CD4-CD8 and CD28 Expression in T Cells Infiltrating the Vitreous Fluidin Patients With Proliferative Diabetic Retinopathy: Title and subTitle BreakA Flow Cytometric Analysis FREE

Proliferative diabetic retinopathy (PDR) is a major cause of adult blindness,and it is characterized by the appearance of neovascularization. The new vesselsare fragile and lack the normal barrier function, thereby permitting extravascularleakage of blood components. In the later stages, fibrovascular proliferativechanges may result in vitreous hemorrhage or tractional retinal detachmentthat often requires surgical intervention. The precise mechanisms involvedin the etiopathogenesis of PDR have not been elucidated, but retinal ischemiaseems to be crucial in the angiogenic stimulus, thus determining the synthesisof several growth factors, such as vascular endothelial growth factor.^{1 - 2}

There is growing evidence that leukocytes are involved in capillarynonperfusion, retinal vascular leakage, and endothelial cell damage in diabeticretinopathy.^{3 - 8} Moreover,leukocytes have a relevant role in the angiogenic and fibrotic processes thatoccur in PDR by means of the secretion of several cytokines and proteases.Several alterations in the properties of leukocytes have been reported indiabetes, such as decreased deformability,⁶ increasedactivation,³ and adhesiveness to vascular endothelium.^{4 - 8} Inaddition, the increased expression of adhesion molecules such as intercellularadhesion molecule 1^{8 - 9} has beenfound in retinal endothelial cells in diabetic retinopathy, facilitating retinalleukocyte stasis (leukostasis). Moreover, our group recently reported thatcellular adhesion and angiogenesis may be linked processes in diabetic patientswith PDR.¹⁰

Although there is much evidence that leukocytes participate in the etiopathogenesisof PDR, there is little information on the specific role of lymphocytes. Thevitreous body is an immune-privileged site protected from systemic circulationby the blood-retinal barrier.^{11 - 12} Thebreakdown of the blood-retinal barrier is a characteristic feature of PDRand, in consequence, could facilitate the passage of immune cells from systemiccirculation into the vitreous body. However, as far as we know, no studieshave investigated T-lymphocyte subsets in the vitreous fluid of diabetic patients.

In the present study, we have flow cytometry to analyze vitreous fluidfrom patients with PDR for the presence of CD4⁺ (helper/inducer)and CD8⁺ (cytolytic/cytotoxic) T cells, as well as the expressionof the costimulatory molecule CD28. In addition, these results were comparedwith those obtained in samples of peripheral blood from the same patients.Finally, we evaluated the relationship between the infiltrating T cells andboth the activity of PDR and the clinical outcome.

Twenty consecutive patients with type 2 diabetes mellitus (11 womenand 9 men; mean ± SD age, 62.82 ± 12.12 years) with PDR in whoma classic 3-port pars plana vitrectomy was performed were initially consideredin this study. Tractional retinal detachment was the main reason for performingvitrectomy in these patients. Six nondiabetic patients (3 women and 3 men;mean age, 47.33 ± 33.50 years) requiring vitrectomy for macular holes(n = 3) and subretinal membranes (n = 3) were also studied. Patients who hadundergone previous vitreoretinal surgery, had had a vitreous hemorrhage inthe preceding 2 months, or had received photocoagulation in the previous 3months were all excluded.

Retinopathy was graded intraoperatively by the same ophthalmologist,taking into consideration the presence of active neovascularization wheneverperfused preretinal capillaries were found and quiescent retinopathy whenevernonperfused vessels or fibrosis was present.

After vitrectomy, an ophthalmologic evaluation was systematically performedeach month for the first 3 months, and subsequently a bimonthy evaluationwas conducted. The overall follow-up was 8.3 ± 2.5 months (range, 3-12months).

