Viral Load Measurements for Kaposi Sarcoma Herpesvirus (KSHV / HHV8): Review and an Updated Assay

Abstract

“When you can measure what you are speaking about, and express it in numbers, you know something about it.”​ is a famous quote attributed to Lord Kelvin. This sentiment puts viral load measurements at the center of virology. Viral load, or more precisely, DNA copy number measurements, are also used to follow infections with human herpesviruses, such as Kaposi’s sarcoma herpesvirus (KSHV) and Epstein-Barr Virus (EBV). EBV and KSHV are associated with human cancers, and determining their DNA copy number in the context of cancer prediction and progression on therapy is of fundamental scientific and translational interest. Yet, there is no generally accepted assay for KSHV DNA quantitation, and KSHV viral load is not used in clinical decision-making. Here, we review the history of KSHV DNA detection assays, explore factors that affect sensitivity and specificity, and describe an automated, high-throughput, real-time quantitative polymerase chain reaction (PCR) assay for KSHV and Epstein-Barr virus (EBV). In conjunction with a digital PCR assay using the same primer/probe combination, we describe how to determine the absolute KSHV genome copy numbers in plasma, peripheral blood mononuclear cells, saliva, and other easily accessible body fluids.

Introduction

As part of this anniversary collection of articles, we first review the history of Kaposi’s sarcoma- herpesvirus (KSHV) viral load measurements and second report the details of our most current KSHV viral load assay. Henceforth, we use viral load measurements equivalent to DNA copy number as KSHV is a human DNA virus and as there exists no plaque assay to measure infectious virus from clinical samples. Kaposi Sarcoma (KS) came to the attention of the medical community and the world at large in 1984. It defined the end-stage malignancy of Acquired Immunodeficiency Syndrome (AIDS), a disease caused by human immunodeficiency virus (HIV) in the US and Western Europe (Centers for Disease 1981). It is the most common cancer in people living with HIV (PLWH) today (reviewed in (Yarchoan and Uldrick 2018)). It disproportionally affects under-represented populations in the US as well as PLWH who reside in low and middle-income countries (McMahon, Maurer, and Freeman 2020; Royse et al. 2017; Salyards et al. 2023). KS herpesvirus (KSHV) was discovered ten years later, in 1994 (Chang et al. 1994; Gatherer et al. 2021).

I. Review of the connection between KSHV and KS

The critical evidence for associating KSHV with KS was that KSHV DNA could be detected in all KS lesions but not in uninfected tissue. In those early days, KS lesions were prominent. CD4 counts were used to stage the progression of PLWH to fulminant AIDS, with KS being associated with low CD4 counts. We call this form of KS “AIDS-KS” since it is associated with high HIV viral loads, low CD4 counts, and often other AIDS-defining conditions as well. HIV viral load assays quickly emerged as an invaluable tool for measuring disease progression and therapy response. It has now replaced CD4 count measurements as the primary means of detecting HIV infection and guiding therapy in PLWH. HIV is an RNA-based human retrovirus. KSHV is a DNA-based human herpes virus.

The first wave of epidemiological studies in KS, as in HIV before, relied exclusively on serological tests targeting the KSHV latency-associated nuclear antigen (LANA) (Kedes et al. 1997; Martin et al. 1998; Rainbow et al. 1997; Kellam et al. 1997; Gao et al. 1996). These proved that prior exposure to KSHV, as measured by seroconversion, preceded disease and that only people who acquired HIV by sexual contact were likely to acquire KSHV as well. PLWH who acquired HIV by contaminated blood transfusion were consistently negative for KSHV antibodies. These studies completed Koch’s postulates to prove that KSHV was a necessary agent of AIDS-KS, but HIV was not.

Polymerase chain reaction (PCR) positivity for KSHV in plasma was reported thirty years ago (Whitby et al. 1995). These studies proved that AIDS patients with overt clinical KS disease carried KSHV DNA in their bloodstream. Subsequently, KSHV was detected in saliva (Koelle et al. 1997; Vieira et al. 1997; Brayfield et al. 2004; Mbulaiteye et al. 2004; Newton et al. 2018; Nalwoga et al. 2020; Blackbourn et al. 1998; Casper et al. 2007; Bender Ignacio et al. 2016). Detection of KSHV in saliva or oral swabs has been a consistent observation since the discovery of this virus. KSHV detection in the saliva is lower and independent of Epstein-Barr virus (EBV) genome copy number (Nalwoga et al. 2023; Nalwoga et al. 2024). EBV is another oncogenic human herpes virus. In AIDS patients, EBV is associated with a localized condition, termed oral hairy leukoplakia, as well as several rare lymphomas (Webster-Cyriaque, Middeldorp, and Raab-Traub 2000; Greenspan et al. 1984). In the oral cavity, EBV replicates in normal epithelial cells independent of systemic latency, replication, and clinical disease (Pegtel, Middeldorp, and Thorley-Lawson 2004; Sitki-Green et al. 2002).

Studies in the combination antiretroviral therapy (cART) era, i.e., from 1997 onwards, show that KS can develop in persons on cART with no detectable HIV viral load and near-normal CD4 T cell counts (Krown et al. 2008; Maurer, Ponte, and Leslie 2007; Moorad et al. 2023). We call this form of KS “HIV-KS” since it is associated with past and now controlled HIV infection, but not terminal AIDS.

Studies in organ-transplantation-associated KS proved that immune suppression of KSHV-infected persons caused KS in the absence of HIV (Barozzi et al. 2003). Moderating the drug-induced CD4 cell suppression is correlated with the disappearance of KS lesions. T cell-specific immune suppressive drugs (cyclosporine) are associated with transplant KS. Switching to a combined B and T cell-targeting suppressant (rapamycin) induces KS regression (Stallone et al. 2005). Rapamycin and other mTOR-targeting therapeutics are active against KS and KSHV-associated lymphoma (Sin et al. 2007; Krown et al. 2012; Caro-Vegas et al. 2019; Ji et al. 2017). We call this form of KS “Transplant-KS” since it is associated with immune suppression, but not HIV infection.

Studies in older men from the Levant proved that aggressive immune deficiency was dispensable for KS disease, and prior exposure to the KSHV virus alone was sufficient. We call this form of KS “classic KS.” Classic KS and, thus, KSHV infections predate the introduction of HIV into the human population. Skin KS cases were described in Vienna as early as 1872 and in Uganda in 1960 (Lothe 1960). Today, we refer to these entities as classic and “endemic KS,” respectively. Antibodies directed against KSHV are detectable in persons with classic or endemic KS. KSHV DNA is present in plasma, peripheral blood mononuclear cells (PBMC), and saliva (Guttman-Yassky et al. 2004; Brown et al. 2006; El-Mallawany et al. 2019; Lidenge et al. 2019; Bender Ignacio et al. 2016). In sum, all forms of KS are associated with KSHV, KSHV viral DNA is detectable in all KS lesions, and anti-KSHV LANA serum antibodies.

KSHV salivary viral loads are highly variable in KSHV endemic regions where this virus is stably maintained, and over 70% of the adult population is seropositive (Phipps et al. 2014; Bender Ignacio et al. 2016). Thus, systemic (blood) and localized (salivary) KSHV copy number measurements may indicate independent underlying pathologies. The levels of CD4 depletion and/or KSHV-reactivation triggers differ between these two compartments, i.e., the two body fluids sample different biological reservoirs. By comparison to KSHV, EBV viral loads are substantially higher in serum and saliva even in asymptomatic patients (Newton et al. 2018; Jacobson et al. 2009), as are human cytomegalovirus (CMV) and human herpesviruses 1 and 2, as well varicella-zoster virus (VZV) viral loads during periods of reactivation. As most persons are infected with multiple herpesviruses, KSHV viral loads must be quantitated in the background of overwhelmingly higher EBV, CMV, and HSV-1 viral loads (Pan et al. 2001). This limits the applicability of multiplexed detection assays.

The information on KSHV viral load or anti-KSHV antibody levels is currently used for the management of KS disease, whether in transplant KS, HIV-associated KS, or for risk modeling in endemic areas or elderly LANA-positive individuals. It is not used for risk modeling in PLWH or young MSM who develop KS but are HIV-negative. After thirty years of study, KSHV serology and KSHV DNA copy number measurements have not left the arena of retrospective research, and KSHV viral load measurements have not moved into clinical practice. This represents a hole and missed opportunity to improve the care of KS patients, regardless of whether they are a part of the classic, endemic, HIV-associated, or transplant populations.

