CXCR4, CXCR5 and CD44 May Be Involved in Homing of Lymphoma Cells into the Eye in a Patient Derived Xenograft Homing Mouse Model for Primary Vitreoretinal Lymphoma

Background: Primary vitreoretinal lymphoma (PVRL), a rare malignancy of the eye, is strongly related to primary central nervous system lymphoma (PCNSL). We hypothesized that lymphoma cells disseminate to the CNS and eye tissue via distinct homing receptors. The objective of this study was to test expression of CXCR4, CXCR5, CXCR7 and CD44 homing receptors on CD20 positive B-lymphoma cells on enucleated eyes using a PCNSL xenograft mouse model. Methods: We used indirect immunofluorescence double staining for CD20/CXCR4, CD20/CXCR5, CD20/CXCR7 and CD20/CD44 on enucleated eyes of a PCNSL xenograft mouse model with PVRL phenotype (PCNSL group) in comparison to a secondary CNS lymphoma xenograft mouse model (SCNSL group). Lymphoma infiltration was evaluated with an immunoreactive score (IRS). Results: 11/13 paired eyes of the PCNSL but none of the SCNSL group were infiltrated by CD20-positive cells. Particularly the choroid and to a lesser extent the retina of the PCNSL group were infiltrated by CD20+/CXCR4+, CD20+/CXCR5+, few CD20+/CD44+ but no CD20+/CXCR7+ cells. Expression of CXCR4 (p = 0.0205), CXCR5 (p = 0.0004) and CD44 (p < 0.0001) was significantly increased in the PCNSL compared to the SCNSL group. Conclusions: CD20+ PCNSL lymphoma cells infiltrating the eye co-express distinct homing receptors such as CXCR4 and CXCR5 in a PVRL homing mouse model. These receptors may be involved in PVRL homing into the eye.


Introduction
Primary vitreoretinal lymphoma (PVRL) is a severe cancer of the eye, which infiltrates the vitreous and retina. It is a rare non-Hodgkin lymphoma (NHL) with an estimated incidence of approximately 0.05/100.000 [1][2][3]. However, PVRL is the most common into the spleen of recipient mice. Intrasplenic (i.s.) transplanted mice developed CNS lymphoma manifestations and in case of PCNSL retinal infiltration, creating a PCNSL and PVRL homing model. To our knowledge, this is the first homing model for PCNSL and PVRL established so far and created the opportunity for the present study (unpublished manuscript under review [36]). Therefore, we have chosen to evaluate the homing receptors CXCR4 and CXCR5 already known to play a role in PVRL pathogenesis as well as CXCR7 and CD44, that have not been analyzed in PVRL so far. The objective of this study was to test expression of the homing receptors CXCR4, CXCR5, CXCR7 and CD44 on CD20 positive B-lymphoma cells in the eyes using a newly established PDX PCNSL respectively PDX SCNSL mouse model.

