Cytokine Profile in Human Eyes: Contribution of a New Cytokine Combination for Differential Diagnosis between Intraocular Lymphoma or Uveitis

Primary intraocular lymphoma (PIOL), also called primary vitreoretinal lymphomas, often masquerades as uveitis. This misdiagnosis can result in subsequent brain involvement and oculocerebral lymphoma (OCL). In this study, we sought to characterize the helper T-cell type 1 (Th1)/Th2 cytokine profile in vitreous samples from patients with PIOL, OCL, uveitis and controls with non-inflammatory disease. Vitreous and aqueous humor samples from 87 patients with PIOL (n = 30), OCL (n = 12), uveitis (n = 34), and retinal detachment (RD) without hemorrhage (n = 11) were analyzed and their concentrations of interleukin (IL)-2, IL-4, IL-6, IL-10, interferon (IFN)-γ, and tumor necrosis factor (TNF)-α were determined by flow cytometric bead arrays (CBA). The IL-10 levels determined by CBA were compared with those by ELISA. IL-10 concentrations measured by CBA and ELISA were highly correlated. IL-2, IL-4, and TNFα were not detected in any sample. The only cytokine detected at a significant level in samples from RD vitreous was IL-6. The IL-10/IL-6 ratio, as previously reported, was slightly higher in PIOL than in uveitis samples, but not for all patients. Cytokine profiles from PIOL and OCL samples did not differ. The combination of the IL-10/IL-6 and IL-10/IFNγ ratios was highly informative for discriminating PIOL/OCL from uveitis samples and for therapeutic follow up of PIOL. This strategy might be very helpful as an initial screening to rule out PIOL in patients thought to have uveitis.


Introduction
Primary intraocular lymphoma (PIOL), also called primary vitreoretinal lymphomas [1], is a subset of primary central nervous system lymphoma (PCNSL) that initially presents in the eye, with or without simultaneous CNS involvement [2,3]. Most PIOL cases are related to high-grade extranodal non-Hodgkin, diffuse large Bcell lymphomas. A rapid definitive diagnosis is required for appropriate treatment to prevent the substantially worse prognosis associated with the spread of the disease to the brain. PIOL often masquerades as chronic uveitis and thus remains difficult to diagnose: a high degree of suspicion is required before testing [4,5]. A high level of IL-10 in pure vitreous or aqueous humor samples or an IL-10/IL-6 ratio greater than 1 in diluted or undiluted samples is considered indirect evidence requiring further diagnostic testing [6,7]. The exact cutoff for the IL-10 concentration or IL-10/IL-6 ratio may vary between laboratories, mainly due to differences in methods and conditions of sample harvesting and storage, techniques, and manufacturers of equipment and supplies, as well as the dilution (known or unknown) of the vitreous samples and the laboratory's own experience. These factors result in false-positive (11%) and false-negative (23 to 30%) PIOL diagnoses [6,8,9,10]. Because no systematic cytological analysis strategies currently discriminate between PIOL and uveitis in patients [11], a better characterization of the molecular microenvironment seems essential to identify new cytokine combinations as diagnostic markers. Accordingly, our aim here was first to compare the reference ELISA technique [6] to a multiplex-based cytometric bead array (CBA) technique that enables the simultaneous analysis of several cytokines without sample volume limitations, as only 25 to 50 mL is needed. Recombinant human IL-10 was used to compare the accuracy, sensitivity, and range of concentration analysis of these techniques, and IL-10 was measured in frozen human samples to confirm the robustness of both techniques. In the second part of this work, we analyzed the Th1/Th2 cytokine profile, including IL-10, IL-6 and IFNc, in vitreous and aqueous humor samples from patients with PIOL, oculocerebral lymphoma (OCL), uveitis, and retinal detachment (as controls). Cytokine ratios from the different groups of samples were compared and allowed us to define a new strategy for discriminating patients with PIOL or OCL from those with uveitis.