II. Review of KSHV DNA detection

Before the KSHV genome was sequenced in 1996 (Russo et al. 1996), there was the KS330Bam probe. This 631 bp DNA fragment was the first piece of KSHV DNA cloned (Chang et al. 1994). It was used for Southern-Hybridization. Subsequently, PCR primers were generated against this region, leading to the first qualitative PCR-based assays for KSHV. Regular PCR represents an endpoint assay design. It outputs a present/absent answer and is highly non-linear. Specificity is determined by the primers alone. Diluting the input material serially allowed for semi-quantitative DNA copy number determination and estimation of the number of infected cells (Chandriani and Ganem 2010; Weck et al. 1999; Decker et al. 1996). Unless explicitly defined as infectious titer and measured in (IU/mL), we will use “DNA copy number” and “viral load” interchangeably and refer to them in copies/mL.

Quantitative real-time PCR (QPCR) operates by measuring the amount of amplified product at each PCR cycle and estimating a crossing point (Cp) or cycle threshold (CT) where the accumulated amount of product can be detected above the background (Ruijter et al. 2013). The higher the CT or Cp value, the more cycles are required to detect the target and the less abundant the target. CT and copy number form an inverse relationship. No signal at the maximum cycle number indicates the absence of a target DNA at a given limit of detection (LOD).

The first QPCR assay targeting KSHV open reading frame (orf) K6, as well as a nested PCR assay, was introduced by Whitby and colleagues (Whitby et al. 1995; Engels et al. 2003; Yuan, Miley, and Waters 2001). Early assays targeted orf26 since the orf26 sequence was the first available KSHV sequence. Initial Orf26-targeting assays were used in endpoint and nested designs (Pan et al. 2001; Chang et al. 1994; Boshoff et al. 1995). The assay target was one of serendipity and tradition, and the amplicon length was 206 bp. We targeted the LANA orf for the same purpose (Dittmer et al. 1999). Others based their KSHV DNA copy number assay on other parts of the viral genome, such as orf K8 (Lin et al. 2009). Today, many QPCR assays can quantify KSHV (Shafiee et al. 2015; Ferraz da Silva et al. 2015; Lin et al. 2009). Others target orf22 or orfK6 (Brown et al. 2005; Brown et al. 2006; Newton et al. 2018; Nalwoga et al. 2020; Nalwoga et al. 2023; Pellett et al. 1999; Biggar et al. 2000; Stamey et al. 2001). We recently developed a multi-herpesvirus assay for concurrently detecting KSHV, EBV, HSV-1, and CMV (Jacobson et al. 2009); this is not a multiplex assay. Even though these assays were thoroughly validated, none were FDA-approved or commercialized. This represents a gap in our clinical tool set and a barrier to KSHV diagnosis for treatment and clinical trials.

In 2001, Yuan, Miley, and Waters published the ERV-3 QPCR primers (Yuan, Miley, and Waters 2001). ERV-3 serves the same purpose as ß-actin or ß-globin for RNA normalization or ß-globin for DNA normalization (Cone et al. 1993), except it targets a single copy of endogenous retrovirus DNA in the human genome. Therefore, it does not pose the danger of inadvertently detecting other isoforms or alleles for ß-actin or ß-globin. Henceforth, the number of KSHV genome copies could be normalized to human genome copy number by the delta-delta crossing threshold (ddCT) method, assuming that the amplification efficiencies R_eff are similar across these primer pairs (Hilscher, Vahrson, and Dittmer 2005). Relative target quantitation can be further improved when experimentally measured actual and individual K_eff are used (Ruijter et al. 2009).

QPCR assays are used to detect and quantitate many viruses, including human herpesviruses (Miller et al. 2006). The most highly developed assays are commercial ones for quantifying the human immunodeficiency virus (HIV). Driven by urgent clinical need and the scale of the HIV epidemic, we have witnessed the continuous evolution of HIV viral load assay sensitivity. The first FDA-approved assays for HIV had a LOD of 400 copies per mL (cps/mL). The LOD for current high-throughput routine assays is 50 cps/mL (Crump et al. 2009). Most recently, single-copy HIV viral load assays have also been introduced, although these require more than 1 mL input volume (Palmer et al. 2003). Similar increases in sensitivity for viral load assays targeting herpesviruses, including EBV and KSHV, have yet to be achieved. This deficiency of sensitive herpesvirus assays represents a gap in our armamentarium to detect infection and assist the treatment of herpesvirus-induced diseases. Until these standardized assays become available on routine and CLIA-certified clinical platforms, applications of KSHV viral load assays to clinical care will remain elusive.

Herpesviruses are associated with multiple human diseases. Disease severity tends to be increased in the context of immunosuppression. In this context, it is typically associated with, if not preceded by, a rise in systemic viral load. KSHV presents with the lowest plasma viral loads among the human herpesviruses during active disease. KSHV serum viral loads among patients with KS or PEL are in the range of 50 – 50,000 cps/mL (Pauk et al. 2000; Tamburro et al. 2012); even in AIDS-KS, viral load in plasma rarely exceeds 100,000 cps/mL. Depending on the population and location, as much as half of all clinically apparent HIV-KS patients have minimal or no detectable circulating KSHV DNA (Spira et al. 2000). Ferraz da Silva et al. observed that KSHV was nearly undetectable in all 157 of their HIV-positive samples, likely due to the viral load being beyond the limit of detection for their assay (Ferraz da Silva et al. 2015). By contrast, KSHV viral load can reach 1 million cps/mL during active episodes of multicentric Castleman’s disease and KSHV inflammatory cytokine syndrome (KICS) (Uldrick et al. 2010; Tamburro et al. 2012; Caro-Vegas et al. 2020; Jary et al. 2018). These general impressions require some cautionary comments, however. Many KSHV-associated diseases (KAD), such as PEL and MCD, occur concurrently with KS (Hansen et al. 2022; Ramaswami et al. 2021; El-Mallawany et al. 2019). They may or may not be detected depending on the clinical capabilities at a location. Thus, systemic KSHV viral load is confounded by occult KAD and other co-infections. Rather than KS lesion burden or response score, these may be the source of high plasma viral loads (Lucena Lage et al. 2024).

Materials and Methods

Cell Culture

Cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 10% FBS, 1% l-glutamine, 100 IU penicillin, and 100 μg/mL streptomycin. Tissue-culture flasks containing the cells and medium were incubated at 37°C with 5% CO2. Cell authenticity was previously verified by the University of North Carolina (UNC) Vironomics Core using Ion Torrent sequencing and short tandem repeat analysis performed by Genewiz, Inc.

Cell Count and Viability

Cells were cultured in growth media and harvested in the log phase, with 1 mL of cells removed for cell counting. From this 1 mL, 20 μL of cells were mixed with 20 μL of 0.2% Trypan blue and were carefully homogenously mixed with a pipette. A hemocytometer was loaded with this 1:1 ratio of cells to Trypan blue solution, yielding a dilution ratio of two. Both hemacytometer counting chambers were loaded with 10 μL of this mix. The hemacytometer was then placed under a microscope to count the viable vs. non-viable cells, where viable cells are defined as those that have not up-taken Trypan blue. Cell count was determined by converting the live cell values into the hemacytometer equation, and cell viability was determined as a factor of viable cells divided by the total cells counted.

DNA Extraction

DNA samples positive for KSHV, EBV, or both were specific for each set of primers. BC-1 cell line DNA was used as a positive control since these cells contain both EBV and KSHV (Cesarman et al. 1995). Cell pellets of 100,000 cells each were reconstituted in 1 mL water. DNA was extracted using MagNA Pure Chemistry and Instruments (Roche Diagnostics, Indianapolis), which yielded 100 μL of purified DNA from 1 mL input. An internal control using a pIDTBlue vector with inserted Drosophila melanogaster DNA (Fly) was included in each sample to determine extraction consistency. The reagent is available from Addgene (#117418).

