lytic mRNA (nut-1) were identified through reverse Northern hybridization of labeled total mRNA from mock- or TPA-treated BCBL-1 cells with a restriction digest of KSHV genomic fragments as a template (75
). A whole-genome Northern blot scan for KSHV transcripts introduced the current classification into type one (latent), two (latent or lytic), and three (lytic only) mRNAs (63
). Similar, more detailed data have since been obtained by using KSHV DNA gene chips, which employed full-length predicted KSHV cDNAs as a template (28
). The latter experiments classified KSHV transcripts into temporal classes (alpha, beta, and gamma), analogous to the other herpesviruses. Because DNA arrays are based on solution hybridization, only substantial differences in the level of each target mRNA could be discerned, and large amounts of sample [ca. 107
cells or ≥1 μg of poly(A) mRNA per routine hybridization] were needed (14
). A variety of template amplification methods are currently under development to address this problem (e.g., linear amplification with T7 polymerase or template switching [41
]). However, these methods may introduce a bias because of sequence-dependent processivity of the polymerase. According to one manufacturer, the concordance of enzymatically amplified cDNA pools compared to nonamplified mRNA is estimated at no more than 80% (Clontech, Inc.). In contrast, real-time quantitative RT-PCR allows the quantification of a smaller set of mRNAs but with, in principal, single cell sensitivity (22
). Since real-time quantitative PCR amplicons are on average 100 bp and all primers fit similar, highly restrictive performance criteria, enzyme processivity and secondary structures pose less of an issue. The real-time quantitative RT-PCR array introduced here can be improved to use as little as 1 ng of poly(A) mRNA or 103
cells per experiment (unpublished observations). Greater sensitivity, i.e., fewer input cells is achievable, but at the limit random target cell variation becomes a concern. While a Northern blot can distinguish between differentially spliced transcripts, DNA array experiments (with PCR-amplified DNA fragments) at present do not. Since a significant proportion of herpesvirus transcripts are spliced and many more are 3′ coterminal, a genome scan by using real-time, quantitative RT-PCR and primers designed for specific splice sites is ideally suited for investigating this genus of human pathogens.
Automated primer design based on predicted KSHV ORFs proved surprisingly successful. More than 90% of the primer pairs, which were designed for real-time, quantitative PCR under universal cycling conditions, worked without further optimization at a single primer concentration (166 nM) and Mg2+-ion concentration. Genomic DNA contamination was eliminated through poly(dT) purification of the input RNA, although limited DNase I digestion suffices for this application as well (unpublished observations). All primer pairs amplified a single KSHV-specific product, which in this study was reliably detected by using a double-strand-specific dye rather than additional, more costly fluorescently labeled probes. Whether this sensitivity and specificity is sufficient for clinical studies is currently under investigation.
Several groups previously used levels of a single KSHV mRNAs or protein as a surrogate marker for KSHV infection. Compared to the whole-genome transcript pattern, we can now evaluate the appropriateness of these choices. (i) orf29 mRNA primers were originally developed for the detection of lytically replicating KSHV by conventional RT-PCR (55
). orf29 protein encodes a capsid component and, based on Northern analysis in BCBL-1 cells, the orf29 mRNA falls into the gamma-2 temporal class. Since mature orf29 mRNA is spliced by removing a 3,251-bp intron, primer pairs that are located on opposite sides of the intron never amplify contaminant DNA. Because of this property, quantification of orf29 spliced mRNA constitutes the most stringent assay for complete KSHV lytic replication. (Since a second mRNA encoding orf48 also splices into orf29 exonB, analysis of intra-exon sequences alone can result in conflicting data.) Adopting the orf29 assay for real-time quantitative RT-PCR allowed us to quantify the inhibition of KSHV de novo lytic replication in SCID-hu Thy/Liv mice by ganciclovir (16
). However, CT
values for spliced orf29 mRNA were consistently higher (5 to 10 cycles) than CT
