Foliate Lymphoid Aggregates as Novel Forms of Serous Lymphocyte Entry Sites of Peritoneal B Cells and High-Grade B Cell Lymphomas Free

This article is featured in In This Issue, p.1

The omentum and mesentery are primarily adipose tissues associated with the gastrointestinal organs, harboring an unusual type of peripheral lymphoid tissues (1, 2). Although the leukocytes dwelling in these serous tissues constitute a distinct immunological compartment, they are fully integrated components of systemic immunity, including their extensive traffic to structured peripheral lymphoid organs during activation and differentiation (3, 4). In addition, the specific significance of omental protection during i.p. inflammations or in perforating abdominal injuries has long been recognized and appreciated (5). Although both the peritoneal and pleural cavities are important reservoirs for B1 B lymphocytes, the tissue characteristics of a local specialized microenvironment that allows the survival and interaction within serous lymphoid organs during immune responses and as peritoneal tumor dissemination sites have only recently been investigated in more detail (6).

As prototypic serous lymphoid organoids, the milky spots (MS) of the omentum were described first. In mice, MS are formed by the clustering of macrophages, dominantly B cells and, to a lesser extent, T cells, without compartmentalization into separate T and B cell zones characteristic of secondary lymphoid tissues (7). MS also function as lymphocyte exit ports from the systemic circulation (8).

More recently, similar leukocyte-rich gatherings (denoted as fat-associated lymphoid clusters [FALCs]) have been identified in the perivascular adipose tissue of mesentery (Supplemental Fig. 1). Regarding their functions, FALCs contain type 2 innate lymphoid cells (ILC2), capable of producing IL-5 and IL-13 in allergic reactions and during protection against helminth infections (9). In FALCs within the mediastinal adipose tissue, the activity of local ILC2 cells can be enhanced by resident stromal cells producing IL-33, leading to protective IgM production by B1 cells (10). Furthermore, omental MS may sustain B cell expansion, leading to germinal center formation coupled with affinity maturation and isotype switch, although without demonstrable follicular dendritic cell meshwork present in peripheral lymphoid organs (7). Similarly to peripheral lymph nodes (pLNs), CXCL13 is a critical chemokine required for B cell entry as well as FALC/MS formation and expansion; however, in contrast to other secondary lymphoid organs, FALC development occurs independently from lymphoid tissue inducer or subsets of ILC3 and LTβR engagement. In contrast, macrophage-derived TNF as well as nonlymphoid tissue inducer/ILC3–type innate lymphoid cells and commensal bacteria are necessary for the appearance of FALCs (9, 10).

These findings indicate that, although intra-abdominal fat-associated lymphoid tissues are anatomically separate and developmentally different from other peripheral lymphoid organs, they, nevertheless, contribute to the systemic immunity, including lymphocyte traffic and distribution. Importantly, their lymphocyte turnover mechanisms and kinetics have only been partially explored, including the access via blood vasculature or lymphatic circulation or migratory routes (11, 12).

In addition to leukocytes, the serous expansion of cancer cells, including colonic, gastric, and ovary tumors, also affects specific regions within the mesentery, in which local chemotactic as well as extracellular matrix (ECM) components may play important roles, including CXCR4/CXCL12 interactions and peritoneal collagen IV and fibronectin (Fn) (6, 13). Further propagation of metastasis may involve dissemination via lymphatic vessels situated within the mesenteric adipose streaks; however, the exact relationship between the cellular entry sites and lymphatic drainage remains to be determined. In addition to these cancers, expansion of high-grade extranodal B cell lymphomas may also lead to peritoneal lymphomatosis and serous effusions causing poor prognosis (14, 15).

In our previous work on a spontaneous mouse high-grade B cell lymphoma, we found a rapid dissemination toward the mesenteric lymph nodes (mLNs) coupled with the lymphomatous transformation of mesentery together with the accumulation of lymphoma cells within the mesenteric lymphatics (16). In the current study, we first investigated the i.p. lymphoma-binding sites of the serous membrane of abdominal cavity, allowing the identification and detailed characterization of a hitherto undescribed variant of serous lymphoid tissues that can efficiently bind i.p. lymphoma cells and normal lymphocytes. Owing to the leaf-like appearance of their main part, connected to omental and mesenteric adipose tissues or peritoneal membrane via a stalk, we chose to denote these structures as foliate lymphoid aggregates (FLAgs). These organoids possess partial lymphoid compartmentalization, which also influences the tissue distribution of high-grade B cell lymphoma subsets, corresponding to the confinement of local homeostatic chemokine dominance. Furthermore, we demonstrate that the removal of LYVE-1–positive macrophages reduces lymphoma cell binding. Finally, we provide evidence that FLAgs can also be destination sites for blood-borne leukocytes via pLN addressin (PNAd)–positive high endothelial venules (HEVs) with the partial use of L-selectin for homing.

Inbred BALB/c, C57BL/6J, and BALB/c^eGFP-transgenic mice (17) were maintained at the specific pathogen-free animal facility of the Department of Immunology and Biotechnology. Prox1-GFP BAC lymphatic reporter–transgenic mice (18) obtained from the Mutant Mouse Regional Resource Center were maintained in heterozygous form and were used for breeding (Prox1-GFP C57BL/6J crossed with BALB/c) F1 mice as recipients for short-term homing. KikGR mice (19) on C57BL/6J background were obtained from The Jackson Laboratory and were backcrossed through 10 generations onto BALB/c background. All procedures involving live animals were carried out in accordance with the guidelines set out by the Ethics Committee on Animal Experimentation (University of Pécs, Hungary) under license number BA02/2000-16/2015, with approval for the use of genetically modified organisms under license number SF/27-1/2014 issued by the Ministry of Rural Development, Hungary.