CD20-Positive Lymphoma Cells Are Mostly Found in the Choroid in the PCNSL Group
Lymphoma cells from the established SCNSL PDX model without PVRL (SCNSL group) as well as from the PCNSL PDX model with PVRL phenotype (PCNSL group) were implanted i.s. into 10 and 13 recipient mice, respectively. The evaluation of staining for human CD20 showed that 0/10 paired eyes of the SCNSL group were infiltrated with CD20-positive lymphoma cells (Figures 1-4). Within the PCNSL group, 11/13 paired eyes showed positive staining for CD20, whereas 2/13 paired eyes were not infiltrated with CD20-positive lymphoma cells. The lymphoma cells were mostly localized in the choroid, showing positive staining in 11/13 paired eyes. The retina was infiltrated in 7/13 cases. Infiltration with few lymphoma cells was also observed in the sub-retinal and sub-RPE space. In 4/13 paired eyes CD20 positive cells were found in the ciliary body.
Int. J. Mol. Sci. 2022, 23,11757 3 of 15 We recently established a novel patient derived xenograft (PDX) PCNSL mouse model with PVRL phenotype, showing CD20 positive lymphoma cells in the retina of the PCNSL PDX model (unpublished, manuscript under review [36]). We hypothesized that CXCR4, CXCR5, CXCR7 and CD44 are expressed on CD20-positive lymphoma cells in the eyes of this PCNSL PDX model. PCNSL and SCNSL PDX models were established via intracerebral (i.c.) implantation of patient stereotactic CNS biopsies and then implanted into the spleen of recipient mice. Intrasplenic (i.s.) transplanted mice developed CNS lymphoma manifestations and in case of PCNSL retinal infiltration, creating a PCNSL and PVRL homing model. To our knowledge, this is the first homing model for PCNSL and PVRL established so far and created the opportunity for the present study (unpublished manuscript under review [36]). Therefore, we have chosen to evaluate the homing receptors CXCR4 and CXCR5 already known to play a role in PVRL pathogenesis as well as CXCR7 and CD44, that have not been analyzed in PVRL so far. The objective of this study was to test expression of the homing receptors CXCR4, CXCR5, CXCR7 and CD44 on CD20 positive B-lymphoma cells in the eyes using a newly established PDX PCNSL respectively PDX SCNSL mouse model.

CD20-Positive Lymphoma Cells Are Mostly Found in the Choroid in the PCNSL Group
Lymphoma cells from the established SCNSL PDX model without PVRL (SCNSL group) as well as from the PCNSL PDX model with PVRL phenotype (PCNSL group) were implanted i.s. into 10 and 13 recipient mice, respectively. The evaluation of staining for human CD20 showed that 0/10 paired eyes of the SCNSL group were infiltrated with CD20-positive lymphoma cells (Figures 1-4). Within the PCNSL group, 11/13 paired eyes showed positive staining for CD20, whereas 2/13 paired eyes were not infiltrated with CD20-positive lymphoma cells. The lymphoma cells were mostly localized in the choroid, showing positive staining in 11/13 paired eyes. The retina was infiltrated in 7/13 cases. Infiltration with few lymphoma cells was also observed in the sub-retinal and sub-RPE space. In 4/13 paired eyes CD20 positive cells were found in the ciliary body. and (G-I) of the ciliary body (marked with *). (A) shows the choroid infiltrated with CD20-positive primary CNS lymphoma cells (PCNSL+), which additionally express CXCR4 (triangles). (D,G) show retina and ciliary body infiltrated with CD20-positive cells (PCNSL+) without co-expression. Neither CD20-nor CXCR4-positive cells were found in the PCNSL negative (PCNSL-) group (B,E,H) and in the SCNSL group (C,F,I). ONL = outer nuclear layer, INL = inner nuclear layer, GCL = ganglion cell layer; scale bar 50 µm.

CD20-Positive Lymphoma Cells Co-Express CXCR4, CXCR5 and CD44 but Not CXCR7
We next evaluated co-expression of CD20 with CXCR4, CXCR5, CD44 and CXCR7. No co-expressing cells were found in the SCNSL group (Figures 1-4). Within the PCNSL group co-expressing cells were predominantly found in the choroid (Figures 1-4). These cells primarily co-expressed CD20/CXCR4 ( Figure 1) and CD20/CXCR5 ( Figure 2). CD20/CD44-positive cells were found sporadically ( Figure 4). Only a few cells were found to co-express CD20 and CXCR7 ( Figure 3). The same pattern was found in sections of the retina, even though fewer co-expressing cells were detected (Figures 1-4). The ciliary body showed little positive staining for CD20 (Figures 1-4). However, some of these cells coexpressed CXCR5 ( Figure 2).