DNA Preparation for PEG-based Assays

For the PEG-based assay of HIV-negative samples, BCBL-1-Trex-RTA cells were seeded at 7.5E5 cells/mL in 25 mL of 10% Tet-free FBS complete RPMI with Puromycin (final concentration of 1.25 μg/mL) and Hygromycin (final concentration of 20 μg/mL). The cells were then activated to induce lytic replication with 1 μL Doxycycline at a concentration of 25 μg/μL and incubated at 37°C for 48 hours before supernatant harvest. After incubation, the cell solution was centrifuged at 3,200 × g at 4°C for 15 minutes, and the supernatant was retained while the pellet was discarded. For samples treated with PEG, 1 mL of 40% PEG was added to 4 mL of harvested supernatant for a final concentration of 8% PEG. Samples were then incubated on an Eppendorf^® Thermomixer at 600 rpm for about 24 hours at 4°C. These incubation conditions were optimized from four sets of conditions that were tested: 2 hours at room temperature (21°C), 2 hours at 4°C, 24 hours at room temperature (21°C), or 24 hours at 4°C. Following incubation, the solution was spun down at 15,000 × g for 30 minutes. This speed was optimized from five speeds that were tested: 2,500 × g, 5,000 × g, 10,000 × g, 15,000 × g, or 20,000 × g for 30 minutes. Most of the supernatant was then poured off, and the pellet was re-suspended in 400 μL DPBS before being volumed up to 884 μL to account for any remaining supernatant on the pellet. Following re-suspension, 100 μL 10X RQ1 buffer was added to each sample, and they were then treated with 10 μL DNase (Promega) at 37°C for 30 minutes. The DNase concentration was optimized from four DNase volumes: 2.5 μL, 5 μL, 10 μL, and 20 μL. After 30 minutes, 6 μL EGTA stop buffer (Promega) was added to each sample, and they were incubated for 10 minutes at 65°C. All samples with a final volume of 1 mL were then processed using MagNA Pure Chemistry and Instruments (Roche Diagnostics, Indianapolis) using the Large Kit system for plasma to produce 100 μL of purified DNA. Before extraction, adding 2 μL of MS2 RNA aided in DNA extraction, and an internal control pIDTBlue with inserted Drosophila melanogaster DNA (Fly) was included in the extraction to observe standardized isolation efficiency. Control tubes of 1 mL supernatant (without PEG) and tubes with 884 μL supernatant and DNase treatment were processed similarly after equivalent incubation times and conditions.

To develop the PEG-based assay of biohazardous HIV-positive samples, re-activated BCBL-1-Trex-RTA supernatant, prepared in the same manner as described above, was used as the experimental sample. The incubation conditions and centrifugation speed for the samples incubated with PEG were the same as the HIV-negative assay described above. Following centrifugation at 15,000 × g, most of the supernatant was poured off, and the pellet was re-suspended in 400 μL DPBS and volumed up to 784 μL to account for any remaining supernatant. 100 μL 10X RQ1 buffer was added to each sample, and they were then treated with 10 μL DNase (Promega) at 37°C for 30 minutes. After 30 minutes, 6 μL EGTA stop buffer (Promega) was added to each sample, and they were incubated for 10 minutes at 65°C. Following DNase treatment, 100 μL of 10mM Tris/2% Triton X-100 Buffer was added and incubated at 65°C for 1 hour to inactivate hypothetical HIV (Ukkonen et al. 1988). All samples were then processed using MagNA Pure Chemistry and Instruments (Roche Diagnostics, Indianapolis) in the same manner as the HIV-negative assay.

An incubation time of about 24 hours was selected to coincide with a typical workday and a temperature of 4°C was chosen to help maintain DNA integrity. The data indicated that incubation conditions of 4°C for 24 hours and 21°C for 2 hours produced relatively similar copy numbers that were higher than the other two sets of incubation conditions, 4°C for 2 hours and 21°C for 24 hours. Centrifugation speeds of 2,500 rcf, 5,000 rcf, 10,000 rcf, 15,000 rcf, and 20,000 rcf were also tested. Although there was no significant difference between detectable copy numbers of the four higher speeds, 15,000 rcf was chosen as the optimal speed because it produced the highest detectable copy number.

Primers

The primers used in this study are summarized in Table 1. The first set of forward and reverse primers amplifies an ERV-3 fragment of the human endogenous retrovirus (HERV). ERV-3 is a full-length human endogenous retrovirus present in known copy numbers in all human cells, allowing it to be used as a control measure for DNA quality and quantity (Yuan, Miley, and Waters 2001). The second set of primers amplifies a fragment from the latency-associated nuclear antigen gene (LANA-1) found in KSHV (Jacobson et al. 2009). The third set of primers also amplifies a fragment within the KSHV LANA-1 orf; however, the forward primer in this pair has been shifted four base pairs to the right of the forward primer sequence for the second set of primers, while the reverse primers are the same. This four-base pair-shift was shown to decrease Cp by about four units. The fourth set of primers amplifies the Epstein-Barr nuclear antigen 3C (EBNA-3C) fragment from EBV or Herpesvirus-4 (HHV-4) and has been used since 2005 (Hilscher, Vahrson, and Dittmer 2005). Finally, the fifth set of primers amplifies a section of the internal control plasmid with inserted Drosophila melanogaster DNA (Fly).

Table 1.

Primer and synthetic oligonucleotide sequences used in the assays in this paper. The second and third sets of primers are both for detecting KSHV and have the same reverse primer sequence; however, the third set has an altered forward primer sequence.

Set	Target	Primer Name	Sequence
1	Human	ERV-f	5’-CATGGGAAGCAAGGGAACTAATG
1	Human	ERV-r	5’-CCCAGCGAGCAATACAGAATTT
2	KSHV	LANA78-F	5’-GGAAGAGCCCATAATCTTGC
2	KSHV	LANA78(2)-R	5’-GCCTCATACGAACTCCAGGT
3	KSHV	LANA78-F altered	5’-GAGCCCATAATCTTGCACGG
3	KSHV	LANA78(2)-R	5’-GCCTCATACGAACTCCAGGT
4	EBV	EBNA3C-F	5’-AAGGTGCATTTACCCCACTG
4	EBV	EBNA3C-R	5’-AGCAGTAGCTTGGGAACACC
5	Fly	Flyflap-f	5’-ATCATAAAGCGTTTTAAGCTCCAACGA
5	Fly	Flyflap-r	5’-AATCATAATTCCTGACTCCCAAGTGGAC

In addition, we synthesized a series of synthetic single-strand oligonucleotides to serve as positive control and for plate-to-plate validation. These were “Original LANA78 oligo”, which has the sequence 5’- GCCTCATACGAACTCCAGGTACGAGAATTCCTCGCAAGATTATGGGCTCTTCC; “New LANA78 oligo”, which has the sequence 5’-GCCTCATACGAACTCCAGGTACGAGAATTCCGTGCAAGATTATGGGCTCTTCC. Each of these target oligos has terminal sequences that match the PCR primers. Instead of amplifying the original ~ 200 bp fragment, they yield a ~100 bp final product that contains an EcoR1 novel restriction site GAATTC in the center.

Primers and oligos were prepared in a dedicated PCR hood in our clean room. Primers and oligo sequences are ordered salt-free and dry from Applied Biosystems using the sequences provided in Table 1. Tubes are briefly centrifuged, and primers resuspended to 100 μM per the unique amounts present in each tube using LOW EDTA 1X TE (10 mM TRIS-HCL, 0.1 mM EDTA). SYBR Primer preparation dilutes the 100 μM stocks into a 2.0 mL tube for 10 μM working stocks by mixing 800 μL of nuclease-free water, 100 μL of 100 μM Forward Primer, and 100 μL of Reverse Primer, and these are then stored at 4° C for up to one month. TaqMan mixes are made at 20 μM by mixing 500 μL of nuclease-free water, 200 μL of 100 μM Forward Primer, 200 μL of Reverse Primer, and 100 μL of 100 μM probe. Our oligo controls contain known targets for KSHV, EBV, and ERV. Oligo control preparation combines 20.76 μL from each 100 μM stock of KSHV, EBV, and ERV and combines them into a 2.0 mL tube, at which point, nuclease-free H2O is used to bring the volume up to 250 μL mix with a total known copy number of 5 × 10¹² copies/μL per target. A serial dilution of the 5 × 10¹² copies/μL stock is diluted to 1 × 10⁷ copies/μL in Low EDTA 1X TE (10 mM TRIS-HCL, 0.1 mM EDTA). This 1 × 10⁷ copies/μL dilution is then diluted to 1 × 10⁵ copies/μL by taking 200 μL of the 1 × 10⁷ copies/μL dilution and putting it in 19.8 μL of nuclease-free H₂O. This mix is then vortexed and aliquoted into 150 μL single-use matrix tubes and stored in a labeled 96-tube 0.2 mL tube matrix rack at −20°C for up to one year.

