High plasma EBV-DNA load and positive EBER status associated with viral recurrence and persistent infection in early treatment of lymphoma

The relationship between plasma EBV-DNA load (PEDL) and Epstein–Barr virus (EBV)-encoded small RNA (EBER) during the early treatment of lymphoma remains unclear. We explored discrepancies in PEDL and variables associated with EBER and evaluated the consistency between EBER and qualitative analysis of PEDL (qPEDL). Serial measurements of PEDL were performed to determine the dynamic changes of PEDL in early treatment of lymphoma. As a result, the median PEDL of non-Hodgkin’s lymphoma NKT cell subtype (NHL-NKT) was higher than that of non-Hodgkin’s lymphoma B cell subtype (NHL-B), the median PEDL of extranodal NK/T cell lymphoma (ENKTCL) was higher than that of diffuse large B cell lymphoma (DLBCL), and the median PEDL of EBER positive was higher than that of EBER negative. Age, Ki-67 ≧ 80%, Bcl-2 ≧ 80%, p53, and qPEDL were related to EBER. The PEDL could distinguish NHL-B, DLBCL, NHL-NKT, and ENKTCL from other lymphoma subtypes. EBER-positive patients spent more time with viral “turn negative (TN)” and “continuous positive (CP)” and less time with viral “continuous negative (CN).” The median PEDL of CP was higher than that of TN. In conclusion, although EBER affects the levels of PEDL in general, it has poor concordance with qPEDL. Our results show, for the first time, that high PEDL and positive EBER present a strong association with viral recurrence and persistent infection in the early treatment of lymphoma.


Introduction
Lymphoma, the most common non-epithelial malignancy of the head and neck region, can be divided into two types, Hodgkin's lymphoma (HL) and non-Hodgkin's lymphoma (NHL) [1]. Epstein-Barr virus (EBV), also referenced as human herpesvirus 4, is a ubiquitous DNA virus belonging to the gamma subfamily of herpesviruses. The normal life cycle of EBV involves the infection of lymphocytes and some epithelial cells [2].
EBV is the first human oncogenic virus ever identified, and EBV infection has been associated with several malignancies [3]. While its contribution to disease development and progression in each lymphoma subtypes is not fully understood, EBV has shown potential as a tumor biomarker, predictor of outcome, and therapeutic target [4]. This characteristic is used to diagnose and monitor after therapy for EBV-associated diseases. For instance, plasma EBV-DNA load (PEDL) can be used as a noninvasive biomarker for diagnosis of EBV-positive Meichun Zeng and Qingjun Jia these authors contributed equally to this research. HL [5]. Furthermore, estimation of PEDL may serve as an excellent noninvasive tool to monitor disease extent and response to therapy [6].
In recent years, PEDL assays are increasingly being employed for EBV association in lymphomas. PEDL can be used as a noninvasive biomarker for the diagnosis of EBVpositive HL [5]. Furthermore, measurements of PEDL may be useful in identifying children with Burkitt lymphoma [7]. Studies have shown that high pretreatment PEDL is a poor prognostic marker in HIV-associated, EBV-negative diffuse large B cell lymphoma (DLBCL) [8]. Additionally, during treatment and follow-up, PEDL should also be monitored as a marker of extranodal natural killer/T cell lymphoma (ENK-TCL) [9]. Researches have reported that DLBCL and peripheral T cell lymphomas (PTCL) are most often associated with detectable PEDL [10]. Although the relationship between EBV and lymphoma risk is thus very robust, the association of parameters of EBV infection, such as PEDL levels, with EBER status in patients with different lymphomas is less clear-cut.
EBER detection by EBER-ISH is the current gold standard for EBV status evaluation. Studies concluded that EBER positive (EBER pos ) should be viewed as a diagnostic marker of more severe condition requiring more aggressive treatment [11]. However, a clear cutoff for positivity has not been defined [12]. Therefore, the clinicopathological features and prognostic value of EBER pos in patients with lymphoma remain controversial [13]. Furthermore, previous reports demonstrated that EBV-DNA positivity was an effective prognostic marker for patients with certain lymphoma subtypes; also, EBER pos was involved in a worse prognosis for DLBCL [14]. Nevertheless, studies also showed that EBER pos greatly improved prognosis compared to EBER negative (EBER neg ) among younger patients with angioimmunoblastic T cell lymphoma (AITL) [15]. Consequently, what role does EBER status play in the prognosis and OS of lymphoma, and the more subtle relationship between EBER status and PEDL need to be explored and investigated in the future.
In the present study, we demonstrated discrepancies in PEDL and variables associated with EBER status and evaluated the consistency between EBER status and qualitative analysis of PEDL (qPEDL). We further performed the diagnostic ability of PEDL levels for several subtypes. Serial measurements of PEDL for patients within one year of diagnosis were performed to determine the dynamic changes of PEDL in early treatment of lymphoma.