CD20-Positive Lymphoma Cells Co-Express CXCR4, CXCR5 and CD44 but Not CXCR7
We next evaluated co-expression of CD20 with CXCR4, CXCR5, CD44 and CXCR7. No co-expressing cells were found in the SCNSL group (Figures 1-4). Within the PCNSL group co-expressing cells were predominantly found in the choroid (Figures 1-4). These cells primarily co-expressed CD20/CXCR4 ( Figure 1) and CD20/CXCR5 (Figure 2). CD20/CD44positive cells were found sporadically (Figure 4). Only a few cells were found to co-express CD20 and CXCR7 (Figure 3). The same pattern was found in sections of the retina, even though fewer co-expressing cells were detected (Figures 1-4). The ciliary body showed little positive staining for CD20 (Figures 1-4). However, some of these cells co-expressed CXCR5 ( Figure 2).

The CXCR5 Receptor Was Most Frequently Co-Expressed among All Examined Homing-Receptors
Analysis of the manually determined proportion of co-expressing cells out of all CD20+ cells identified CXCR5 as the most frequently co-expressed receptor (Supplementary Materials Figure S1).

Ciliary Body
In the ciliary bodies of the PCNSL group, 1/4 (25%) of the CD20-positive cells also showed co-expression with CXCR5. 2/9 (22%) of cells showed co-expression with CXCR4. CXCR7 was co-expressed in 4/19 (21%) of CD20-positive cells. No CD20-positive cells showed additional staining for CD44. We counted significantly lower infiltration with CD20-positive cells in the ciliary body compared to the retina (Supplementary Materials Figure S1).

CD20, CXCR4, CXCR5 and CD44 Are Expressed Significantly Higher in the PCNSL-Group Compared to the SCNSL-Group
We next used an immunoreactive score (IRS) to quantify expression of CD20, CXCR4, CXCR5 and CD44 in the eyes in the PCNSL group in comparison to the SCNSL group ( Figure 5). To determine the IRS, staining intensity and the percentage of positive cells were scored. We used this score as quality control for our immunoreactive double stainings as well as to evaluate overall expression levels of all cells in the eyes and not only of CD20 positive cells.

Choroid
In the SCNSL group, no CD20-positive cells were found, resulting in an IRS of 0. In contrast, the expression of CD20 in the PCNSL group showed a wide range of variation of IRS values from 0 to 9. The median IRS was 3, indicating that human CD20 positive lymphoma cells were exclusively detected in the PCNSL model. In comparison to the SCNSL group, CXCR5 (p = 0.0004), CXCR4 (p = 0.0205) and CD44 (p < 0.0001) were significantly higher expressed in the PCNSL group. CXCR7 showed weak expression in both groups, with an IRS median of 2 with no differences between the two groups (p = 0.7408) ( Figure 5).

Retina
Overall, the CD20 IRS values indicated that there were few weakly CD20 positive lymphoma cells in the retina. Again, positive staining for CD20 was only present in the PCNSL group. CXCR4 was moderately expressed in the retina; expression was significantly higher in the SCNSL group than in the PCNSL group (p = 0.0008). CXCR5 (p = 0.9706) and CD44 (0.0594) also showed a weak expression pattern in both the SCNSL and PCNSL groups and did not differ significantly. CXCR7 (p = 0.9706) showed weak to moderate expression in both groups with no significant difference ( Figure 5).

CD20, CXCR4, CXCR5 and CD44 Are Expressed Significantly Higher in the PCNSL-Group Compared to the SCNSL-Group
We next used an immunoreactive score (IRS) to quantify expression of CD20, CXCR4, CXCR5 and CD44 in the eyes in the PCNSL group in comparison to the SCNSL group ( Figure 5). To determine the IRS, staining intensity and the percentage of positive cells were scored. We used this score as quality control for our immunoreactive double stainings as well as to evaluate overall expression levels of all cells in the eyes and not only of CD20 positive cells. The IRS was significantly higher for CD20 (marked with asteriks *** p < 0.001), CXCR5 (marked with asterisks *** p < 0.001), CXCR4 (marked with asterisks * p < 0.05) and CD44 (marked with asterisks *** p < 0.001) in the PCNSL compared to the SCNSL group. (B) Box plots of IRS values in the retina of the PCNSL vs. SCNSL group. The IRS was significantly higher for CD20 (marked with asterisks ** p < 0.01) and CXCR4 (marked with asterisks *** p < 0.001) in the PCNSL compared to the SCNSL group. (C) Box plots of IRS values in the ciliary body of the PCNSL vs. SCNSL group. The IRS was significantly higher for CXCR5 (marked with asterisks ** p-value < 0.01) in the PCNSL compared to the SCNSL group. ns = not significant.