Patients and samples
Ethics Statement: this study was performed in accordance with the Declaration of Helsinki. The patients, who were recruited from the Ophthalmology department of the Pitié-Salpêtrière Hospital (Paris), all provided written informed consent after the nature of the study had been fully explained to them. The Pitié-Salpêtrière hospital review board approved the study. All data were treated confidentially.
The results involving vitreous samples included 60 patients (Table 1), distributed in four groups. The first comprised 17 patients with PIOL, the second 9 patients with OCL diagnosed by cytologic analysis of the vitreous and positive MRI for OCL, the third 23 patients with diagnosed uveitis, and the last (negative control) group, 11 patients with non-hemorrhagic retinal detachment. Vitreous humor specimens were obtained through standard three-port pars plana vitrectomy, as previously described [12]. Tissue culture medium (balanced salt solution, BSS) enriched with 10% fetal calf serum was added to the collection chamber to improve cell viability. An initial 250 mL specimen of pure vitreous humor was collected separately into a microtube for IL-10 quantification only; the remainder was diluted into BSS supplemented with 10% fetal calf serum, and 2 mL of this preparation were delivered immediately, in a syringe, to the Hematology Laboratory. The vitrectomy cutter was maintained within the vitreous humor to obtain a dilute specimen with minimal disruption. The remainder of the diluted vitreous humor specimen, about 20-30 mL, was harvested in a collection cassette and immediately processed for cytological and immunocytochemical analyses.
After an anterior chamber puncture during or before surgery, 100 mL of aqueous humor was taken from 27 patients for cytokine measurement. If performed before vitrectomy, the anterior chamber puncture was done under topical anesthesia with a 30gauge needle in the operating theater with a surgical microscope. If it was performed during the surgery or vitrectomy, the patient was under general or local anesthesia. Patients analyzed at diagnosis included a first group of 11 patients with PIOL, a second of 3 patients with OCL, and a third of 11 patients with uveitis. Kinetic analyses were performed for two other patients (patients #1 and #2), after a recurrence; both were treated intravitreously with methotrexate.

Cytokine assays
Samples were stored at 280uC, and cytokine assays were performed on freshly thawed samples. IL-10 concentrations were determined from minimum 100 mL vitreous samples, with a standard quantifiable sandwich enzyme immunoassay technique (QuantikineH; RD Systems, Abingdon, UK), as previously described [13].
For cytokine measurements by the cytometric bead array (CBA) technique, samples were assayed for IL-2, IL-4, IL-6, IL-10, TNFa, and IFNc (human Th1/Th2 CBA kit; BD Biosciences), according to the manufacturer's recommendations. Briefly, six capture bead populations with distinct fluorescence intensities and coated with cytokine-specific capture antibodies were mixed together in equal volumes: 50 mL of each sample and 50 mL of PE-conjugated detection antibodies were added to 50 mL of mixed-bead populations. The mixture was incubated for 3 hours at room temperature in the dark to form sandwich complexes. The beads were then washed with wash buffer, and data acquired with an LSR II flow cytometer (BD Biosciences). FACSDiva and FCAP software (BD Biosciences) were used for the analyses. When a value was negative, the theoretical limit of detection value was used for Log calculations.

Statistical analysis
The PIOL, OCL, and uveitis groups were compared by the Mann-Whitney non-parametric statistic test or the Student t-test. P,0.05 was considered significant.

IL-10 concentrations measured by CBA and correlation with ELISA
High levels of IL-10, usually measured by ELISA within pure vitreous or aqueous humor samples, are considered indirect evidence of a PIOL diagnosis [6]. Before using the CBA multiplex technology to analyze the Th1/Th2 cytokine profiles of PIOL, OCL, and uveitis patients, we compared the IL-10 results produced by ELISA and CBA. As Figure 1 shows, measurement by these two different techniques yielded IL-10 concentrations that were very highly correlated (R 2 = 0.994 to 0.999) for the recombinant human IL-10 provided with each kit. IL-10 concentrations were next analyzed in the vitreous samples from PIOL, OCL, and uveitis patients ( Figure 2). Again, the two techniques produced good correlations (R 2 = 0.701). Comparative Th1/Th2 cytokine profiles of PIOL, OCL, and uveitis samples Vitreous samples from 60 patients (Table 1) with PIOL (n = 17), OCL (n = 9), uveitis (n = 23) or retinal detachment (RD) without hemorrhage (n = 11) were analyzed for IL-2, IL-4, IL-6, IL-10, IFNc, and TNFa concentrations by CBA. IL-2, IL-4, and TNFa were detected only in a few samples and then with values near the threshold of detection (Figure 3). In the control RD samples, 82% (9/11) of cases had a significant level of IL-6, the only cytokine significant in this group. IL-6 was detected in 100% of the uveitis samples, IL-10 in 60%, and IFNc in 48%. Interestingly, the PIOL and OCL cytokine profiles did not differ significantly, except for IL-10: the mean IL-10 concentration in PIOL samples was 724.126676 pg/mL and in OCL samples, 247.416207 pg/mL (P = 0.009). As previously described, the mean IL-10 concentration was much higher for PIOL (724.126676 pg/mL) than for uveitis (12.61623 pg/mL) samples (P,0.001). Inversely, the mean IL-6 concentration was lower for PIOL (41.4635 pg/mL) than for uveitis (118462394 pg/mL) samples (P = 0.018). Nonetheless, some PIOL patients had low IL-10 levels and some uveitis patients high levels.
Since these overlapping samples were not necessarily for the same patient on each graph, we decided to combine the IL-10/IL-6 and IL-10/IFNc ratios on a single graph ( Figure 5). Using this strategy, we identified a cluster (upper right) containing exclusively PIOL and OCL samples. In the left parts of the graph, only 1 PIOL sample was colocalized with uveitis samples. We performed the same analysis on aqueous humor from PIOL, OCL, and uveitis patients and obtained similar results ( Figure 6A). As aqueous humor can be harvested several times from the same patient, we were thus able to measure cytokine ratios over time during treatment (i.e., intravitreous injections of methotrexate). Remission was identified by a decrease in both ratios (see Figure 6B), whereas relapse was correlated with a new increase back in the upper right quadrant of the graph ( Figure 6C).