FLY spike in control

The spike in control was custom synthesized and cloned into vector pIDTBlue. It is available from Addgene (#117418). The sequence is 5’- AGCGTTTTAAGCTCCAACGATCTCTGTGTAAAGATAAAATATTGTAATTCATTTAAATGAACAAATCAGGCAGAGATCGAGGCTAAAAAGAACTGTCCACTTGGGAGTCAGGAA. The primer binding sites are underlined.

Real-time QPCR for Volume-based Assays

The larger-volume QPCR reactions had a final volume of 94.5 μL: 45 μL sample, 47.14 μL 2X SYBR, and 2.36 μL of combined forward and reverse primers at a 5 μM concentration for KSHV, EBV, ERV, or Fly primers. The reactions were set up using the CAS-1200 pipetting robot (Corbett Research Inc., Australia) in Lightcylcer 480 96-well QPCR plates. This robot uses filtered carbon-graphite pipette tips with liquid-level sensing capability to reduce human error. Lightcycler 480 96-well QPCR plates were run on the 96-well Lightcycler 480 (Roche Diagnostics, Indianapolis). The thermocycling conditions consisted of a pre-amplification step for 5 minutes at 95°C, followed by 40 cycles of 15 seconds of denaturation at 95°C and 1 minute of annealing at 62°C. This was followed by a melting curve step for 5 seconds at 95°C and 1 minute at 60°C with 5 data acquisition points every degree. The raw data extraction via a CSV file of crossing points (Cp) and melting temperatures (Tm) was performed using Lightcycler 480 SW 1.5 software.

A combination virus oligonucleotide standard with a known initial copy number was used as a positive control. Upon receiving each of the synthetic oligonucleotides, they were re-constituted to 100 μM, and by using Avogadro’s number (6.022×10²³ particles per mole), we can calculate the molecules per microliter to one significant digit, which is 6 × 10¹³ molecules per microliter. Molecules per microliter can be equated to single-strand copies per microliter. We then calculated the volume needed in a combination oligonucleotide standard of 250 μL when we wanted a final concentration of 5 × 10¹² cps/μL. After making the combination standard, we do serial dilutions to reach a final working stock concentration of 10 × 100,000 cps/μL. The oligonucleotide working stock then underwent four serial five-fold dilutions (1:5 = 20,000 cps/μL, 1:25= 4,000 cps/μL, 1:125 = 800 cps/μL, 1:625 = 160 cps/μL) for each sample PCR plate.

For comparison of robots, reactions with the same dilution series samples were set up for the original-volume assay, 18.9 μL, using both the CAS-1200 pipetting robot and the Tecan Freedom EVO 100 (Roche Diagnostics, Indianapolis) pipetting robot. The original-volume assay had a reaction volume of 18.9 μL: 9 μL sample, 9.43 μL 2X SYBR, and 0.47 μL of combined forward and reverse primers at a 10 μM concentration for either KSHV, EBV, ERV, or Fly. The thermocycling conditions were the same as that of the larger-volume assay.

Real-time QPCR for PEG Assay

The QPCR plates were set up using a Tecan Freedom EVO 100 in Lightcycler 384-well plates (Roche Diagnostics, Indianapolis) following the original-volume assay with KSHV and Fly primers only. All plates included with-PEG and without-PEG samples and combination virus oligonucleotide dilution series standards. The plates were then processed on the 384-well Roche Lightcycler 480 (Roche Diagnostics, Indianapolis). The thermocycling conditions were the same as those of the volume-based assays.

Production-grade Real-Time QPCR for Five-Plex KSHV Assay

Preparation of extracted DNA for QPCR was performed using a Tecan Freedom EVO 100 (Tecan Material No. 10641100) with a DiTi-only Air LiHa 4-Channel Arm (Tecan Material No. 30063958) using two in-lab programmed Tecan scripts. Tecan Script 1 can be used as an intermediate step before QPCR in which 96-well plates (Eppendorf: Ref. No. 0030129580) are loaded with samples and water or cDNA master mix, if necessary, in preparation for Tecan Script 2. Tecan Script 2 utilizes the plate from Tecan Script 1 to prepare the QPCR plate containing our sample reaction mixes and controls. These scripts were designed to enable flexibility using DNA or RNA to maintain a consistent amount of raw template volume throughout our QPCR protocols.

Tecan Script 1 transfers and mixes 45 μL of nucleic acid from each MagNA Pure 24 100 μL eluate tube and 15 μL of nuclease-free water into individual wells of this empty 96-well plate. The 96-well sample plate prepared by Tecan Script 1 is then sealed with MicroAmp^™ Optical Adhesive Film (Applied Biosystems, Catalog # 4311971) and centrifuged at 2000 rpm for 1 minute to remove any air bubbles. Pre-made 10 μM primer mixes are then used to create our QPCR master mixes by pipetting 25 μL of our 10 μM primer mix into 500 μL of our SYBR^® Select Master Mix (Applied Biosystems: Cat. No. 4472920) or TaqMan Universal Master Mix II with UNG (Applied Biosystems: Cat. No. 4440045) pre-purchased mixes in an empty 2.0 mL tube. These mixes contain either uracil-DNA glycosylase (UDG) in SYBR or uracil-N-glycosylase (UNG) in TaqMan, which are evolutionarily conserved DNA repair enzymes that function to remove uracil from DNA, serving as a tool for eliminating carry-over contamination in PCR reactions. Five primer master mix tubes are created in which two primers target KSHV, one EBV, one ERV, and one targets our FLY 3.0 internal control. This setup allows Tecan Script 2 to contain DNA samples targeted by five primers in individual wells (non-multiplexed), an NTC, and an oligo control on each QPCR plate. A separate 2.0 mL tube is filled with 1,800 μL of nuclease-free water. These tubes are then placed onto a cooled carrier, holding the primer mixes, 1,800 μL of nuclease-free water, and four empty 2.0 mL tubes for our oligo dilutions using the 1,800 μL of nuclear-free water. The centrifuged 96-well sample plate is then placed back onto the Tecan Freedom EVO 100 deck onto another PCR-Cooler 0.2 mL along with an empty MicroAmp^™ EnduraPlate^™ Optical 384-Well Clear GPLE Reaction Plates with Barcode (Applied Biosystems: Cat. No. 4483319) on a cooled carrier, and a matrix rack containing the pre-made oligo control.

Tecan Script 2 is then run once everything is loaded onto the Tecan Freedom EVO 100 deck and results in five reactions for each sample containing 9.429 μL of SYBR or TaqMan, 0.471 μL of primer or primer/probe mix, and 9.0 μL of DNA in each well for a reaction volume of 18.9 μL and a final primer concentration of 0.250 μM for SYBR or 0.500 μM for TaqMan reactions. The plates also contain five NTC wells for each primer mix and four 1:5 oligo dilutions, which are the basis for creating the standard curve. The plates are then sealed with a new optical adhesive film and loaded onto the QuantStudio^™ 7 Pro Real-Time PCR System, 384-well (Thermo Fisher/ABI: Cat. No. A43164). The run template is then uploaded onto the instrument, and the QPCR run is performed using the following cycling conditions: Hold at 50 °C for 2 minutes and 95 °C for 2 minutes (or 10 minutes for TaqMan), followed by PCR with 40 cycles of 95 °C for 15 seconds and 60 °C for 1 minute. If using SYBR, a melt curve is run immediately after PCR with the following cycling conditions: 95 °C for 5 seconds, 60 °C for 1 minute, 95 °C for 15 seconds, and then holding the plate at 50 °C for 5 minutes after completing the melt curve. Results are obtained and analyzed using Design & Analysis Software Version 2.5.1 (Thermo Fisher Scientific).

Further Optimization

Additional optimization was carried out to improve the viral load assay sensitivity further. A more optimal primer set (Set 3 in Table 1) targeting LANA with four base pair changes in the forward primer decreased the crossing point (Cp) for samples by about four units at an annealing temperature of 62°C versus only two units at 60°C than the previously published primer set (Set 2 in Table 1), which could allow for detection of less concentrated samples. The annealing temperature was also found to be more optimal at 62°C for both pairs of primers.