For flow cytometry rat mAbs against mouse, CD5 (YTS121.5), CD19 (1D3), MHC class II (IBL-5/22), CD21 (7G6), CD23 (B3B4), LFA-1/CD11a/CD18 (M17.7 and IBL-6/2), MAC-1/CD11b/CD18 (M1/70), CD45 (IBL-3/16), B220/CD45R (RA3-6B2), IgM (B7.6), Thy-1/CD90 (IBL-1), ICAM-1/CD54 (YN1/1), MAdCAM-1 (MECA-367), VCAM-1/CD106 (M/K-2.7), and L-selectin/CD62L (MEL-14) were used as hybridoma supernatants. Rat mAb against mouse CD138 (281-2) and PE-conjugated rat mAb against mouse CXCR5 (2G8) were obtained from BD Biosciences (Diagon, Budapest, Hungary) and biotinylated anti-CCR7 mAb (4B12) from BioLegend (Biomedica Hungaria, Budapest, Hungary), the latter detected with streptavidin–PE/Cy5 conjugate (from BD Biosciences). Rat mAb against LYVE-1 (223322) Ag and goat polyclonal Abs against mouse CCL21 and CXCL13 were purchased from Bio-Techne (R&D Systems, Diagon, Budapest, Hungary), FITC-conjugated donkey anti-goat Abs and HRP-conjugated sheep anti-FITC Abs were purchased from SouthernBiotech (Bio-Kasztel, Budapest, Hungary). For immunohistochemistry, rat mAbs were detected using ImmPRESS goat anti-rat IgG-HRP polymeric conjugate (Vector Laboratories, BioMarker, Gödöllő, Hungary). Rabbit anti-Fn and tetramethylrhodamine-labeled goat anti-rabbit polyclonal Abs were purchased from Abcam (Bio-Kasztel, Budapest, Hungary). Chemicals for buffers and histochemical substrates were purchased from Sigma-Aldrich.

For the labeling of CXCR5 and CCR7 A20 and Bc.DLFL1, lymphoma cells were incubated with biotinylated or PE-conjugated mAbs against chemokine receptors in mixture with anti-B220 mAb conjugated with Alexa Fluor 647 (prepared at the Department of Immunology and Biotechnology) at room temperature, followed by washing. After further incubation with streptavidin–PE/Cy5 conjugate, the samples were washed and fixed in 1% buffered paraformaldehyde. For other cell surface markers, the lymphoma cells were incubated on ice with rat mAbs listed above, followed by washing and incubation with goat anti-rat IgG FITC conjugate. After fixation, 10,000 events gated on forward light scatter (FSC)/side light scatter (SSC) parameters and B220 reactivity, or KikGR expression combined with CD45 expression, were collected and analyzed using BD FACSCalibur and CellQuest Pro software package.

After the removal of the entire gut complex from esophagus until the upper third of rectum, the gut was placed in a petri dish, with its folds arranged, and fixed in 4% buffered paraformaldehyde for 10 min. After rinsing in PBS, a short hematoxylin staining was performed, followed by rinsing in tap water. Either the whole omentum or the mesentery or some selected regions were isolated under stereomicroscopic dissection. For immunofluorescence, the samples were rinsed in PBS containing 0.1% saponin, followed by incubation in PBS containing 0.1% saponin and 5% BSA for 1 h. The samples were incubated with FITC anti-IgM, TAMRA anti–Thy-1 or Alexa Fluor 647 anti-CD45 mAbs overnight, followed by extensive washing in PBS containing 0.1% saponin and 0.1% BSA. For indirect immunofluorescence FITC or PE-conjugated anti-rat, anti-goat, or anti-rabbit secondary Abs (adsorbed for mouse IgG) were added and incubated overnight, followed by extensive washing for at least 6 h. Confocal fluorescence images were taken using an Olympus FluoView FV1000 laser scanning confocal imaging system (Olympus Europa SE & Co., Hamburg, Germany).

For immunohistochemistry, the tissue samples were incubated in 2 mg/ml phenylhydrazine hydrochloride in PBS containing 0.1% saponin to quench endogenous peroxidase activity, followed by extensive washing. The tissues were blocked with a 1:1 mixture of 20% normal goat serum and 5% BSA for 1 h, followed by the addition of rat mAbs in the presence of 5% DMSO. For detecting CFSE-labeled lymphoma cells, HRP-conjugated sheep anti-FITC Abs were used. The samples were incubated overnight with continuous agitation, followed by washing in PBS–0.1% saponin. The bound rat Abs were detected using goat anti-rat IgG-HRP conjugate following overnight incubation and were visualized using diaminobenzidine (DAB)–H₂O₂ (Dako, Kromat, Budapest, Hungary). After mounting, the sections were viewed under an Olympus BX61 microscope. The acquisition of digital pictures with a charge-coupled device camera was performed using the ZEN software; the pictures were processed using Adobe Photoshop 6.0 with adjustments for brightness contrast and color balance applied for the entire images.

Dissected mouse intestines and harvested FLAgs were fixed overnight at 4°C with 4% buffered paraformaldehyde, washed in PBS, and postfixed with 2% glutaraldehyde, followed by dehydration in graded ethanol. Tissue samples were treated with 1% osmium tetroxide (Polysciences, Warrington, PA) for 2 h and embedded in a Polybed/Araldite 6500 mixture (Polysciences). The 1-μm-thick semithin sections were stained with toluidine blue. The ultrathin sections were contrasted with uranyl acetate and lead citrate and studied with a Hitachi Electron microscope type H-7600.