Study population
We reviewed the medical records of 1,910 patients who diagnosed as lymphoma at our hospital between May 2020 and May 2022. Baseline blood samples of these patients were collected in the first 3 days after admission to the hospital. The diagnosis of lymphoma was based on the criteria of the 2017 World Health Organization classification, and all pathological biopsies were double-blinded and reviewed by at least two pathologists. All of the patients were treated with rituximab plus chemotherapy or chemotherapy alone. The research protocol was approved by the institutional review boards or research ethics committees at participating sites. All patients provided written informed consent before enrollment.

Quantitative analysis of PEDL
Plasma samples collected at diagnosis were used for PEDL testing. We extracted DNA from plasma samples using EBV-PCR Fluorescence Quantitative Diagnostic Kit (Shen Xiang Gene Co. Changsha, China). Amplification data compiled by the ABI Applied Biosystems 7300 system were analyzed using the Sequence Detection System software developed by ABI Applied Biosystems. We calculated the targeted concentration of pretreatment PEDL based on the standard curve. The lowest boundary of test sensitivity was 500 copies per milliliter. Samples with values less than 500 copies per milliliter were EBV-DNA-negative, including potential subjects with no copies and subjects with values of 1 to 500 copies per milliliter.

EBER detection by in situ hybridization (ISH)
After being completely fixed in 10% buffered formalin, tissue samples were dehydrated by gradient alcohol and embedded with paraffin. Three-micrometer sections were cut from tissue blocks dewaxed in xylene, treated with proteinase K, and hybridized with a digoxigenin-labeled probe. EBER in situ hybridization (EBER-ISH) (EBER PNA Probe, ISH-7001UM; ZSGB-BIO, Beijing, China) was performed according to the manufacturer's instructions using a fluorescein-conjugated EBER oligonucleotide probe and the purified IgG fraction of a mouse monoclonal anti-fluorescein antibody. Studies demonstrated that fewer EBER pos cells were considered to indicate nonpathogenic latent infection and might have no effect on disease progression or prognosis [16]. Therefore, we defined EBER-positive neoplastic cells per high-power field (HPF) exceeded 50 (> 50/HPF) as EBER pos . Samples with scattered (the number of positive neoplastic cells in the highest density region was less than 50/HPF), inadequate or equivocal positive neoplastic cells (a cutoff value < 50/HPF) were as EBER neg .

Follow-up and definitions
We completed the follow-up by reviewing the patients' medical records from May 19, 2020, to May 10, 2022, and calculated the time intervals of corresponding PEDL tests. Patients who completed PEDL tests once a month for 5 months or twice a month for 4 months within one year were included.

Statistical analysis
Data were presented as numbers (percentages) for categorical variables. The Pearson's chi-square statistical test was used to explore differences in variables between males and females. The Mann-Whitney U test and the Kruskal-Wallis rank test were performed to compare differences in the median PEDL. Univariate and multivariate logistic regression analyses were used to determine variables associated with the EBER status of subtypes. All variables with P < 0.1 in univariable analysis were kept in the multivariable analysis by using forward selection. Multivariable logistic regression was used to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for EBER status (1 = positive, 0 = negative) adjusted by age, Ki-67 ≧ 80% (1 = yes, 0 = no), Bcl-2 ≧ 80% (1 = yes, 0 = no), p53 (1 = positive, 0 = negative), and qPEDL (1 = positive, 0 = negative). All statistical tests were two-sided, and the statistical significance was established at P < 0.05. The concordance between EBER status and qPEDL was evaluated by the kappa coefficient (k). Agreement was classified into categories: poor (k < 0.4), moderate (k = 0.41 to 0.75), and good (k > 0.76). Predictive cutoff values of PEDL for lymphoma subtypes were determined using receiver operating characteristic (ROC) curve analysis and survival curve comparison using log-rank (Mantel-Cox) test. Statistical analyses were performed with IBM SPSS version 20 for Windows software program (IBM Corp, Armonk, NY, USA) and GraphPad Prism 5.