Discussion
The incidence of PCNSL has multiplied more than tenfold over the past decade. The incidence of PIOL, a subtype of PCNSL, has also increased albeit more slowly, thus leading to increased concern about the diagnosis of this disease [14,15]. Moreover, PIOL is one of the most challenging masquerade syndromes, [5] especially for differential diagnosis with uveitis. The lack of information about this disease is mainly due to the limited number of patients diagnosed and the small amount of ocular fluid harvested. The principal reliable criterion for PIOL diagnosis is the presence of lymphoma cells in patient samples, but the cytological procedure is very delicate: lymphoma cells are particularly fragile ex vivo and sometimes difficult to find or to recognize; they can be analyzed only by highly trained hematologists. A high level of IL-10 in the vitreous of patients is also highly indicative of intraocular lymphoma but is not sufficient for a definitive conclusion [16]. This study describes the Th1/Th2 cytokine profile in undiluted vitreous and aqueous humor samples from PIOL patients, determined with a multiplex-based technology requiring only 25 to 50 mL of sample. First, we confirmed that IL-10 concentrations measured by CBA and ELISA were highly correlated, both for recombinant cytokine and human vitreous. Next, we observed that the results did not display either a conventional Th1 or Th2 profile: IL-2, IL-4 and TNFa were not detected in any sample. In the RD control group, the only cytokine detected at a significant level, and only in some samples, was IL-6. In contrast, as expected, samples from patients with PIOL and OCL both had high concentrations of IL-10; to a lesser extent, these samples also contained IL-6 and in some cases IFNc. Interestingly, we noted that the cytokine profile of samples from patients with OCL at diagnosis was similar to that of samples from PIOL patients, except that the OCL patients had a slightly lower IL-10 concentration. Although the mean level of IL-6 and IL-10 expression in uveitis vitreous samples differentiated them from tumor samples, overlapping was found in some cases. Indeed, 48% (11/23) of the uveitis samples had an IL-6 concentration in the same range as PIOL/OCL samples and 38% (10/26) of the PIOL/OCL samples an IL-10 concentration in the same range as uveitis samples, consistent with a previous report [17]. Adding the IL-10/IFNc ratio increased accuracy for PIOL diagnosis, but some samples still overlapped. Moreover, the exact cutoff ratio for a positive test result may vary between laboratories and by technique. The combination of the IL-10/IL-6 and IL-10/IFNc ratios was highly informative in its discrimination between PIOL/ OCL and uveitis samples. A PIOL/OCL specific cluster (100%,  25/25) was found and only 1 PIOL/OCL sample of the 26 (3.8%) was graphically colocalized with uveitis samples. In a recent paper, Ecker and colleagues [18] analyzed cytokines in the vitreous and aqueous humor of patients with posterior segment diseases and showed that the cytokine ratio was constant although cytokine quantities differed in the two locations. Similarly, we obtained the same cytokine ratios with aqueous humor as with vitreous. This is particularly important because aqueous humor is easier to tap and can be harvested several times during the disease. It could therefore be very useful for rapid prediction of patients' response to treatment, as reported recently by Kawamura and co-workers [19].
Accordingly, we recommend that ocular samples be frozen (280uC) quickly in aliquots after harvesting. The IL-10/IL-6 and IL-10/IFNc ratios in aqueous humor should then be routinely analyzed in aqueous humor from all patients with ocular inflammation at diagnosis and during treatment for PIOL/OCL, such as methotrexate and/or anti-CD20 monoclonal antibodies. We further encourage the use of multiplex technologies, such as cytometric beads arrays (CBA, BD Biosciences) that allow simultaneous measurement of several cytokines from a very low volume of liquid. Finally, the graphic representation of the combination of the Logarithmic (Log 10 ) conversion of IL-10/IL-6 and IL-10/IFNc ratios helps to cluster the PIOL/OCL versus uveitis patients at diagnosis, with the upper right cluster specific for PIOL/OCL patients.
In conclusion, the combination of the IL-10/IL-6 and IL-10/ IFNc ratios could be very helpful as an initial screening to rule out PIOL in uveitis patients. Whereas immunocytology and retinal histology remain the gold standard for confirming a diagnosis of PIOL, multiplex technologies are relatively simple to use for ruling this diagnosis out, they are accurate, and offer the opportunity to combine several markers for diagnosis and therapeutic follow up since only 25 to 50 mL of sample are needed to perform all these tests simultaneously.