A20 cells were cultured in RPMI 1640 (Sigma-Aldrich) supplemented with 10% heat-inactivated FBS (EuroClone, Biocenter, Szeged, Hungary) and transduced using replication-defective retroviral vectors containing the bacterial gene for β-galactosidase (LacZ) using the supernatant of PA317 packaging cells (kindly provided by Dr. C. L. Cepko) followed by G418 selection (0.5 mg/ml final concentration; Sigma-Aldrich). LacZ detection was performed using standard X-gal (Sigma-Aldrich) staining. Bc.DLFL1 cells (16) were maintained as serial i.p. passage, where the lymphoma cells were collected from the tumor-laden mLNs. For tracing, the injected cells were either labeled with 5 μM CFSE (Thermo Fisher Scientific, Life Technologies, Budapest, Hungary) as recommended, or with 10 μM CellTrace Far Red (CTFR) (20) kindly provided by Dr. K. R. Gee, Thermo Fisher Scientific, Eugene, OR). Normal peritoneal B cells were purified from BALB/c mice using FITC-labeled anti-B220 and anti-FITC beads, followed by separation on VarioMACS (Miltenyi Biotec). For near-infrared (NIR) fluorescence bioimaging of MACS-purified B cells or lymphoma cells, we used lipophilic XenoLight DiR dye (PerkinElmer) as recommended by the vendor. After labeling, the cells were washed in RPMI 1640 basal medium and were injected i.p. at 2 × 10⁶ cells per recipient dosage.

MACS-purified B cells or lymphoma cells (A20 or Bc.DLFL1) were labeled with XenoLight DiR dye, followed by i.p. injection. The fluorescence in the excised gut samples was measured using the IVIS Lumina III (PerkinElmer, Waltham MA) in vivo imaging system with the following parameters: autoacquisition time, F/stop = 1, Binning = 2, excitation: 740 nm, and emission filter: 790 nm. The tissue binding of Bc.DLFL1 lymphoma cells was quantified by NIR analysis of microdissected serous lymphoid organoids placed in Greiner Bio-One CELLSTAR plate after the hematoxylin staining of 4% paraformaldehyde–fixed omentum and mesentery of mice previously injected i.p. with XenoLight DiR–labeled lymphoma cells. The data were processed and analyzed using Living Image software (PerkinElmer) by displaying the fluorescence intensity using the total radiant efficiency ([photons per second per square centimeter per steradian]/[microwatt per square centimeter]) as a pseudocolor overlay image.

KikGR photoconversion of lymph nodes was performed by placing whole inguinal lymph nodes in a drop of sterile PBS, followed by exposure to a custom-made illumination device (Optics Engineering, Budapest, Hungary) through 6-mm-diameter fiber optics using indium gallium nitride–based 410-nm chip light-emitting diode (no. APGC1 410, 125 mW) for 2 × 5 min from two sides at room temperature. After the mechanical release of lymphocytes, KikR cells were placed on ice, whereas lymphocytes from unconverted (KikG) lymph nodes were incubated with 10 μg/ml anti–L-selectin/CD62L mAb MEL-14 (21). After incubation, the lymphocytes were washed and resuspended at 1:1 KikG/KikR ratio (with the final donor cell mixture verified by flow cytometry) and were injected in the tail vein in BALB/c recipients at 5 × 10⁷ total cells per recipient in 250 μl volume. The degree of inhibition was determined by the KikG/KiR ratio following anti-CD45 labeling, corrected with the KikG/KikR ratio of donor cell mixture.

Total RNA from inguinal lymph node, omentum, and mesentery homogenates was isolated using NucleoSpin RNA (Macherey-Nagel). Purity and concentration of RNA was analyzed by NanoDrop. cDNA was synthetized using High-Capacity cDNA Reverse Transcription Kit (Life Technologies). RT-PCR was run on an Applied Biosystems PRISM 7500 machine in duplicates using previously described SYBR Green primers for PNAd core proteins and glycosylation enzymes, creating the MECA-79 epitope (22). Results are shown as fold change of target gene relative to the β-actin housekeeping gene mRNA level, defining the relative value pLN as 1.

Data analysis was performed using SPSS 22.0 (IBM). Normality of data distribution was assessed by Shapiro–Wilks test. A t test or Mann–Whitney U test was employed to compare two groups with normally distributed and nonnormally distributed data, respectively. Data are represented as mean ± SEM. A p value < 0.05 was considered statistically significant.

Our previous study of the spontaneous extrafollicular high-grade B cell lymphoma Bc.DLFL1 in BALB/c mice revealed spreading via the lymphatic capillaries embedded in the perivascular fat pads of the mesentery and subsequent expansion restricted to mLNs and spleen (16). As this condition represented an advanced stage of disease, we aimed at studying the early steps of lymphoma expansion. Therefore, we first investigated the initial adhesion site of the serous surface, followed by the lymphatic propagation and eventual expansion along the perivascular fat. Four hours after i.p. injection of XenoLight DiR–labeled lymphoma cells followed by NIR fluorescence detection, transabdominal imaging revealed three main positive regions, seemingly corresponding to the mLNs and spleen as main target tissues for Bc.DLFL1 seeding in addition to the injection site. However, after opening up the abdomen, we observed that the omentum was the major dominant fluorescent signal-emitting site, together with a series of small foci arranged in a bead-like pattern along the edge of mesenteric fat streaks connected to the middle segment of small intestine, whereas the mLNs and the spleen were label-free (Fig. 1A–F).