Both venous blood and vitreous samples were collected simultaneouslyat the time of vitreoretinal surgery. Undiluted vitreous samples (0.5-1 mL)were obtained at the onset of vitrectomy by aspiration into a 1-mL syringeattached to the vitreous cutter (Series Ten Thousand Ocutome; Alcon Laboratories,Irvine, Calif) before the intravitreal infusion of balanced saline solutionwas started. The vitreous samples were transferred to a tube and deliveredto the laboratory at room temperature as rapidly as possible after collection,usually within 30 minutes.

Vitreous hemorrhage was identified when either macroscopic blood wasobserved in the vitreous or hemoglobin was detected within the vitreous fluid.For this purpose, in patients without macroscopic blood in the vitreous, hemoglobinlevels were measured in the vitreous fluid by spectrophotometry (Uvikon 860;Kontron Instruments, Zürich, Switzerland) according to the classic methodof Harboe¹³ for measuring plasma hemoglobinin micromolar concentration. This method has recently been validated,¹⁴ and in our hands the lower limit of detection was0.03 mg/mL. Only the patients in whom hemoglobin level was below the detectionlimit were considered free of intravitreous bleeding.

The protocol for sample collection was approved by the hospital ethicscommittee, and informed consent was obtained from patients.

Monoclonal antibodies (mAbs) against CD3, CD28, CD8, and CD4 conjugatedto fluorescein isothiocyanate, phycoerythrin, peridium chlorophyll protein,and allophycocyanin, as well as isotype IgG control mAb, were obtained (Becton,Dickinson and Co Immunocytochemistry Systems, San Jose, Calif).

Cells were analyzed immediately after blood collection by means of multiparameterflow cytometry according to standard protocols described by the manufacturer.Briefly, 4-color staining of surface markers was performed by incubation ofthe whole blood with saturating amounts of the different mAbs mentioned previously.Stained cells were then washed with staining buffer (phosphate-buffered salineplus 1% fetal calf serum plus 0.1% sodium azide), and red blood cells werelysed with an automated sample preparation system (FACS Lyse; Becton, Dickinsonand Co Immunocytochemistry Systems). After washing, cells were resuspendedand then analyzed with a double laser flow cytometer (FACS Calibur; Becton,Dickinson and Co Immunocytochemistry Systems). All samples were protectedfrom light during incubation throughout the procedure. Negative control isotypeIgG-matched mAbs were used for each case. Cell analysis was performed withan acquisition and analysis program (CellQuest; Becton, Dickinson and Co ImmunocytochemistrySystems).

Vitreous specimens were centrifuged at 2500 rpm for 15 minutes at 4°Cand the pellet obtained was resuspended in staining buffer (phosphate-bufferedsaline plus 1% fetal calf serum plus 0.1% sodium azide). Surface markers werestained by incubation with saturating amounts of the different mAbs. Stainedcells were then washed with staining buffer, resuspended, and analyzed withthe flow cytometer. All samples were protected from light during incubationthroughout the procedure. Negative control isotype IgG-matched mAbs were usedfor each case. Cell analysis was performed with an acquisition and analysisprogram (CellQuest; Becton, Dickinson and Co Immunocytochemistry Systems).In case of samples with vitreous hemorrhage, red cells were lysed as describedearlier. Results were expressed as percentages of positive cells.

Statistical analysis was performed with a microcomputer version of SPSS(SPSS Inc, Chicago, Ill). For comparisons between peripheral blood and vitreousfluid in the same patient, the Wilcoxon test was used. Levels of statisticalsignificance were set at P<.05. Data are presentedas median and range.

Three diabetic patients were excluded at the time of analysis becauseof the presence of nonviable cells in the vitreous fluid. Therefore, datafrom 17 diabetic patients were considered suitable for analysis. Six (35%)of the 17 patients with PDR and none of nondiabetic controls had vitreoushemorrhage. For the purpose of the study, data from diabetic patients withvitreous hemorrhage were analyzed separately.