Ciliary Body
The CD20 IRS values revealed few CD20-positive lymphoma cells in the ciliary body in the PCNSL group and no CD20-positive cells in the SCNSL group. CXCR5 showed a weak expression pattern in the PCNSL group, being significantly higher than in the SCNSL group (p = 0.0075). The receptors CXCR4 (p = 0.1573), CXCR7 (p = 0.9355) and CD44 (p = 0.6809) showed no significant differences between the two groups ( Figure 5).

Discussion
In this study we investigated a novel CNS lymphoma PDX mouse model for a PVRL phenotype. We compared the eyes of a lymphoma PDX model derived from a PCNSL versus SCNSL patient stereotactic CNS biopsy regarding infiltration of human CD20 positive lymphoma cells and immunohistochemical expression of the homing receptors CXCR4, CXCR5, CXCR7 and CD44.  [39,40].

CD20+ Cell Infiltration
In our PDX mouse model, we found eye infiltration after heterotopic (intrasplenic) implantation of human PCNSL xenografts, supporting the concept of homing of CNSL cells to the brain and eyes. Here, CD20-positive lymphoma cells were found primarily in the choroid, but the retina and ciliary body were also infiltrated (Figures 1-4). Even though an infiltration of the choroid cannot be seen clinically, the choroid may function as a point of entry for lymphoma cells into the retina and is therefore an important anatomical structure to examine in PVRL. Interestingly, we found no infiltration in the vitreous, the anterior chamber, the iris, or the conjunctiva. These slight differences may result from the different methodology used for this mouse model. In this PDX model, DLBCL lymphoma xenograft cells from established PCNSL or SCNSL PDX were injected heterotopically into the spleen of NSG/NOG mice (unpublished manuscript under review [36]). In the case of PCNSL, this led to retinal infiltration, resulting in a homing model for PVRL. Our homing model therefore contrasts with other PVRL mouse models, in which lymphoma cells were injected directly intravitreally, i.e., orthotopically. We were able to show that 11/13 paired eyes of the PCNSL model but none of the paired eyes of the SCNSL model were infiltrated with CD20-positive lymphoma cells, confirming the PVRL phenotype.
In this study, we detected co-expression of CD20 and CXCR4 on lymphoma cells in the PCNSL group but not in the SCNSL group. Thus, our model recapitulates results of previous studies in primary lymphoma samples. When we evaluated the overall expression of CXCR4 in both groups using the IRS, CXCR4 was significantly more highly expressed in the retina in the SCNSL group, contrary to our expectations. However, CXCR4 can be expressed in healthy retina by photoreceptor cells, by the RPE as well as by endothelial cells [22]. In summary, CXCR4 was expressed by CD20-positive lymphoma cells in our PCNSL/PVRL model. Thus, the CXCR4-CXCL12 interaction could mediate the chemotaxis of CD20/CXCR4-positive lymphoma cells in PVRL [41,42].
Our results show that CXCR5 was co-expressed on CD20-positive lymphoma cells in the PCNSL group. In contrast, no co-expression of CXCR5 was detected in the SCNSL group. Of all the receptors examined, CXCR5 was most frequently co-expressed. For that reason, we suspect that the CXCR5-CXCL13 axis could be particularly relevant for homing of lymphoma cells and pathogenesis of PVRL and PCNSL.