Although the focal pattern of lymphoma accumulation within mesenteric fat and the omentum suggested specific tissue-binding pattern to exclude potential labeling artifacts because of the accumulation of lipophilic dye–tagged lymphoma cells in an adipose tissue environment, we performed a similar short-term transfer experiment using CFSE-labeled cells, followed by anti-FITC whole-mount hapten immunohistochemistry. This approach allows light-microscopical inspection over a wide range of magnification of the entire gastrointestinal tract while eliminating high-level autofluorescence of intestines related to alimentary compounds typically emitting at fluorescein spectrum (23). Using anti-FITC–PO immunohistochemical detection, we observed a distribution pattern of CFSE-marked lymphoma cells similar to NIR imaging (Fig. 1G, 1H). Quantitative comparison of omental and mesentery-associated XenoLight DiR fluorescence signal emission indicated that although the individual lymphoma-binding sites are relatively small, their total lymphoma-binding capacity equals that of the omentum (Fig. 1I).

To test whether normal B cells also display similar selective homing sites to the serous surface, MACS-purified peritoneal B cells were labeled with XenoLight DiR dye, and their location was investigated using NIR imaging. We found that, similarly to the distribution of Bc.DLFL1 cells, several B cell foci alongside the mesentery were present 4 h after the injection, in addition to the robust omental localization of peritoneal B cells (Fig. 2A). Interestingly, the quantitative analysis revealed that, compared with the similar diffuse large B cell lymphoma (DLBCL) retention by omentum and mesentery, the capacity of the entire mesentery for the uptake of normal peritoneal exudate cell (PEC) B cells was significantly higher than that of the omentum (Fig. 2B).

To further investigate the adherence kinetics of lymphocytes to the serous tissues in comparison with the entry into mLNs, we next injected normal BALB/c mice i.p. with lymphocytes isolated from enhanced GFP–reporter mice, followed by their detection in the omentum, mesentery, and mLNs and in the PEC pool at 2, 6, 24, and 48 h after injection. We found that the frequency of residual donor cells in PEC compartment continuously decreased after 6 h. In mLNs, the increase of GFP-positive cells’ frequency was first observed at 24 h, and it was maintained until 48 h. Interestingly, we observed a significant reduction after 24 h in the mesentery, whereas in the omentum, the donor cell frequency remained stable until 48 h (Fig. 2C).

These findings establish that, in addition to the omentum, the perivascular adipose streaks of mesentery contain specialized sites that can bind both normal B lymphocytes and high-grade B cell lymphoma cells. We found, in this study, that the transit time for the bulk of lymphocytes entering and departing the serous lymphoid tissues is between 6 and 48 h, with a delayed departure from the omentum compared with the mesentery within this period.

To characterize the mesenteric lymphocyte–binding regions, we next removed the whole gut–omentum–mesentery complex and stained it with hematoxylin, combined with lymphoma tracing. We found leaf-like formations, connected to the mesenteric adipose tissue either directly or via a slender stalk (Fig. 3A, 3B), both types showing robust uptake of Bc.DLFL1 cells (Fig. 3C). In addition, similar formations were also observed in the membrane lining of the omental bursa, appearing in paired or, less frequently, in single arrangement (Fig. 3D–F). Owing to their particular appearance, we denote these formations as FLAgs. Their dimensions were in the range of 250–450-μm length from base to tip, 150–250-μm width, and 30–80-μm thickness, respectively. The most typical location for FLAgs were the omental bursa, but in the omentum and in the mesenteric fat streaks, ∼5–15% of lymphoma-binding foci also comprised adipose-associated FLAgs in young adult mice (6–10 wk of age). Furthermore, several distinct leukocyte-rich FALCs and crescent-shaped protrusions (Fig. 3B), both capable of DLBCL binding (Fig. 3C), were also present in both the omentum and mesentery, in addition to the FLAgs.

To compare the lymphoma-binding efficiencies of the FLAgs with FALCs and protrusions, next, we performed a quantitative measurement using XenoLight DiR fluorescence. Hematoxylin staining revealed that FLAgs and protrusions represent more condensed leukocyte clustering compared with FALCs and omental MS located on the adipose surface with less-defined boundaries. Moreover, protrusions connect with broader attachment area to the underlying adipose base and lack identifiable stalks, thus allowing the microdissection of these different omental and mesenteric adipose lymphoid tissues after i.p. injection of XenoLight DiR–labeled Bc.DLFL1 cells. We observed the highest fluorescence signal in FLAgs (and with FALCs generating the lowest signal) at both the 4- and 24-h time points. In contrast, FALCs showed a marked increase of lymphoma binding by 24 h, whereas in FLAgs, the increase was less pronounced (Fig. 3G).

To determine whether these FLAgs occur only in BALB/c mice as a strain-specific trait, we also inspected C57BL/6J mice as a standard mouse strain used in immunological studies. We found in our C57BL/6J specific pathogen-free colony a similar occurrence of FLAgs in mice older than 8–10 wk. Their number is variable, usually in the range of 10–30 per mouse, with approximately twice as many sharply demarcated protrusions, and ∼50–80 FALCs/MS together in the omenta and mesentery, although latter structures are sometimes difficult to evaluate as one or more confluent clusters using hematoxylin staining (data not shown).