values of other spliced gamma-2 amplicons, such as orfK8.1. In other words, a much higher proportion of late lytically infected cells must be present in the culture to yield a significant orf29 signal. This suggests that correctly spliced orf29 mRNA is present in low quantities, late in the infectious cycle, with either a short half-life or unstable and that this assay therefore underestimates the number of lytically infected cells. (ii) Earlier on, we also developed conventional primers for spliced v-cyclin latent mRNA, as well as corresponding real-time quantitative PCR primers (16
). Like the primers for orf29, the primers for v-cyclin mRNA are located on opposite sides of a large (4,037-nucleotide) intron. Hence, primers for spliced v-cyclin mRNA provide a specific probe for latently infected cells. (iii) Other real-time quantitative PCR primers for KSHV have been published, but the corresponding amplicons do not cross splice sites (3
). This has the advantage that the same primer-probe combination can be used to quantify KSHV viral load, but it relies on careful and more extensive provisions to remove contaminating viral DNA if these primers are used for transcript analysis or transmission studies. Based on these insights we compiled a set of primer pairs and corresponding TaqMan probes for high-throughput, high-sensitivity, high-specificity staging of KSHV-infected cells (see Table ). These encompass well-characterized spliced latent, alpha, beta, and gamma mRNAs, most of which have been independently validated by protein expression studies and/or in situ analysis in KS tumors.
The principal question of this inquiry was the assignment of LANA-2/vIRF-3 as a class one latency mRNA, similar to LANA, v-cyclin and v-FLIP. LANA-2/vIRF-3 is comprised of the predicted K10.5 and K10.6 orfs and was initially identified based on its sequence similarity to viral interferon regulatory factors (vIRFs) (37
). Indeed, like KSHV vIRF-1/K9 and vIRF-2/K11.1, LANA-2/vIRF-3 inhibits interferon signaling in experimental systems. LANA-2/vIRF-3 stands out, however, because of its uniform, widespread expression in PEL and multifocal Castleman's disease. Overall, the transcriptional organization of the K10 region is quite complicated. LANA-2/vIRF-3 is encoded by a 1,701-bp ORF, which is readily detected in untreated PEL cell lines (37
). All three primer pairs, which were specific for either exon I or exon II or which spanned the splice junction in between, yielded concordant results, and these results paralleled the pattern of LANA/v-cyc/v-FLIP transcription. Hence, LANA-2/vIRF-3 is encoded by a latency type one mRNA. In contrast, vIRF-1/K9 and vIRF-2/K11.1 mRNAs are induced by TPA with early kinetics in PEL cells, as evidenced by the simultaneous comparison presented here.
Of major concern in comparing transcription patterns between various studies are the myriad technical differences such as timing, drug concentration, and single gene probe characteristics. In the case of KSHV-infected PEL cell lines, culture conditions and time in culture exert a dramatic influence on the timing and efficiency of viral reactivation. Relating the transcription profiles for all KSHV mRNAs provides a common frame of reference within in a single experiment, but this is still a reflection of the particular experimental setup.
While KSHV lytic transcription appears to be an all-or-nothing response, which proceeds to completion once orf50 is expressed, KSHV can enter very distinct latency programs in terms of temporal regulation (latency type one or two) and tissue specificity (KS tumor versus B-cell lineage). All class one latency mRNAs are resistant to the effects of other chemical inducers such as ionomycin or butyrate, as well as TPA (data not shown), whereas no other KSHV transcripts share this feature. This suggests that LANAp, the promoter for LANA, v-FLIP, and v-cyclin (15
), and the LANA-2/v-IRF-3 promoter share common regulatory features. The fact that latency type one mRNAs do not increase upon viral reactivation in BCBL-1 cells suggests that their expression might be latency specific and that latency type one mRNAs might be expressed in latently infected cells to the exclusion of lytic mRNAs. Alternatively, latency type one mRNAs might be constitutively transcribed, but their promoters are isolated from the orf50-mediated upregulation of neighboring lytic transcription units.