T lymphocytes were detected in all diabetic patients with vitreous hemorrhage,in 6 (55%) of 11 diabetic patients without blood in the vitreous, and in noneof the nondiabetic patients. A flow cytometric analysis of both blood andvitreous fluid from a representative patient with PDR is displayed in Figure 1. The percentages of T cells obtained,as well as their phenotypical analysis in peripheral blood and in the vitreousfluid according to the presence or absence of vitreous hemorrhage, are presentedin Table 1. As expected, the percentagesof T cells, as well as the percentages of CD4⁺ and CD8⁺ inthe vitreous fluid and in the peripheral blood, were similar in patients withvitreous hemorrhage. In these patients, the percentage of CD4⁺ CD28⁻T cells was higher in the vitreous fluid than in the peripheralblood, but the differences were not of statistical significance (40.99% [24.32%-100.00%]vs 20.64% [10.34%-41.25%]). In addition, we did not observe any statisticaldifference in the percentage of CD8⁺ CD28⁺ T cells betweenthe vitreous fluid and the peripheral blood (21.05% [0.00%-45.21%] vs 19.85%[17.73%-68.93%]).

Patients without vitreous hemorrhage showed a slight, nonsignificantreduction in the percentage of CD4⁺ and a mild increase in thepercentage of CD8⁺ in the vitreous fluid in comparison with theperipheral blood (50.59% [21.05%-57.11%] vs 55.15% [36.42%-72.64%], and 49.40%[42.88%-78.94%] vs 44.85% [27.36%-63.57%], respectively). However, the percentageof CD4⁺ CD28⁻T cells in the vitreous fluid wassignificantly higher than in the peripheral blood (33.34% [20.75%-100.00%]vs 8.45% [2.43%-56.59%]; P = .02), and this differencewas observed in every patient (Figure 2).By contrast, CD28 was present in a greater proportion of CD8⁺ Tlymphocytes in the vitreous fluid than the peripheral blood, although in thiscase the differences were not statistically significant (67.32% [0.00%-100.00%]vs 44.16% [18.04%-66.31%]).

The relationship between the flow cytometric analysis and both the activityof retinopathy and clinical outcome is displayed in Figure 3. All patients with PDR and no vitreous hemorrhage who showedT lymphocytes in the vitreous fluid had quiescent retinopathy, and only 1of these patients developed early intravitreal bleeding. The remaining diabeticpatients continued free of rebleeding after a follow-up of 8 ± 2 months.By contrast, 3 of 5 diabetic patients without T lymphocytes in the vitreousfluid had early intravitreal hemorrhage.

Analysis of the vitreous fluid obtained from diabetic patients subjectedto vitreoretinal surgery is a useful means of indirectly exploring the eventsthat are taking place in the retina. However, there is growing evidence thatthe vitreous participates actively in the etiopathogenesis of PDR by meansof accumulating angiogenic factors such as vascular endothelial growth factoror by losing angiogenic inhibitors such as pigment epithelium–derivedfactor.^{15 - 17} Inaddition, an enhancement of intravitreous concentrations of several cytokineshas been reported,^{18 - 26} thusunderlining the importance of the inflammatory process in the developmentof PDR. In the present study, we investigated the presence of T-lymphocytesubsets in the vitreous fluid and compared the results with those obtainedin the peripheral blood at the time of vitreoretinal surgery. One of the majorproblems in any technique for studying these vitreous cells is to obtain anadequate number for analysis. Many cells in the vitreous fluid are alreadynonviable, and the remainder can disintegrate very quickly after collectionof the sample. Immunocytochemical techniques have been used to study immunecells infiltrating epiretinal membranes in patients with PDR,^{18 ,22} but,to our knowledge, flow cytometry has not been previously applied to immunophenotypingT cells in the vitreous fluid of patients with PDR. The experimental approachused (vitreous fluid and whole-blood immunofluorescence, performed immediatelyafter collection) allows us to simulate the in vivo scenario as closely aspossible. The second problem when vitreous fluid is analyzed is to excludevitreous hemorrhage. To circumvent this problem, the hemoglobin levels withinthe vitreous fluid were measured by spectrophotometry in all samples. Thus,only patients in whom hemoglobin was undetectable were considered free ofvitreous hemorrhage.