Peritoneal lavage is a regular procedure to obtain murine PEC for the analysis of the leukocyte composition and characteristics of this compartment. To investigate whether lavage affects the presence of FLAgs, we compared their appearance and frequency after hematoxylin staining but found no difference to untreated controls, indicating that during PEC collection the detachment of cells from FLAgs is minimal if any (data not shown).

Subsequent electron microscopic analysis using semithin sections revealed that the FLAg structures are enveloped in a mesothelial layer. Underneath this capsule, the cortical area is richly perfused with capillaries, whereas the medullary region is enriched for lymphocytes (Fig. 4A–C). The serous surface of mesothelial cells covering FLAgs display several microvilli, whereas in the deeper regions, CD138-positive plasma cells can be frequently observed intermingled with reticular cells, often in paired arrangements, comprising around 2.5–6.5% of leukocytes (35–55 cells stained positive for CD138 from a midlevel focus plane of FLAgs containing 850–1250 cells) (Fig. 4D–F).

According to these findings, the murine visceral serous surface contains preformed leukocyte-rich areas with different appearances that can bind DLBCL cells. Among these structures, the FLAgs represent a uniquely shaped variant present in both membrane-attached and adipose-associated locations, and these formations collect most efficiently the lymphoma cells injected i.p.

The striking form of FLAgs clearly distinguishable from the more diffuse forms of serous leukocyte clusters (such as MS or FALCs) prompted us to investigate their lymphoid subset composition using whole-mount immunohistochemistry. To verify their leukocyte content, we first used anti-CD45 labeling, resulting in a robust reactivity throughout the FLAgs, compared with the slightly less–dense labeling of the FALCs on the serous surface of adipose cuffs (Fig. 5A). Anti–Thy-1/CD90 labeling demonstrated a concentrated accumulation of T cells within the medullary region of FLAgs, whereas the anti-B220 labeling revealed a more diffuse B cell distribution (Fig. 5B, 5C). The marginal area of the FLAgs contained a dense network of LYVE-1–positive cells (Fig. 5D), previously described as FALC/MS-associated macrophages (10). Using dual immunofluorescence for Thy-1 and IgM, followed by confocal microscopy, we confirmed that Thy-1–positive T cells accumulated separately from the IgM-positive B cells, latter typically located at the peripheral regions toward the stalk connection and the tip regions as well as on the surface (Fig. 5E).

The partially segregated pattern of T and B cells suggested the presence of homeostatic chemokines recruiting specific lymphocyte subsets to separate domains within FLAgs. Labeling for Thy-1, B220, and either CCL21 or CXCL13 revealed discrete regions with different chemokine production. Thus, CCL21 production was confined to the central region of the FLAgs coupled with T cell clustering, whereas CXCL13 expression appeared more widespread (Fig. 5F, 5G). These findings suggest that within FLAgs a central region with T cell enrichment is partially separate from the peripheral B cell rich regions, parallel to corresponding homeostatic chemokine dominance.

To test whether the central CCL21-positive domain and the peripheral CXCL13-producing region can also affect the positioning of B cell lymphoma cells with different lymphoid compartment derivation, we compared the distribution of Bc.DLFL1 as extrafollicular DLBCL and A20 as centroblastic DLBCL (24), respectively. Cell surface phenotyping for B cell–associated markers revealed that whereas both Bc.DLFL1 and A20 cells demonstrated similar levels of surface CD19, in Bc.DLFL1 cells, the expression of CD21 and CD23 were reduced, whereas MAC-1/CD11b/CD18 expression was increased compared with A20 cells (Fig. 6A). Importantly, A20 cells display CXCR5 and lack CCR7, whereas Bc.DLFL1 cells have the opposite chemokine receptor pattern (Fig. 6B). Spectral karyotyping analysis (performed at the Molecular Cytogenetics Facility MD Anderson Center, Houston, TX) also revealed a different karyotype for Bc.DLFL1 cell, including 39–41 XX with t(6:10); t(10:6); t(10:11), in contrast to 39×/−, t (2, 15), del (6), del (9), dup (14) reported for A20 (24). Thus, these two DLBCL lines represent genetically and phenotypically two distinct lymphoma variants.

We found that 2 h after i.p. injection of mixed lymphoma cells, A20 cells accumulate at the cortical rim of the FLAgs, whereas Bc.DLFL1 cells readily enter the central region of the FLAg body where, under normal circumstances, T cells congregate. Interestingly, within the omental or mesentery-associated flat FALCs, such segregation did not occur, the A20 and Bc.DLFL1 cells remained mixed (Fig. 6C, 6D). To exclude that the differential migration within FLAgs was due to a competition between two different cell types being simultaneously present, we repeated this experiment using A20^Lacz cells alone. We found that, at the same time point, A20^Lacz cells remained excluded from the center of the FLAgs and accumulated in the cortical rim, whereas in FALCs, they were diffusely distributed (Fig. 6E, 6F), thus their exclusion from the FLAg center was not due to the presence of competing cells.

These data establish that, within FLAgs, different chemokine domains exist that segregate T and B cells and can also influence the specific accumulation of DLBCL subtypes as defined by their chemokine receptor display, whereas in FALCs, such lymphoma segregation is absent.

The presence of capillaries and reticular cells in FLAgs suggested the prolonged persistence of these lymphoid formations, raising the question of organization of their nonhematopoietic constituents. We first tested if FLAgs contain Fn as major ECM component. We found that FLAgs in both membrane- and adipose-associated locations contain an extensive Fn meshwork throughout the FLAg body, extending into the stalk regions (Fig. 7A). Using VCAM-1 as a broadly expressed stromal marker expressed in peripheral lymphoid organs, we found a pronounced capillary labeling along the FLAgs stalk, as well as cells with reticular fibroblastic appearance (Fig. 7B).