Digital PCR (dPCR)

Digital PCR analysis was performed using a QuantStudio Absolute Q Digital PCR System (Thermo Fisher Scientific: Cat. No. A52864) according to the manufacturer’s protocols for Absolute Q^™ DNA Digital PCR Master Mix (5X) (Applied Biosystems: Cat. No. A52490) using a 20X dPCR assay reaction mix. MAP16 plates were loaded with samples in a dPCR mix. The following dPCR cycling conditions were used: 96 °C for 10 min followed by 40 cycles of 96 °C for 5 seconds and 60 °C for 15 seconds. Sample quantification was analyzed using the Absolute Q Digital PCR Software Version 6 (Thermo Fisher Scientific).

Gel Electrophoresis and Melting Curves for Target Confirmation

The Caliper Labchip GX Touch HT (Perkin Elmer), a highly sensitive gel electrophoresis device, was used to run positive NTC samples and samples at a Cp of 35 or above to confirm the amplified target. Before being run on Caliper, post-PCR samples were diluted, 3 μL sample into 37 μL mH₂O. Synthetic oligonucleotide contamination could be detected at around 50 base pairs (bp) on the gel, while the KSHV target could be detected at around 200 bp.

Processing the Assay with Clinical Samples

For serum samples processed by automatic extraction, the manufacturer’s protocol was used for serum samples processed using the PEG protocol.

Statistical Analysis

Cp and Tm raw data extraction vis CSV file was performed using Lightcycler 480 SW 1.5 software (Roche Diagnostics, Indianapolis, IN). The preliminary analysis was performed using Excel 2016 (Microsoft Inc., Redwood, WA). A logarithmic curve was plotted for each PCR plate using the Cp values of the four oligonucleotide dilution standards and their corresponding known copy numbers. The known copy numbers for the oligonucleotide dilution standards were calculated from the original concentration of the synthetic oligo and molecular weight (see below). The copy numbers of the serial dilution samples were then derived by substituting the respective Cp values into the equation of the standard logarithmic curve. Logarithmic curves were required to have an r² value of at least 0.96 for quality control; otherwise, the data from that run would not have been used. Additional analysis was performed using R statistical software (R Core Team, Vienna, Austria) and the ggplot2 package (Elegant Graphics for Data Analysis, Verlag, NY) and ANOVA calculations.

Results

A high throughput, high sensitivity KSHV viral load assay

The impact of assay volume.

We hypothesized that a larger volume assay would be more sensitive than a smaller reaction volume assay. To test this hypothesis, the sensitivity of a larger volume assay requiring 96-well plates was compared to that of a small volume assay suitable for traditional 96-well or high throughput 384-well plates. The absolute LOD was set as 100 copies per ml. The large-volume assay (94.5 μL) was more sensitive than the small-volume (18.9 μL) assay for both EBV and KSHV because more of the extracted DNA could be added to each QPCR reaction (Figure 1A). The small-volume assay (red) recovered less input than the large-volume assay. This was because the larger volume assay accommodated five times as much input from a fixed DNA extraction volume as the small volume assay (with both standard 2x QPCR SYBR mix and the same pipetting robot). The median difference between reaction volumes was 0.89 * log10 copies with a ^95%CI of 0.63 – 1.2. This difference was significant (p ≤ 0.05, by analysis of variance (ANOVA)) and was larger for the EBV assay than the KSHV assay for reasons unknown. All data were normally distributed as assessed by Shapiro-Wilk statistics. The variance between the two assays (large in 96- and small in 384-well plates) was not significantly different based on the F-test. Most hardware can handle small and large volumes with similar accuracy.

Figure 1: Recovery of input virus by assay volume and instrument accuracy.

(A) Box and whisker plot comparing the sensitivity of the smaller volume (18.9 μL) assay in red with the larger volume (94.5 μL) in blue for two primer pairs EBV and KSHV. The starting point in each case is the same 1 mL sample, i.e., a fixed amount of virus, processed through DNA isolation and eluted in 100 μL. The vertical axis shows the number of copies recovered. The bolded horizontal line within the box plot indicates the median of the data set. The box itself represents the interquartile range (IQR). The ends of the whiskers represent the minimum and maximum values that are not considered outliers. Outliers are not shown. The absolute limit of detection was set to 100 cps/ml (dash-dotted line). (B) Box and whisker plot comparing the sensitivity of the small (CAS-9) liquid handling robot, in green, versus the large (Tecan) liquid handling robot, in black, using the same small (18.9 μL) volume assay. The vertical axis shows the number of copies recovered.

Adding five times more volume into the qPCR assay resulted in a greater than five-fold improvement in sensitivity because DNA purification by column or magnetic beads places a lower limit on the final elution volume, which is 100 μL. A higher elution volume is more efficient and dilutes contaminants, such as residual ethanol, from any washing steps, which can otherwise lower PCR efficiency.

The impact of pipetting accuracy.

Pipetting accuracy translates into assay reproducibility and sensitivity. Two robotic pipetting systems, the CAS-1200 (now Qiagen) and the Tecan Freedom Evo, were compared for pipetting efficiency and reproducibility using the small-volume assay. This Tecan design used fixed tips, and the instrument costs about five times more than the smaller benchtop robot. The variance within the Tecan data set generated was less than that within the data set generated from the CAS-1200 setup plates by the F-test to compare two variances. The median difference was 0.35 log10 with a 95%CI of 0.11 – 0.60. This was significant (p ≤ 0.007 by ANOVA) (Figure 1 B). The average LOD for the EBV small-volume viral load assay for the CAS-1200 robot (shown in green) was 1,600 cps/mL, compared to 600 cps/mL for the Tecan robot (shown in black). For KSHV, the pattern repeated itself with a LOD of about 3,200 cps/mL for the CAS-1200 compared to 1,000 cps/mL for the Tecan robot. In sum, when high accuracy is required, as in comparing differences between samples of less than 10-fold, investments in diagnostic-grade hardware pay off. If only differences between samples of greater than one log10 are scientifically meaningful, “personal” automation equipment suffices.

The impact of PEG enrichment.

PEG can concentrate viruses from large volumes. Poisson statistics stipulate that greater than three mL input is needed for a theoretical LOD of one copy per mL. Hence, larger overall input volumes yield more sensitive viral load assays. While larger volumes of body fluids can be collected, and five mL blood draws are routine, manual or automated liquid handling and DNA extraction are centered around 2 mL tubes. Eppendorf’s introduction of 5 mL tubes and a centrifuge to accommodate larger volumes has allowed for an optimized workflow to develop. Adopting earlier protocols for other viruses (Annunziata et al. 2002), 1 mL 40% PEG 8,000 was added 4mL samples and nutated overnight at 4°C to yield a 1mL resolved precipitate to feed into automated extraction. Known biohazardous specimens, such as HIV-positive samples, were supplemented with 10mM Tris/2% Triton X-100 Buffer before PEG incubation and centrifugation. This step did not affect the recovery of viral DNA. Across different concentrations, the measured copy numbers were about one log higher, starting from a four-fold greater input volume (Figure 2). The mean log difference was 0.806 log10 copies across the three runs, and the difference when considering all replicates was significant (p≤ 10⁷ by ANOVA). When considering the different input volumes of 1 mL versus 4 mL, the mean log difference was 0.204 log10 copies /mL. Thus, PEG pre-incubation of a larger input increased DNA recovery and improved assay sensitivity but perhaps not as dramatically as expected for the 1 mL to 5 mL increase in input. Operationally, PEG precipitation increased turnaround time by twenty-four hours.

Figure 2: PEG-based Enrichment.

Average copy number recovered in samples incubated with PEG (Blue) versus samples with no incubation in (Red) for the same input. Each data set on the horizontal axis, A-C, represents a separate run with different amounts of input virus. The vertical axis shows the number of copies recovered. The bolded horizontal line within the box plot indicates the median of the data set. The box itself represents the interquartile range (IQR). The ends of the whiskers represent the minimum and maximum values that are not considered outliers. Outliers are not shown.