T lymphocytes were not detected in the vitreous fluid of nondiabeticpatients. This finding supports the concept that the disruption of the blood-retinalbarrier is crucial for permitting the access of inflammatory cells into thevitreous body. By contrast, T lymphocytes were detected in all patients withblood in the vitreous, and the percentage of T cells as well as the patternof CD4⁺ and CD8⁺ was very similar to that obtained inthe peripheral blood. This was not a surprising result because intravitrealcellularity probably reflects peripheral blood entering the intraocular cavity.It must be emphasized that this event was found at micromolar concentrationsof hemoglobin and demonstrates that a little bleeding is sufficient to changethe pattern of T cells detected within the vitreous of diabetic patients withPDR.

In patients without detectable hemoglobin, T lymphocytes were detectedin only 55% of cases, and the T-cell count was lower than that obtained inpatients with blood in the vitreous. The substantial variations obtained inthe percentages of T cells and their percentages within the total number ofviable elements in the vitreous suggest that the participation of the immunesystem in the etiopathogenesis of PDR could be different in each patient.In addition, in these patients we found a slight deficit of CD4⁺ (helper-inducer)T cells and a mild increase of CD8⁺ (cytolytic-cytotoxic) T cellsin relation to peripheral blood. However, the most important finding of thepresent study was the significant enhancement of CD4⁺ CD28⁻detected in the vitreous fluid in comparison with peripheralblood in patients without vitreous hemorrhage. Activation of T lymphocytesrequires 2 signals. The first signal, induced by the interaction of the antigen–majorhistocompatibility complex with the T-cell receptor, determines the antigenspecificity, whereas the second costimulatory signal determines the activationthreshold and the functional outcome of the antigen-specific activation.^{27 - 28} CD80/CD86–CD28/CTLA-4 is themost important and best-studied costimulatory pathway. CD80 and CD86 moleculesare expressed on activated antigen-presenting cells and bind to their ligandsCD28 and CTLA-4.^{29 - 31} CD28costimulation of human T cells increases the expression of the intrinsic cellsurvival factors Bcl-X_Land interleukin 2 (IL-2), which correlatewith enhanced resistance to apoptosis.³² Inaddition, it has been demonstrated that ligation of CD28 results in the activationof protein kinase B,³³ a key mediator of growthfactor–induced cell survival.³⁴ CD28also up-regulates expression of both the IL-2 receptor and several other cytokinesor chemokines (ie, IL-4, IL-3, interferon γ, tumor necrosis factor,granulocyte-macrophage colony-stimulating factor, IL-8, and RANTES [regulatedon activation of normal T cells expressed and secreted]), and also modulatesexpression of the chemokine receptors CCR5, CCR1, CCR4, CXCR1, and CXCR2.^{35 - 36} Furthermore, manipulation of theCD28 pathway of costimulation can prevent the initiation of an autoimmuneresponse, as well as suppress an ongoing autoimmune process.³⁷ Forthese reasons it could be speculated that, although we have detected a highpercentage of CD4⁺ CD28⁻in the vitreous fluidof diabetic patients, these T cells are lacking appropriate activation. However,the high intravitreous levels of several cytokines and inflammatory mediatorsdetected in diabetic patients with PDR argue strongly against this hypothesis.Recently it has been demonstrated that CD4⁺ CD28⁻clonesare CD28-costimulatory independent and exhibit a full agonist signaling activationpattern, prominent T_H1 cytokine production, and a prolonged responsein vitro.³⁸ Long-term memory CD4⁺ CD28⁻cells produce high amounts of interferon γ and maximallyup-regulate interferon γ and IL-12Rβ2 chain expression in the absenceof costimulation. In addition, they have an increased survival after apoptoticstimuli, probably related to their persistent lack of CTLA-4 surface expression.^{38 - 40}