Using anti-CD31 immunohistochemistry, we found that both the stalk and body parts of FLAgs contain an elaborate meshwork of CD31-positive blood capillaries. Furthermore, within the stalk-proximal parts of both omental and mesentery-associated FLAgs, short PNAd-positive vascular segments could be detected (Fig. 7C). Using anti–MAdCAM-1, CR1/2, and anti-ICAM immunohistochemistry, the FLAgs did not display any discernible follicular stromal reactivity suggesting follicular dendritic cells or mucosal-type HEVs (data not shown).

We also sought to identify lymphatic capillaries staining for hyaluronan receptor LYVE-1 (25) or the expression of lymphatic endothelial cell (LEC)–specific transcription factor Prox-1 (26). We found that FLAgs contained only a few scattered cells expressing GFP in Prox-1 reporter mice (Fig. 7D). Furthermore, whereas macrophages with intense LYVE-1 labeling were present in a discontinuous arrangement at the FLAg edges, as described for omental FALCs and MS (10), they showed no vascular organization, supporting the lack of lymphatics (Fig. 7E, 7F).

These findings reveal that the nonhematopoietic structure of FLAgs includes VCAM-1–positive stromal cells embedded in a Fn-rich ECM, and FLAgs are supplied by blood capillaries containing short PNAd-positive segments, but FLAgs lack a defined LEC vasculature.

To chart the lymphoma-binding and dissemination routes following i.p. injection involving FLAgs, next, we used various labeling strategies. First, we investigated whether Bc.DLFL1 cells bind to the LYVE-1–positive macrophages. Using CFSE-labeled lymphoma cells and combined immunofluorescence for LYVE-1 and Fn as FLAg ECM compound, we found that 1 h after injection, CFSE-labeled lymphocyte accumulate at the tip part of the FLAgs enriched for LYVE-1–positive macrophages, whereas a small fraction had already entered the FLAg body. Interestingly, both migrating lymphoma cells and LYVE-1–positive macrophages could be observed in a close association with Fn bundles (Fig. 8A).

To confirm that it is indeed macrophages that mediate the early binding of lymphoma cells, next, we studied how their depletion by the i.p. administration of clodronate liposomes (27) affects lymphoma homing. We found that a 24-h-long pretreatment with clodronate liposome led to an ∼80% reduction in the subsequent mesenteric and omental binding of XenoLight DiR–labeled Bc.DLFL1 cells after a 4 h period (Fig. 8B, 8C) together with the substantial loss of LYVE-1–positive macrophages from the peritoneal membrane and FLAgs (Fig. 8D–F) compared with PBS-liposome control treatment, indicating effective removal of macrophages that, in turn, led to the significant reduction of serous lymphoma adhesion.

As the progress of Bc.DLFL1 lymphoma involves spreading into the mLNs, next, we studied whether injected lymphoma cells can be located within the mesenteric lymphatic capillaries. Using CFSE-labeled lymphocytes and anti-FITC whole-mount immunohistochemistry, first, we found thin-walled capillaries within the adipose streaks of mesentery in a close vicinity of original lymphoma attachment at 24 h after the injection, suggestive of lymphatic vessels (Fig. 8G). To verify the entry of lymphoma cells into lymphatic capillaries, we used (Prox1-GFP crossed with BALB/c) F1 backcross mice as lymphoma recipients expressing GFP in LECs (18); thus, the positioning of CTFR (red) fluorescence labeled lymphoma cells within capillaries formed by GFP-marked LECs could be analyzed by confocal microscopy without immunological rejection. We found that, as early as 4 h postinjection, CTFR lymphoma cells enter the GFP-positive lymphatic capillaries (Fig. 8H, 8I) in the mesentery.

According to our findings, the initial event for the serous propagation of DLBC lymphoma cells is their attachment to LYVE-1–positive macrophages in FLAgs and less efficiently in FALCs. Next, the lymphoma cells migrate toward the lymphatic capillaries within the mesentery, where the lymphoma cells can enter the draining capillaries, thus gaining access to the mLNs for subsequent expansion. Moreover, as the FLAgs (although they efficiently bind B cells and DLBCL cells) are devoid of demonstrable lymphatic capillaries, the entry sites of i.p. injected lymphoma cells into the mesenteric lymphatic drainage are located outside the FLAgs.

The presence of PNAd-positive segments in both mesenteric and omental FLAgs and FALCs raised the possibility of their involvement in the local leukocyte extravasation mediated by PNAd ligand L-selectin, similar to pLNs. To determine whether the blockade of L-selectin abrogates the serous extravasation of lymphocytes from blood vessels, we performed a competitive adoptive cell transfer experiment using KikGR detection (19). In this study, KikG cells pretreated with saturating amount of anti–L-selectin mAb MEL-14 were coinjected i.v. with untreated KikR cells (with the Kik fluoroprotein switched to red variant upon photoconversion of pLNs from KikG mice), and their relative appearance in the omentum and mesentery was determined by flow cytometer following CD45 labeling (Fig. 9A). We found that, 1 h after the injection, a substantial blockade occurred in the pLNs, whereas no measurable alteration of the original KikG/KikR ratio was observed in the spleen with L-selectin–independent homing. In the mesentery, a partial blockade could be observed, at around 50% less inhibition of homing compared with pLN (Fig. 9B). In the omentum, we detected only very sparse Kik donor cell appearance in this period, hampering the precise determination of the degree of inhibition (data not shown).