Combining different primers

All primer pairs yield slightly, but not majorly different, amplification products that can be distinguished from each other (Table 1). DNA from three virally infected human cell lines, BCBL-1, Namalwa (Nam), and BC-1, containing known targets for these primers, were used to assess primer amplification products. All primer pairs yielded slightly different amplicons that could be distinguished from each other based on their amplicon size, as assessed using automated electrophoresis (Figure 3A). As expected, all cell lines yielded a 134 bp amplification product with ERV because they were of human origin. As a KSHV mono-infected cell line, BCBL-1 additionally yielded a 198 bp PCR product when amplified by the KSHV-specific primers. The EBV mono-infected cell line, Namalwa, which carries a single integrated copy of EBV, yields a 100 bp PCR product using the EBV-specific primers. Unlike the other two cell lines, BC-1 is co-infected with both KSHV and EBV. As a result, DNA from the cell line yields PCR products for both viruses, a phenomenon not observed in the other two cell lines, which only amplify a single viral target. The non-template control (NTC) yields no products.

Figure 3. Endpoint PCR analysis.

(A) Automated electrophoresis was conducted on QPCR products from three virally infected human cell lines: BCBL-1, Namalwa (Nam), and BC-1. Also shown is the non-template control (NTC). Amplicon size is measured in base pairs (bp) with 1500 bp as the upper limit marker (UM) and 25 bp as the lower limit marker (LM). (B) Microfluidics-based electrophoresis of a ~100 bp EBV amplicon (32 run time/s) was shown to elute after our smaller 70 bp oligo control amplicon (28 run time/s). Excess primer or primer dimer is around 22 run time/s by the lower limit (LM). “UM” indicates the upper size limit that the gel matrix can resolve. (C) Correspondence between melting temperature and amplicon size as measured by microfluidics-based electrophoresis. “a.” points to the QPCR amplification product from the sample (“s.1,” “s.2”) or positive control (“pos”), “b.” to that of the control oligo (“O), “c.” to excess primer. No products are seen in the NTC lanes. Relative molecular weights (“mw”) are shown in lane 1. “UM” and “LM” refer to the upper and lower resolving limits of the gel matrix, respectively.

Amplification products are easily distinguished from primer dimers by their melting temperature and size on agarose-gel or microfluidics-based electrophoresis (Figure 3 B, C). In the case of EBV, as shown here, the Tm difference between the sample-derived product and oligos-standard derived product is 83°C – 78°C = 5 °C. Melting temperatures can be determined to be within ±1°C. A melting curve is not available when TaqMan probes are used as a means of detection. Using these methods, we typically evaluate positive reactions outside the assay’s linear range. Borderline and suspected false NTCs were run on a microfluidics gel electrophoresis device that has high sensitivity and does not require an intercalating dye. Thus, both single-stranded primers and double-stranded products are detected. Synthetic oligonucleotide contamination could be detected at around 50 base pairs (bp) on the gel, while the KSHV target could be detected at around 240 bp.

An Extraction Spike-In Control Lowers the Rate of False Negatives

Viral load assay improvement hinges on simultaneously reducing the false negative and false positive detection rates while maintaining the most extensive linear range possible. False-negative results arise not only from a lack of sensitivity but also from assay failures anywhere in the workflow. These can be systematic or random. For instance, the level of PCR inhibitors varies among analytes, and pipetting errors can happen anytime. At the LOD, extraction efficiency contributes as much to overall sensitivity as a primer and PCR reaction design. To factor this into our analysis, we developed the FLY spike-in control for our KSHV viral load assay. It consists of a piece of Drosophila melanogaster DNA (see methods) cloned into a plasmid, which we add at an equal concentration to every sample before extraction. The FLY primers target this region, allowing us to control DNA extraction efficiency. In contrast to GAPDH, Actin, or ERV, it does not depend on the background DNA in the sample. Thus, it applies to swabs that collect viruses, such as those used to collect SARS-CoV-2 during the COVID-19 pandemic, from specimens including, but not limited to, cells, cell supernatant, urine, and saliva.

Complete extraction failures are easily distinguished from successful extractions (Figure 4A). Using FLY, extraction outliers can be excluded from further processing and downstream analysis as these samples would yield unreliable results and skew the KSHV copy number. If we include these samples in the analysis, there would be false positive data that would compromise our results, as the assay’s sensitivity is no longer guaranteed. Complete extraction failures are infrequent. Often, sample batches contain inhibitors that lead to a systematic bias in extraction efficiency, such as differing degrees of red blood cell lysis. By including the FLY control, we add that information as a confounder in our multivariate analysis for normalization to avoid such skewing (Figure 4B).

Figure 4. Examples of a spike-in control and batch normalization.

(A) This is a gray violin plot with the data points overlayed as red dots. The raw Cp values are shown on the vertical axis, and random QPCR plate identifiers are shown on the horizontal axis. Each plate contains 24 samples. The black arrows indicate samples for plates 717 (n=1) and 735 (n = 2), where the FLY extraction control failed. (B) Batch normalization across multiple PCR plates using Cp’ values The raw Cp values are shown on the horizontal axis, and the batch-normalized Cp’ values are on the vertical axis. For each plate (indicated by different colors), at least two data points, typically four, are used to calculate a standard curve.

Unless predetermined numbers of purified cells or PBMC are used, e.g., one million, extraction efficiency introduces random error, sometimes by up to a factor of five. This can be observed through the distribution of CT values for the FLY target. As long as this variation is random and the magnitude of variation remains constant across a batch of samples, bias will not be introduced.

Including an Endogenous Human Gene Improves Normalization Accuracy

Once extraction efficiency is controlled, the next step is to normalize the amplification efficiency and the amount of input material. This is easily achieved for cell-associated viruses, where viral load is reported as copy number per microgram of DNA or copy number per one million cells. The human endogenous retrovirus 3 (hERV-3) is present at one copy per haploid genome (Yuan, Miley, and Waters 2001). Thus, the absolute copy number of ERV, as determined by digital PCR, allows for calculating an overall conversion factor of assay efficiency. This means that regardless of the QPCR chemistry used, one million cells input into a KSHV viral load assay should always read out as one million copies of ERV. Hence, introducing ERV into any KSHV viral load assay allows for standardization across assays, irrespective of where and when they are performed.

ERV-based standardization is superior to GAPDH-based normalization of reverse transcriptase efficiency in RT-QPCR assays, as the level of any “housekeeping” gene can change under some culture conditions (Ruijter et al. 2013). For PBMCs, the copy number of ERV DNA per cell remains constant unless all cells in the populations synchronously enter mitosis. However, the ERV copy number per cell for tumor cells is not necessarily the same. Most tumor cells are no longer diploid, and their chromosomes are often unstable. ERV-based normalization is not applicable for acellular analytes, as the amount of cellular DNA in saliva, plasma, urine, or cell supernatant depends on the underlying degree of cell lysis, which is sample storage and disease-specific.

Most KSHV and EBV DNA is Not Virion Encapsulated

Viruses incorporate mechanisms of protecting their viral nucleic acids from degradation, often in the form of a protective protein capsid or lipid envelope (Allander et al. 2001). A high genome copy number to particle ratio and DNA to infectious unit ratio are common for herpesviruses. They likely represent circulating DNA (ctDNA) from a lysed tumor or lytically replicating cells. This is also true for KSHV. To determine the abundance of non-human DNA in a specimen (Figure 5A) Here, 0 μl, 2.5 μL, 5 μL, and 10 μL DNase I was added before automated nucleic acid extraction. The results revealed that the copy number of circulating KSHV DNA was indeed susceptible to DNase in a dose-dependent manner, suggesting that this DNA is non-encapsulated. At the optimal DNase concentration, all human ctDNA in the sample was degraded; however, KSHV DNA was not (Figure 5B). This experiment established a free-to-encapsulated DNA ratio of approximately 100:1 for KSHV.

Figure 5: Elimination of Non-Virion DNA.

(A) Three concentrations of DNase (Promega) or mock control were added to samples that were concentrated with PEG. (B) A density plot of 80 primer pairs that were designed to span the entire length of the viral genome. KSHV (top), but not human DNA (bottom), is protected from DNase I digestion. (C) Individual Primer pair performance and non-template control (NTC). KSHV-positive human specimens were treated with DNase (blue) or mock (red). Technical triplicates for all primers were measured using a standard QPCR detection level (40-CT).