Comparative analysis of pretreatment PEDL in lymphoma subtypes
The pretreatment PEDL was detected in 414 patients with a positive rate of 30.02% (414/1,379). Figure 2a shows a statistically significant difference between NHL-B and NHL-NKT. In fact, the median PEDL of HL was 5,111 copies/ml, the median PEDL of NHL-NKT was 3,406 copies/ml, and it was higher than that of NHL-B (1,831 copies/ml) (P < 0.0001). More specifically, the median PEDL of ENKTCL (6,631 copies/ml) was higher than that of DLBCL (1,773 copies/ml) (P < 0.005) (Fig. 2b). Figure 2c also illustrates that EBER status was greatly related to the PEDL. As a result, the median PEDL of EBER pos group (5,096 copies/ml) was higher than that of EBER neg group (1,601 copies/ml) (P < 0.0001) (Fig. 2c).

Diagnostic ability of pretreatment PEDL levels using thresholds in lymphoma subtypes
To validate the discriminatory competence of PEDL for lymphoma subtypes, we performed ROC analyses when the sample size was greater than 300 (Fig. 3). Results revealed that the differences in PEDL of NHL-B, DLBCL, NHL-NKT, and ENKTCL were statistically significant (all P < 0.05), and the area under curve (AUC) was from 0.582 to 0.66 (Table 3 and Fig. 3a). Of note, using a threshold of 5, 070 copies/ml, the AUC of NHL-B was 0.742 (95%CI 69.04 to 79.32, P < 0.0001). Similarly, when the threshold was 1,300 copies/ml, the AUC of DLBCL was 0.85 (95%CI 81.02 to 88.91, P < 0.0001). The AUC of NHL-NKT was 0.737 (95%CI 68.74 to 78.56, P < 0.0001) with a threshold of 1,387 copies/ml and the AUC of ENKTCL was 0.738 (95%CI 66.26 to 81.34, P < 0.0001) with a threshold of 1,656 copies/ml (Table 3 and Fig. 3b).

EBER status associated with dynamic PEDL change during early lymphoma treatment
Finally, 467 patients met the criteria and were analyzed. Thereof, 340 were EBER neg and 127 were EBER pos . Overall, 302 were continuous negative (CN), 104 were turned negative (TN), and 61 were continuous positive (CP) (Fig. 4). In the CN group with EBER neg , the duration of undetectable PEDL was longer than that of EBER pos (P = 0.0011) (Fig. 4a). In the TN group with EBER neg , the duration of PEDL turned negative was shorter than that of EBER pos (P = 0.0052) (Fig. 4b). Similarly, patients in the CP group with EBER pos spent more time on continuous detectable EBV-DNA than that with EBER neg (P < 0.0001) (Fig. 4c). Additionally, the median PEDL of CP (3,090 copies/ml) was higher than that of TN (1,300 copies/ml) (P < 0.0001) (Fig. 4d). Furthermore, the median PEDL of CP (2, 650  copies/ml) was higher than that of TN (1,140 copies/ml) in the EBER neg group (P < 0.005), and the median PEDL of CP (6,480 copies/ml) was also higher than that of TN (1,440 copies/ml) in EBER pos group (P = 0.009) (Fig. 4e).