CD20/CXCR7 Co-Expression
CXCR7 mRNA expression was associated with good prognosis in CXCR4 positive systemic DLBCL [49,50]. To our knowledge, it was not evaluated in PVRL so far. However, in a CXCR7 knockout mouse model of DLBCL, CXCR7 WT mice showed CNS infiltration as opposed to CXCR7 knockout mice, suggesting CXCR7 as an important receptor for CNS homing [49].
The results of our work show that in the choroid as well as in the retina and the ciliary body there were hardly any CD20-positive lymphoma cells which co-expressed the scavenger receptor CXCR7. However, we did not analyze CXCR7 mRNA expression. Thus, the significance of CXCR7 in homing of lymphoma cells to the brain and eye remains to be further evaluated.

CD20/CD44 Co-Expression
CD44 has not yet been examined in PVRL but is expressed by lymphoma cells and vascular endothelium in PCNSL and PTL [11,46,[51][52][53][54][55]. In PCNSL, lymphoma cells, which were often found perivascular, were CD44-positive [56]. Furthermore, CD44-positive B lymphocytes can infiltrate the white matter of the brain by binding to HA. This infiltration decreased when both the lymphocytes and the brain tissue were treated with either hyaluronidase or CD44 antibodies [51]. In PCNSL, expression of CD44 was associated with a shorter overall survival [56]. In PTL, CD44-expression seems to correlate with the late-stage of disease [55].
In our mouse model, CD44 was co-expressed only by a few CD20-positive lymphoma cells in the choroid and retina of the PCNSL group. Interestingly, CD44-positive cells were found in capillaries near CD20-positive lymphoma cells. We therefore assume that these cells may be CD44-positive monocytes or macrophages, which are involved in the extravasation of the lymphocytes by presenting chemokines, such as CCL4 and CCL5 [31][32][33]57]. It was shown that HA, the ligand of CD44 was expressed on vascular endothelium cells. Additionally, HA is also one of the main components of the extracellular matrix of the brain as well as the vitreous body of the eye [58]. Thus, CD44 positive lymphoma cells might migrate towards a HA gradient of the brain and eye. The CXCL12-CXCR4 axis can affect the expression and activation of CD44 [59]. In our study, the 15% proportion of CD20/CXCR4-positive cells in all CD20-positive lymphoma cells may have negatively impacted the expression of CD44 on lymphoma cells. CD44 remains an interesting candidate receptor for PCNSL and thus also for PVRL, which may well contribute to homing of lymphoma cells into the eye.
Together, our data for the first time demonstrate homing of human PCNSL, but not SCNSL, xenografts to the eye after heterotopic implantation and suggest that expression of chemokine receptors CXCR4, CXCR5 and CD44 may be involved in homing of lym-phoma cells into the eye.

Limitations
Although we used human CNSL xenografts, it must be emphasized that the data for this study using a PDX mouse model cannot be directly extrapolated to humans. In contrast to PVRL in humans [6], no lymphoma cells were present in the RPE in this model, even though a few lymphoma cells were found sub-RPE. This may result from specific representation of lymphoma cells by immunofluorescent staining of the mouse retina due to its slightly different morphology [60].
However, due to the scarcity of tumor material in PVRL, this mouse model creates the opportunity to expand human tumor material in vivo, allows us to specifically determine the exact localization of human CD20+ lymphoma cells in the eyes of PCNSL and SCNSL and is a valid model to investigate the significance of respective homing receptors for lymphoma manifestation in the CNS and eyes. Additionally, PDX mouse models unlike conventional xenograft mouse models preserve genetic heterogeneity over many passages. Therefore, a xenograft mouse model working with human lymphoma cells from patient samples offers a unique opportunity to study PVRL in vivo, creating the opportunity to identify mechanisms critical for homing of lymphoma cells that might be accessible for therapeutic intervention.