Next, we compared the expression of mRNA for the core proteins and glycosylation enzymes necessary for the production of MECA-79 epitope of PNAd. Although the FLAgs are more abundant in the mesentery in variable distribution, we chose omenta as sample, owing to their more uniform tissue size and composition, including the presence of FLAgs. With quantitative PCR of omental samples, we found that Cd34, endomucin (Emcn) and podocalyxin (Podxl) core proteins mRNA and β1-6GlcNAc transferase (Cgnt2) enzyme mRNA levels variably increased, whereas GLYCAM (Glycam1) core protein, intercellular cell adhesion molecule/ICAM (Icam) as well as Cgnt1, β-1,3-N-acetylglucosaminyltransferase (B3gnt3) and fucosyltransferase (Fut7) enzyme mRNA increased to different degrees in comparison with pLN reference of target gene/β-actin mRNA ratios (Fig. 9C, 9D).

These data indicate that the extravasation of blood-borne lymphocytes to serous lymphoid tissues partially depends on the interaction between L-selectin and MECA-79–positive PNAd endothelial ligands, generated by the concerted action of several core proteins and modifying enzymes with detectable level of mRNA production, probably via the simultaneous involvement of different serous lymphoid organoids.

In our present work, we establish that, following i.p. injection, normal B cells and DLBCL cells readily colonize various serous lymphoid organoids in the abdominal cavity whence, despite the lack of lymphatic vessels, they gain access to the mesenteric lymphatic capillaries. These capillaries, in turn, thus represent a dual afferentation route for the mLNs, draining both the gut and the peritoneal cavity. As early serous docking sites, morphologically different lymphoid formations may be seeded by lymphoma cells (6) and normal B cells, including a hitherto undescribed structure denoted as FLAg in a process involving LYVE-1–positive peritoneal macrophages. FLAgs show a partial T/B compartmentalization and the presence of domains producing CXCL13 and CCL21 homeostatic chemokines. Furthermore, in contrast with the previously described diffuse MS and FALC structures (1, 2), FLAgs are completely enveloped into mesothelial cells, suggesting a hitherto unknown type or stage of adipose-associated peripheral lymphoid tissues.

The degree of compartmentalization of the bulk of omental lymphoid tissues into T and B cell zones is substantially less than that of the mucosal draining sites, such as mLNs or Peyer’s patches (1, 2, 10). However, in FLAgs, a concentrated T-zone can typically be identified, corresponding to the region producing CCL21, whereas the peripheral region of FLAgs is enriched for CXCL13 production, which is necessary for omental B cell entry (7, 10). In the omentum, CXCL13 production has been attributed to both hematopoietic cells and, to a lesser extent, nonhematopoietic stromal cells (8, 10). In FLAgs, the relative enrichment of VCAM-1–positive stromal cell in the center suggests possible involvement in CCL21 production, whereas the more prominent production of CXCL13 in the periphery can be linked to LYVE-1–positive macrophages (10), in addition to VCAM-1 being a possible key vascular adhesion molecule for omental homing of recirculating B cells (11). Whether follicular dendritic cells as cardinal CXCL13-producing stromal cells in secondary lymphoid tissues are also present in FLAgs requires further investigations, including phenotypic analysis and functional assays, particularly germinal center formation and long-term immune complex retention. Although the treatment of EL4 T cell lymphoma cells with pertussis toxin did not affect their accumulation in omental foci (7), T and B cells and various DLBCL variants within FLAgs show evidences for partial segregation, indicating that the various forms (MS and FALCs) of serous lymphoid aggregates have different levels of chemokine-mediated organization affecting T/B positioning. It is not yet known whether B-1 B cells within the B cell–rich areas accumulate to separate regions relative to the B-2 subset. Thus, for determining the significance of pertussis toxin–sensitive chemokine signaling in the subsequent omental or mesenteric migration of B cells and DLBCL cells within different types of serous lymphoid tissues, further investigations are needed.

Importantly, previous studies demonstrated that the surgical removal of omentum did not prevent the departure of peritoneal B cells, hinting at other potential exit sites (11), with mesenteric FALCs and FLAgs offering possible candidates as demonstrated in our present work. Our findings indicate mesenteric lymphoid organoids (including FALCs and FLAgs) as alternative lymphocyte-binding sites for subsequent exit routes, at an efficiency of lymphoma binding nearly equal to omental MS lymphoma adhesion. Among various formations, we observed different lymphoma-binding kinetics, indicating faster homing to FLAgs compared with FALCs, which may mediate both more efficient adherence and faster exchange kinetics, as the FLAg-bound lymphoma cell load did not change significantly over a 24-h period.