To test the hypothesis that there was no biased extrusion of some regions of KSHV DNA, rather than the whole genome, our set of approximately 80 primer pairs spanning the entire viral genome was utilized, not to measure differential RNA transcription as published (Dittmer 2003), but to show equivalent ctDNA and encapsulation across the length of the genome. Figure 5C shows the raw detection values for independent QPCR reactions from the same input sample. DNA corresponding to every primer position, each located in a different viral ORF, was detectable and equally affected by DNase treatment. Most of the viral primers in this array yielded comparable results as indicated by similar CTs. By contrast, multiple primers targeting human genes failed to amplify the DNase-treated material. Despite similar design parameters, similar predicted Tm, similar amplicon lengths, and different real-time QPCR primer pairs, they have slightly different amplification efficiencies (K_eff). Hence, not all signals were perfectly equal.

Real-Time QPCR Matches Digital PCR

There are two types of dPCR: chip-based dPCR, where the sample is partitioned into microwells with the aim of one target molecule (or not) per well using microfluidics, and digital droplet PCR, where the sample is emulsified into oil droplets with the aim of one target molecule per droplet. The measurement output is the same: the number of single positive reactions per sample. One can then calculate the number of molecules per μL based on Poisson statistics. Note that because of the nature of the Poisson statistics, the copy number determination is most accurate when 63.21% of events are positive (independent of technology). Fewer positive events increase the error rate (s.d.), as the variance equals the mean for a Poisson distribution. Thus, if only 1% of wells are positive, the error rate is ~10%, which is less accurate than the error rate qPCR. This can lead to seemingly absurd reporting, where the error estimate is much greater than the measurement, e.g. 10±100 copies per μL. Hence, we use dPCR exclusively to calibrate the qPCR assay.

A comparison of digital PCR to qPCR shows a very good correlation and correspondence between the two platforms. Figure 6 compares measurements on defined cell numbers of varying input and clinical PBMC samples (1 Million input). In the case of the single copy gene human ERV target, cell counts, absolute digital PCR, and real-time QPCR results correlate to r²=0.99, n = 9 for cell lines (Figure 6 A). Note that the blue robust regression line and gray ^95%CI are calculated based on the cell line data. ERV is an, albeit expensive, alternative to cell counting, including ERV in the KSHV viral load assay guards against degraded or miscounted samples, such as D240002. Normalizing to the ERV qPCR signal is more accurate than cell counting for reporting DNA genome copy number per cell equivalent (typically 1 Million). A similar correlation also holds for EBV; however, the two cell lines contain different amounts of EBV genomes (Figure 6 B). Namalwa cells have one integrated copy of the EBV genome per cell, similar to ERV. BC-1 cells have multiple extrachromosomal copies of the EBV genome plasmid (Cesarman et al. 1995).

Figure 6. Correspondence of real-time QPCR and digital PCR.

Shown are the DNA copy numbers per μL as ascertained by digital PCR on the vertical axis (in log 10 scale) and the raw Cp values for the same sample on the horizontal axis for (A) EBV, (B) ERV, and (C) KSHV as well as clinical samples (D24). Also shown are the linear regression lines and ^95%CI in gray. The regression line is calculated based on cell lines with known cell numbers.

For KSHV, no cell line with just a single copy of the virus integrated into every cell exists. All stable KSHV cell lines contain multiple episomal copies of the viral plasmid. Absolute episome numbers center around a cell-line-specific mean (Bigi et al. 2018; Chiu et al. 2017), but there is substantial variation among cell lines (Figure 6 C). Using digital PCR and Poisson statistics is the only means to obtain consistent absolute DNA copy numbers for KSHV.

In sum, digital PCR using the same TaqMan primers as real-time qPCR represents a new tool for KSHV viral load determination. It has three different use cases. First, providing an absolute DNA copy number for body fluids, such as plasma or saliva, where an internationally agreed-upon standard and reference material is not available. For KSHV, no diagnostic standards of any kind exist. Second, providing a high-sensitivity assay for KSHV detection since individual molecules are counted. Digital droplet PCR is considered more sensitive than regular PCR since the assay volume is much smaller. Third, providing high accuracy for reasonably abundant viral loads, as the number of replicates (typically greater than 5000, as compared to 5 for qPCR) is much larger.

Discussion

KSHV can lead to KS, PEL, MCD, and acute complications, such as KS-IRIS or KICS (reviewed in (Yarchoan and Uldrick 2018; Cesarman et al. 2019)). Multiple viral genes, in conjunction, are necessary and sufficient to cause KS (Sin et al. 2024). In PEL, viral gene expression is continuously required for tumor cell survival (Godfrey et al. 2005; Bigi et al. 2018); however, the time between KSHV infection and clinically apparent KS disease is highly variable. In classic KS, viral infection precedes skin lesion development by decades. Many who are KSHV-infected never experience clinical disease. In endemic areas, the majority of the population is KSHV-infected. KS disease prevalence and onset are variable, even in areas with broad childhood transmission. Some who become KSHV infected develop KS as children, others as adults, and the majority not at all. This pattern of delayed disease onset is common to all cancers and herpesvirus infections in general, including EBV-associated Burkitt and diffuse large B-cell lymphoma. For viral cancers, decades separate exposure from disease. Currently, no predictive biomarker exists for impending KS onset. No prospective biomarker for monitoring KS response to therapy exists either.

Under particular circumstances, the connection between KSHV exposure and disease manifestation becomes more direct. Under these circumstances, the incubation time shortens. (A) In endemic regions, co-infections and environmental factors increase viral loads and hasten pediatric KS analogous to EBV-associated pediatric Burkitt lymphoma (Nalwoga et al. 2023; Sabourin et al. 2021; Oluoch et al. 2020; Nalwoga et al. 2020; Nalwoga et al. 2018; Wakeham et al. 2013). (B) In the setting of uncontrolled HIV infection, the incubation time between KSHV exposure and KS disease is shortened to three to ten years (Gao et al. 1996; Martin et al. 1998). HIV-associated immune deficiency before KSHV acquisition tends to shorten the time to KS development. However, in those scenarios’ multiple exposure events in high transmissibility settings may also contribute to this acceleration of disease emergence. End-stage AIDS KS and associated KAD correlate directly with high HIV viral load and low to absent CD4 counts. (C) In the setting of iatrogenic KS, the level and type of immune suppression regulate KS disease; either reducing immune suppression or changing the type of immune suppressive regimen can induce KS regression (van Leeuwen et al. 2010; Stallone et al. 2005; Krown et al. 2012; Ji et al. 2017). To summarize, immune competency curbs KS disease but not KSHV DNA copy numbers, except in edge cases.

KSHV-infected persons shed the virus intermittently in nasal secretions and saliva (Koelle et al. 1997; Vieira et al. 1997; Blackbourn et al. 1998). In saliva, KSHV DNA copy numbers are much higher than in blood but not as high as EBV or HCMV levels in saliva (Newton et al. 2018; Jacobson et al. 2009). The range of KSHV DNA copy numbers is extremely variable; in general, persons with high levels of KSHV also shed the virus for more days within a given observation window (Mbulaiteye et al. 2004; Nalwoga et al. 2020; Pauk et al. 2000; Bender Ignacio et al. 2016). In endemic areas, both men and women shed KSHV in saliva (Phipps et al. 2014). On average, even in the context of HIV infection and even in KSHV endemic countries, half of KSHV seropositive individuals do not have detectable KSHV DNA in their oral cavity. Not every KS patient sheds KSHV, either. Replication in the oral cavity is essential to KSHV infection and transmission; however, it does not correlate with KS disease or systemic (plasma) DNA copy number.

In pediatric KS, the KSHV DNA copy number seems to correlate with different patterns of disease (El-Mallawany et al. 2019). In other KAD, KSHV DNA copy number also shows some dependency on disease status. For instance, PEL presents with consistently high KSHV viral load in blood and effusions (Lurain et al. 2019). In MCD, KSHV genome copy number in plasma correlates with disease episodes (Bower et al. 2011; Jary et al. 2018; Ramaswami et al. 2021). In KICS, the KSHV genome copy number in plasma is disease-defining (Hansen et al. 2022; Caro-Vegas et al. 2020; Tamburro et al. 2012; Uldrick et al. 2010; Polizzotto et al. 2016). A limitation of many of these findings is that no distinction was made between encapsulated virus and circulating viral tumor DNA. We assume, based on lytic proteins expressed in the lesion, that the high KSHV plasma viral load originates from virions released by replicating virus in the disease lesions; however, cell death is typically concurrent with high levels of KSHV virion release, leading to a spike in unencapsulated, extruded viral DNA as well.