PDX Mouse Model
In this study we used a PDX PCNSL mouse model with a PVRL phenotype and compared it to a PDX SCNSL mouse model. These models were established by Isbell et al. (unpublished manuscript under review [36]). Tumor tissue from diagnostic stereotactic brain biopsies from patients with suspicion of PCNSL or SCNSL were implanted i.c. via the foramen postgleonidale into 4-6 weeks old recipient NSG mice (NOD NOD/Shi-scid/IL-2Rγnull; Charles River, France). DLBCL was detected in all patients, regardless of whether they were PCNSL or SCNSL. Monoclonality of lymphoma cells was demonstrated by FACS analysis of infiltrating lymphoma cells in the spleen of both PDX mouse models.
Secondary implantations were carried out after the depletion of mouse cells ((#130-104-694, Miltenyi Biotec, Bergisch Gladbach, Germany). A PDX was defined as established when stable growth was observed over at least 3 transplantation passages and regrowth from xenograft tumour tissue stored in liquid nitrogen was seen. Lymphoma cells from established PDX models were implanted i.s. into recipient NSG mice. Mice were euthanized after development of signs of disease. The eyes were enucleated and embedded into paraffin.

Primary and Secondary Antibodies
For indirect immunofluorescence staining, the following primary and secondary antibodies were used: Human CD20 (

Double Indirect Immunofluorescence Staining
0.5 µm thick sections were cut from paraffin embedded eyes and mounted on slides. The slides were deparaffined using graded alcohol solutions. After boiling in citrate buffer (pH 6.0) and rinsing in phosphate buffered saline (PBS) the slides were incubated with 5% fetal bovine serum. The staining with primary antibodies against CD20 combined with primary antibodies against CXCR4 (1:25)/CXCR5 (1:100)/CXCR7 (1:200) or CD44 (1:400) took place during incubation overnight at 4 • C. The next day, the slides were washed in PBS and then stained with matching secondary antibodies for 1h at room temperature. For evaluation, the staining was imaged with immunofluorescence microscope (Leica DMI 6000 B, Leica, Germany). On each section, CD20-positive cells and co-expressing cells in the choroid, retina and ciliary body were counted using Fiji/ImageJ [68]. For evaluation with the immune reactive score (IRS), representative shots of the eyes were analyzed as explained below.

Cell Counting
To indicate a proportion, CD20-positive and co-expressing cells were counted manually by using the cell counter of Fiji/ImageJ [68]. The proportion then was calculated by dividing the number of CD20 and homing-receptor co-expressing cells by only CD20-positive cells.

Evaluation with the IRS
For every representative image of the choroid, retina, and ciliary body an IRS was determined. Firstly, the staining intensity of positive cells was estimated and scored as 0 = no reaction, 1 = weak reaction, 2 = moderate reaction and 3 = strong reaction. The percentage of positive cells was determined and scored. 0% positive cells were scored as 0, less than 10% as 1, 10-50% as 2, 51-80% as 3 and more than 80% as 4. The IRS was then calculated by multiplying the scores for staining intensity and percentage of positive cells [69,70]. We considered an IRS greater than 0 to be positive.

Statistical Analysis
Statistical analysis of ordinal scaled and not normally distributed IRS values was performed with the Mann-Whitney-U-Test using GraphPad Prism version 8.0.2 for Windows, GraphPad Software, San Diego, California, USA, www.graphpad.com (accessed 1 October 2022). A p-value less than 0.05 was considered statistically significant.

Conclusions
In this study, we were able to show that CD20-positive lymphoma cells in a PCNSL PDX homing mouse model with a PVRL phenotype co-express the chemokine receptors CXCR4, CXCR5 as well as CD44 in comparison to a SCNSL PDX mouse model. These receptors may therefore be involved in homing of B-lymphoma cells into the eyes in PVRL. The expression patterns of these homing receptors and their ligands CXCL12, CXCL13 and HA should be further researched, in knock-out models and in knock-out DLBCL cell lines, for example. Furthermore, downstream investigations of homing receptors are of particular interest as well as alteration of these chemokines in the process of homing within this mouse model. Chemokines and the receptors themselves may help us to understand this disease better and therefore might introduce possible new targets for therapeutic approaches for PVRL in the future.