Discussion
In this retrospective study of 1,379 participants, we demonstrate that it is inconsistent between qPEDL and EBER status, although the median pretreatment PEDL in EBERpositive lymphoma patients is higher. Pretreatment PEDL in diagnostic samples can be a potential biomarker associated with EBV involvement in DLBCL. Moreover, the high The PEDL distributions grouped by lymphoma subtypes illustrated statistically significant differences. Figure 2a reveals that the median PEDL of NHL-NKT (3,406 copies/ml) is higher than that of NHL-B (1,831 copies/ml). Figure 2b shows that the median PEDL of ENK-TCL (6,631 copies/ml) is higher than that of DLBCL (1,773 copies/ ml). Figure 2c shows that the median pretreatment PEDL of the positive EBER group (5,096 copies/ml) is higher than that of the negative EBER group (1,601 copies/ml). In this study, "*" means a P value less than 0.05, "* *" means a P value less than 0.005, and "* * *"  Figure 3b indi-cates that using thresholds of 3.705 log 10 copies/ml, 3.114 log 10 copies/ml, 3.142 log 10 copies/ml, and 3.219 log 10 copies/ml the AUC of NHL-B, DLBCL, NHL-NKT, and ENKTCL was 0.742, 0.85, 0.737, and 0.738, respectively (all P < 0.0001)  [13]. In this research, we investigated the EBER status of four lymphoma subtypes (NHL-B, NHL-NKT, DLBCL, and PTCL) and found that five factors were related to EBER status: age, Ki-67 ≧ 80%, Bcl-2 ≧ 80%, p53, and qPEDL ( Table 2). Our results indicated that age was the only factor associated with EBER status of DLBCL, and it was similar to studies by Gao et al. [13]. We discovered that EBER status of these lymphoma subtypes was affected by various factors. Interestingly, qPEDL was only associated with the EBER status of NHL-B and NHL-NKT. Nevertheless, no statistical correlations were found between qPEDL and EBER status of DLBCL (n = 556) and of PTCL (n = 198). Consistent with these differences, studies demonstrated that qPEDL was detected in up to 72% of EBER-negative DLBCL patients [17]. Therefore, our data indicate that the relationship between EBER status and qPEDL should be combined with specific subtypes, rather than a general analysis.
The kappa statistic result between EBER status and qPEDL is only 0.19, corresponding to poor agreement and with an overall agreement rate of 68.02%. Therefore, while preliminary, our results (Table 2) demonstrate the poor consistency of them. Furthermore, more studies should be done to figure out the relationships between Ki-67 ≧ 80%, Bcl-2 ≧ 80%, p53, and EBER status. Moreover, there may be a few variations in the intensity of EBER staining under a certain level in tissues and the amount of plasma EBV-DNA, which may also contribute to the discrepant associations between the EBER status and qPEDL [18].
Previous studies reported that EBV-DNA positive was a more effective prognostic marker than EBER status for patients with certain lymphoma subtypes [17]. Therefore, we analyzed the pretreatment PEDL of different lymphoma subtypes and found that the median PEDL of NHL-NKT was higher than that of NHL-B. Furthermore, we explored finer subtypes and found that the median PEDL of ENKTCL was also higher than that of DLBCL (Fig. 2 a and b). Similarly, the median PEDL of the EBER pos group was higher than that of the EBER neg group (Fig. 2c). Our results tie well with previous studies wherein the PEDL in different EBER groups [19]. Our data are similar to studies from Shen et al. [19]; nevertheless, the median pretreatment PEDL is much lower than that of them. It is worth noting that they have used whole blood samples including circulating tumor cells and other EBV-infected cells. Nonetheless, plasma has been confirmed to be superior to whole blood for EBV-DNA quantification [20].
In this study, we investigated the pretreatment PEDL as a diagnostic marker for lymphoma subtypes. Results indicated Fig. 4 Associations of dynamic PEDL changes and EBER status during early lymphoma treatment. Figure 4 reveals that the PEDL of CN, TN, and CP has statistically significant differences according to different EBER status; furthermore, the PEDL of CP was higher than that of TN. Figure 4a shows that the duration of CN for patients with EBER negative (n = 238) is significantly longer than that with EBER positive (n = 64) (P = 0.0011). Figure 4b shows that the duration of TN for patients with EBER negative (n = 67) is significantly shorter than that with EBER positive (n = 37) (P = 0.0052). Figure 4c shows that the duration of CP for patients with EBER nega-tive (n = 35) is significantly shorter than that with EBER positive (n = 26) (P < 0.0001). Additionally, Fig. 4d shows that the median PEDL of CP (3,090 copies/ml) is higher than that of TN (1,300 copies/ml) (P < 0.0001) and Fig. 4e reveals that the median PEDL of CP (2,650 copies/ml) is higher than that of TN (1,140 copies/ml) in the EBER-negative group (P < 0.005), and the median PEDL of CP (6,480 copies/ml) is also higher than that of TN (1,440 copies/ml) in EBER-positive group (P = 0.009). "EBER pos " means EBER positive and "EBER neg " means EBER negative that these thresholds of PEDL for NHL-B, DLBCL, NHL-NKT, and ENKTCL were 5,070 copies/ml, 1,300 copies/ ml, 1,387 copies/ml, and 1,656 copies/ml (Table 3). When the threshold was less than 1,300 copies/ml, the AUC of DLBCL increased from 0.582 to 0.85; however, the sensitivity was 43.22% and the specificity was 70.95%. Similarly, despite the application of thresholds, the sensitivities and specificities of NHL-B, NHL-NKT, and ENKTCL were 75.88%, 77.22%, 88.89%, 44.81%, 43.36%, and 43.93%, respectively. Therefore, although the differences in each subtype are statistically significant, the specificity of using PEDL to distinguish NHL-B, NHL-NKT, and ENKTCL from other subtypes is poor. Studies have shown that measurements of EBV copy numbers in plasma may be helpful in identifying HL and endemic Burkitt lymphoma [5,7]. However, our results show that the accuracy of DLBCL is improved using a threshold, suggesting that there may be an optimized threshold to distinguish DLBCL from other subtypes.