DNase I treatment confirmed that in human samples, a large portion of the KSHV copy number signal stemmed from cell-free circulating DNA, analogous to circulating tumor virus DNA, rather than virion-encapsulated and presumably infectious, linear viral genomes. We estimate the copy number to infectious units’ ratio to be 1:100 to 1:1000 for human samples. It will be interesting to explore whether total or virion DNA is a better potential biomarker.

Multiple studies have tried to link KSHV viral load to KS severity at presentation, progression, or response to therapy. Most of these studies were not powered to detect anything more than general trends or large effects. Hence, no clear conclusions can be drawn. The viral load data of the two most extensive KS trials remains outstanding (Krown et al. 2020; Hosseinipour et al. 2018). We noted that the KSHV genome copy number at presentation does not correlate with lesion count or disease severity, CD4 count, or HIV viral load, except for participants with very advanced AIDS-KS, where we could not exclude concurrent KAD (Moorad et al. 2023). The diagnostic value of measuring KSHV viral load in saliva or blood, thus, may be to uncover oblique and obscured KAD in KS patients who do not respond to cART alone. That in itself, however, is of immense clinical value since KAD rather than skin KS lesions seem to drive mortality in people living with HIV and in KSHV endemic areas.

This study aimed to improve the sensitivity and specificity of a KSHV DNA copy number detection. We explored SYBR green and related DNA intercalating dyes. These detect double-stranded products at each cycle, whereas TaqMan and hydrolyzable probes result in the accumulation of fluorescence (Tuomi et al. 2010). SYBR is cheaper but has not been validated for digital PCR. The relative merits of sensitivity and specificity for virus and viral quasi-species detection have been discussed previously (Papin, Vahrson, and Dittmer 2004). The experience with SARS-CoV-2 confirmed that TaqMan probe failures can result in selective undercounting of specific variants, which poses a problem in detecting rapidly evolving RNA viruses but is not as relevant to more evolutionarily stable DNA viruses.

Optimization of QPCR conditions with regard to both annealing temperature, primer sequences, and the use of a calibrator oligonucleotide standard on each plate increased the sensitivity and reliability of the assay. Increasing the sample and PCR reaction volume increased assay sensitivity. This result mirrors the experience with HIV viral load assay. Here, single-copy and sub-single copy assays use as much as one hundred milliliters of blood to detect free virus or single virus-infected cells (Palmer et al. 2003; Bruner et al. 2019; Stuelke et al. 2020). Precipitating the virus (and cell-free DNA aggregates) with PEG from a larger sample volume before DNA purification increased the sensitivity, similar to other studies (Colombet et al. 2007; Kaaden et al. 1971).

Note that manufacturers have changed ramping time capabilities, the composition of PCR buffers, and the properties of polymerases over calendar time. When KSHV was discovered, unmodified Taq-polymerase was the standard, which required a three-step protocol (55°C, 72°C, 92°C) for a maximum of thirty cycles. Today, real-time QPCR is conducted as a rapid two-step protocol (62°, 94°). Today’s polymerases are more heat resistant; hence, denaturation temperature and time could be increased. Today’s modified and differently sourced polymerase can extend primers at 60°C instead of 72°C. The primers used here were validated to anneal between 60°C and 64°C with negligible loss of efficiency.

Assay specificity is not a problem for KSHV (and other herpesviruses), which allows reliable extrapolation of viral genome copy number determination based on QPCR. These herpesvirus genomes are DNA-based and stable. Herpesviruses do not exist as quasi-species. Sequence variation in core genes is minimal within the host and across populations (Moorad et al. 2022; Marshall et al. 2022; Santiago et al. 2021). The genome was large enough to place one hundred QPCR primers and probes with identical annealing temperatures.

Assay sensitivity, however, is a problem for KSHV genome copy number determination based on QPCR, since KSHV accumulates to much lower levels than any other herpesvirus in body fluids and does not plaque on cultured cells (Nalwoga et al. 2023; Jacobson et al. 2009). Thus, no method exists to quantify the infectivity of clinical isolates, and therefore, there is no absolute standard to calculate genome-to-pfu ratios. The digital PCR assay described here and validated previously (Bigi et al. 2018), provides a means to attain absolute DNA genome copy numbers; however, it is based on cell-associated KSHV plasmids that vary in number by cell line. The equivalent of the Namalwa cell line with one integrated genome copy per haploid human cell genome does not exist for KSHV as it does for EBV.

Targeting PCR primers to repeat regions has been reported and is used to detect many infectious agents (Hofmann et al. 2015). The principal idea is that by targeting multiple copies, a higher rate of PCR efficiency (Lewis and Metcalf 1988) is attained. For a ten-repeat target, twenty amplification products are generated by cycle one of the reactions instead of two amplification products generated for a single copy target. Instead of a maximum theoretical rate of 2^{cycle number}, 10^{cycle number} and higher can be achieved. This effect is exploited in assays targeting the EBV Bam H W region (Ryan et al. 2004; Tan et al. 2020). Circulating EBV isolates contain five to eight W repeats (Tierney et al. 2011). There are two disadvantages to this approach. First, repeats are unstable even though their sequence is conserved. The limit of detection for a high repeat number strain is lower than for a low repeat strain, as is the effective amplification efficiency (Sanosyan et al. 2017). Hence, the viral loads for the two samples cannot be directly compared. Second, repeat regions generate complex secondary structures. It is unknown how these behave during PCR. There is the possibility for “super-stable” intermediates that would not melt as efficiently and self-priming that would artificially inflate the apparent primer concentration in the reaction. Either effect would lower the reproducibility and quantitative nature of real-time PCR.

One can debate the relative merits of single and multiplex QPCR. We found today’s automated pipetting robots accurate enough that an internal normalizer for PCR efficiency within a tube is no longer required. The primary source of variation for clinical material seems to be extraction efficiency and co-purifying impurities such as glycogens or lysed red blood cells during storage and processing at the collection site. We found that PCR reagent prices no longer constrain volume or the number of reactions one could afford. Instead, the biggest concern for multiplex QPCR on viruses and other infectious agents is that large amounts of one virus, read QPCR target, can alter the sensitivity for a second target in the same tube by competing for reagents. We routinely observe that EBV is present a 100 – 100,000 higher copy number than KSHV in the same sample. Hence, we use parallel single plex QPCR, whereby the sample is split after DNA extraction, and each PCR tube only has a single primer pair, typically in technical duplicates. Of note, FDA-approved quantitative assays, e.g., for HIV, are all conducted as single-plex reactions, while the FDA-approved multi-respiratory virus is an endpoint PCR assay. Maintaining accuracy and linear range across all combinations of all possible virus concentrations in multiplex reactions would be exceedingly difficult.

How much sensitivity is needed to develop a biomarker of reasonable positive predictive value for KS? Positive predictive value depends on the prevalence of the disease in the at-risk population. Thus, the viral load accuracy needed to have a clinical impact depends on the different types of KS. In endemic regions, every adult is reactivating KSHV with some frequency, independently of overt skin KS. KS is the most common cancer in these regions. Here, KSHV reactivation may be induced by SARS-CoV-2 infection (Lambarey et al. 2024). In high-risk populations, such as MSN with and without concurrent HIV infection (Salyards et al. 2023; Jary et al. 2020), KS disease and KSHV prevalence is much lower, even if elevated, compared to the general population. A similar calculation can be made for transplant recipients of Mediterranean descent in Turkey and the Arabian Peninsula. Here, KS disease and KSHV seroprevalence are elevated. Thus, a (i) KSHV viral load, (ii) advanced Age, (iii) levels of immune suppression, and (iv) perhaps levels of specific cytokines can become a predictor of occult KS or KS-associated KAD that require more intensive treatment than cART alone (Epedlegui et al. 2023; Openshaw et al. 2023). What is needed to define a clinically useful nomogram based on these factors is a cutoff for clinically actionable changes in KSHV plasma viral load. Right now, such a cut-off exists only for KICS, not KS.

Data Availability Statement

No data sets were generated in this review, and methods manuscript Primers, probe sequences, and detailed procedures are presented in the methods. Code and figure data are available on request.