The reason for the enhancement of CD4⁺ CD28⁻observedin the vitreous fluid of diabetic patients is as yet unknown. One possibilityis that costimulation-independent autoreactive CD4⁺ cells undergoactivation in the periphery by the mechanism of molecular mimicry or bystanderactivation. Subsequently, activated CD4⁺ CD28⁻cellsmay preferentially tend to leave the bloodstream, migrate through the impairedblood-retinal barrier, and initiate an inflammatory response within the eye.In this regard, costimulation-independent activation is particularly importantfor antigen recognition within the posterior chamber of the eye, where competentantigen-presenting cells are sparse. Alternatively, the resistance to theapoptosis reported for CD4⁺ with lack of CD28 surface expression^{38 - 40} could be a reliableexplanation for its increased percentage within the vitreous fluid of diabeticpatients.

We have observed a tendency toward a lower proportion of CD8⁺ CD28⁻T lymphocytes in the vitreous fluid of diabetic patients withoutvitreous hemorrhage compared with peripheral blood. Similar results have beenreported by Svenningsson et al⁴¹ in the cerebrospinalfluid of patients with either multiple sclerosis or central nervous systeminfectious disease. There are conflicting results in the literature regardingthe role of CD28 expression in determining functional characteristics of CD8⁺ T lymphocytes. Original data supporting primarily suppressor functionsfor the CD28⁻populations^{42 - 43} haverecently been challenged.⁴⁴ It is thus notpossible at present to assign a lack of CD28 expression as being specificfor cells with suppression function among CD8⁺ T lymphocytes, andso a reliable marker for this functional subset is still lacking.

One interesting finding observed in the present study is the relationshipdetected in patients without blood in the vitreous between flow cytometricanalysis and both the activity of retinopathy and the outcome in terms ofearly bleeding after vitrectomy. Thus, all patients in whom T lymphocyteswere detected in the vitreous fluid had quiescent retinopathy, and only 1patient developed early intravitreal bleeding. Therefore, it seems that Tcells infiltrating the vitreous cavity have a protective role in the outcomeof PDR. In this regard, it should be emphasized that the neuroprotective effectof autoimmune T cells has recently been reported.^{45 - 46} Inaddition to anti-inflammatory cytokines like IL-10 or transforming growthfactor β, neurotrophic factors could be potential candidates to explainthe protective effect of T cells on PDR outcome.^{47 - 48}

In summary, in patients without intravitreous hemorrhage in whom T cellsare detectable within the vitreous fluid, we found a high percentage of CD4⁺ CD28⁻in comparison with peripheral blood. In addition,all of these patients had quiescent retinopathy and their outcome was betterthan that in either patients with blood in the vitreous or patients in whomintravitreous T cells were undetectable. The different pattern of T cellsidentified in the vitreous fluid of diabetic patients with PDR requires furtherfunctional characterization, which should provide us with a better understandingof events involved in the development of autoimmune response in diabetic retinopathyand would help us in the search for effective treatment for this disease.

Corresponding author and reprints: Rafael Simó, MD, DiabetesResearch Unit, Division of Endocrinology, Hospital General Universitari Valld'Hebron, Pg Vall d'Hebron 119-129, 08035 Barcelona, Spain (e-mail: rsimo@hg.vhebron.es).

Submitted for publication March 19, 2003; final revision received September25, 2003; accepted October 14, 2003.

This study was supported by grants from the Ministerio de Ciencia yTecnología (PM-99-0136), Madrid, Spain; Instituto Carlos III (G03/212and C03/08), Madrid; and Novo Nordisk Pharma SA (01/0066), Madrid.

We thank Carlos Mateo, MD, for his generous collaboration in the collectionof vitreous samples. In addition, we thank X. Montalban, MD, for his technicalsupport and Michael Willi for his assistance with manuscript preparation.