In a recent report from Japan, results demonstrated the EBER-positive group of DLBCL had the worst prognosis [21]. Multivariate analysis indicated that EBER status was the only independent factor affecting OS among plasmablastic lymphoma (PBL) [22]. Koh et al. confirmed that patients with EBER pos had a lower 5-year OS rate than that with EBER neg [23]. Studies also reported that high PEDL was an adverse prognostic marker for patients with DLBCL [8]. Previous studies concluded that serial measurement of PEDL was strongly recommended [9], as any sign of an increase in PEDL indicated that relapse was inventible, even when a patient had been in remission for years [24]. However, so far, we have found no reports on the relationship between EBER status and the duration of dynamic PEDL changes among lymphoma patients during early lymphoma treatment.
Based on 467 patients, we described the EBER status and dynamic changes of PEDL in lymphoma patients at followup. We found that EBER status had some subtle relationships with the dynamic changes of PEDL. Our results indicated that patients in the EBER pos group had shorter CN time when compared with that in the EBER neg group (Fig. 4a). On the contrary, patients in the EBER pos group demonstrated longer TN time and CP time than that in the EBER neg group (Fig. 4 b and c). EBV viral proteins participate in the regulation of proliferation and metastasis [25]. Latent membrane protein 1 (LMP1), an EBV-associated oncogenic protein, can downregulate LMP1-triggered erythropoietin-producing hepatocellular receptor A4 (EphA4) in lymphoma [26]. Moreover, EBV microRNAs are expressed in patients with chronic lymphocytic leukemia (CLL) and correlated with OS [27]. However, whether the relationship between EBER status and dynamic changes of PEDL in lymphoma suggests a role for certain EBV-associated proteins is unknown and requires further investigation.
Studies confirmed that post-treatment EBV-DNA positivity could predict early relapse and poor prognosis for patients with lymphoma [28]. Furthermore, He et al. also demonstrated that elevated EBV-DNA after treatment indicated recurrence, metastasis, and unfavorable prognosis for EBV-associated nasopharyngeal carcinoma (NPC) [29]. In our study, pretreatment PEDL was compared between the TN and CP groups; we found that the median pretreatment PEDL of CP was 2.4 times higher than that of TN (P < 0.0001) (Fig. 4d). Results indicated that as expected, the median pretreatment PEDL of CP was 2.3 times higher than that of TN in the EBER neg group (P < 0.005); also, the median pretreatment PEDL of CP was 4.5 times higher than that of TN in the EBER pos group (P = 0.009) (Fig. 4e). These results confirmed that patients with EBER pos may be related to worse OS [14] and high pretreatment PEDL may also be associated with a significantly worse OS [30], and risk of relapse and mortality [31]. To sum up, we can conclude that the high PEDL and positive EBER status can predict viral recurrence and persistent infection.
In conclusion, the most significant finding of the present study is that EBER status is related to dynamic PEDL during early treatment of lymphoma. Meanwhile, although we have performed different subtypes analyses, EBER status and qPEDL show poor agreement, which is different from other studies [32,33]. Despite this, PEDL levels are higher in patients in the EBER pos group. We also conclude that higher pretreatment PEDL levels may indicate longer EBV-DNA duration in the plasma; thus, it may disclose high pretreatment PEDL levels that lead to inferior prognosis and the risk of relapse. Our results suggest that continuous monitoring is necessary for patients with EBER pos and high pretreatment PEDL at diagnosis. Moreover, the study is helpful to urge clinicians to provide individualized treatment for plasma EBV conversion and avoid viral recurrence and persistent infection in patients with EBER pos and high pretreatment PEDL. To some extent, this provides accurate risk stratification for EBV-associated lymphoma treatment and management.
To summarize, the retrospective nature of this study, with incomplete data, such as the follow-up that was conducted by reviewing the patients' medical records, is the major limitation. We set at least one month as an interval for PEDL testing when screening for follow-up targets, which resulted in the following limitations: First, we miss some patients who were occasionally token PEDL tests in other hospitals; second, the specific PEDL testing date is at random; consequently, the calculation of date for each patient may not be precise enough; and third, there is the fact that there is no established criteria for the proportion of EBER-positive cells in lymphoma tissues [16,34], which is a limitation when evaluating EBER status.