In contrast to the abundance of blood vasculature, in the omentum, we found only occasional LEC-marker expression, unlike in the mesentery, where typically several confluent lymphatic vessels displaying Prox-1 and LYVE-1 are present. Moreover, relying on these markers, we failed to detect lymphatic connection between the FALCs or FLAgs as mesenteric lymphoma-docking sites and the deep-running lymphatic vessels. Therefore, we can rule out direct lymphatic communication connecting either FLAgs or FALCs and mesenteric lymphatics, clearly necessitating further investigations to clarify the movement of FLAg/FALC-attached lymphocytes and lymphoma cells toward the mesenteric lymphatic vessels. Nevertheless, our kinetic analyses suggest that relatively early (4 h) after the i.p. injection, DLBCL cells are already detectable within the mesenteric lymphatic capillaries. Furthermore, the kinetics of normal B cell distribution following i.p. injection also suggests a faster departure from the mesentery than from the omentum after 24 h, possibly because of the drainage via the mesenteric lymphatic vessels, which appear absent in the omentum. Moreover, similar early accumulation of both normal B cells and DLBCL cells in human tumors with disrupted Gα13 (thus compromising S1P-mediated inhibition of migration induced by CXCL12) was also noted in the parathymic lymph nodes, where lymphocytes could reach this site either across the lymphatic stomata of the diaphragma or via lymphatic vessels (11, 28). In addition, radiodiagnostic imaging of human patients has also revealed the propagation of Burkitt lymphoma and DLBCL via the gastrocolic, gastrosplenic, and other peritoneal ligaments toward the mesocolon (29). Subsequent inflammatory reactions may expand the pre-existing lymphatic vasculature by LEC proliferation upon the effect of VEGF-C produced by perivascular smooth muscle cells and macrophages or by the direct integration of LEC-switched macrophages (30).

The capacity of lymphoma cells to adhere to such serous lymphoid formations along the mesentery represents a significantly expanded spectrum of potential propagation sites for this type of malignancies in addition to the omental MS also targeted by gastric, colonic, and ovary cancer cells (6, 31, 32). As both mesenteric FALCs and omental MS are able to host local immune responses, the entry of tumor cells into these lymphoid territories may elicit an immune reaction against such cells (1, 2, 6). However, the adipose microenvironment also offers a supportive milieu for cancers, thus antagonizing the efficiency of antitumor immune responses (31, 33, 34), in which cytotoxic reactions can also be modulated by ILC2 and myeloid-derived suppressor cells (8, 35). This tumor growth–promoting function may manifest in several ways, including the niche function of local microenvironment to support cancer stem cell growth. Alternatively, these sites may attract putative tumor–associated stroma precursors, including myofibroblast precursors or mesenchymal stem cells that exert immunosuppressive activity (36, 37) as well as chemotactic signals, including CXCL12–CXCR4 and CCL22–CCR4 interactions (38, 39). In omental tissue, the local microvascular network of actively sprouting CD105-positive capillaries influenced by VEGF-A binding several tumor types (32) may facilitate both the entry for recirculating stroma precursors and exit for cancer cell dissemination. In addition, local mesothelium–mesenchymal transition may also promote the appearance of cancer-associated fibroblasts, whereas adipocytes may also contribute to vessel expansion following their dedifferentiation and endothelial redifferentiation (40, 41).

The i.v. entry of lymphocytes to the MS and their retention, and their peritoneal entry are likely to be mediated by different adhesion events, as the egress from the peritoneal cavity was independent from β7 integrin ligand binding (11). The detectable level of mRNA for various PNAd core proteins and modifying enzymes involved in the generation of MECA-79 epitope in selective segments of vasculature within the omenta and mesentery supports the view that the local formation of available endothelial ligand permits L-selectin–mediated homing of blood-borne leukocytes, similar to pLN HEVs. However, the precise comparison with pLN HEVs necessitates the isolation of FLAgs for the enrichment of MECA-79–positive endothelial cells. Furthermore, anti–L-selectin treatment could only partly inhibit the entry of blood-borne lymphocytes to serous lymphoid tissues, indicating other potential endothelial recognition mechanisms, also likely to be independent from MAdCAM-1 binding, as the FLAg-associated vascular segments lacked this addressin. Because of the small size and transparent appearance of FLAgs, our attempts to isolate them in unfixed and unstained conditions have remained unsuccessful. Nevertheless, the presence of PNAd-positive vessels as well as the efficient binding through LYVE-1–positive macrophages together (thus offering access for both blood-borne and serous leukocytes) may explain the increased lymphocyte density within FLAgs compared with omental or mesenteric FALCs. Further analyses are required to identify other endothelial ligand–receptor pairs that facilitate serous homing of recirculating leukocytes.

In summary, we propose that the serous lymphoid aggregates function as peritoneal collection sites for both normal B cells and DLBCL cells. These lymphoid clusters have diverse structural variants displaying different degrees of lymphocyte compartmentalization, with FLAgs demonstrating a more organized variant. From these sites, the peritoneal lymphocytes may subsequently reach the mesenteric lymphatic capillaries, which, in turn, thus represent shared transit routes for both leukocytes, leaving the mucosal and the serous compartments, whereas the dissemination of peritoneal leukocytes and lymphoma cells may involve also other routes, particularly toward the mediastinal cavity and pleural surface. As initial contact partner cells binding B cell and DLBCL, LYVE-1–positive macrophages may play crucial roles; however, the identity of chemotactic and other factors guiding the lymphocytes toward the mesenteric lymphatic capillaries remains to be determined.

We gratefully acknowledge Dr. Constance L. Cepko (Departments of Genetics and Ophthalmology, Howard Hughes Medical Institute, Harvard Medical School, Boston, MA) for the PA317 packaging cells producing replication-defective retroviral vectors containing the bacterial gene for β-galactosidase and Dr. Kyle R. Gee (Thermo Fisher Scientific, Eugene, OR) for providing CTFR dye. Spectral karyotyping was performed at the Department of Genetics, Division of Basic Science Research, The University of Texas MD Anderson Cancer Center (Houston, TX), headed by Dr. Asha S. Multani.

The authors have no financial conflicts of interest.