mTORC1 activation in B cells confers impairment of marginal zone microarchitecture by exaggerating cathepsin activity

Summary

Mammalian target of rapamycin complex 1 (mTORC1) is a key regulator of cell metabolism and lymphocyte proliferation. It is inhibited by the tuberous sclerosis complex (TSC), a heterodimer of TSC1 and TSC2. Deletion of either gene results in robust activation of mTORC1. Mature B cells reside in the spleen at two major anatomical locations, the marginal zone (MZ) and follicles. The MZ constitutes the first line of humoral response against blood‐borne pathogens and undergoes atrophy in chronic inflammation. In previous work, we showed that mice deleted for TSC1 in their B cells (TSC1^BKO) have almost no MZ B cells, whereas follicular B cells are minimally affected. To explore potential underlying mechanisms for MZ B‐cell loss, we have analysed the spleen MZ architecture of TSC1^BKO mice and found it to be severely impaired. Examination of lymphotoxins (LTα and LTβ) and lymphotoxin receptor (LTβR) expression indicated that LTβR levels in spleen stroma were reduced by TSC1 deletion in the B cells. Furthermore, LTα transcripts in B cells were reduced. Because LTβR is sensitive to proteolysis, we analysed cathepsin activity in TSC1^BKO. A higher cathepsin activity, particularly of cathepsin B, was observed, which was reduced by mTORC1 inhibition with rapamycin in vivo. Remarkably, in vivo administration of a pan‐cathepsin inhibitor restored LTβR expression, LTα mRNA levels and the MZ architecture. Our data identify a novel connection, although not elucidated at the molecular level, between mTORC1 and cathepsin activity in a manner relevant to MZ dynamics.

Abbreviations

KO

knockout

LT

lymphotoxin

LTβR

lymphotoxin receptor

MARCO

macrophage scavenger receptor with a collagenous structure

mTORC1

mammalian target of rapamycin complex 1

mTOR

mammalian target of rapamycin

MZBs

marginal zone B cells

MZ

marginal zone

PTEN

phosphatase and tensin homologue

S1P

sphingosine‐1‐phosphate

SDS–PAGE

sodium dodecyl sulphate–polyacrylamide gel electrophoresis

TSC

tuberous sclerosis complex

WT

wild‐type

Introduction

Marginal zone B cells (MZBs) bridge the innate and adaptive immune response. This innate‐like activity is achieved by their ability to quickly respond to B‐cell receptor and toll‐like receptor stimulations by rapid proliferation and secretion of antibodies, primarily of the IgM isotype. This humoral response is critical to protect against a large variety of pathogens until the germinal centre reaction develops. Furthermore, MZB‐derived antibodies contribute to the natural antibody pool. The marginal zone (MZ) itself is strategically located at the interphase with the bloodstream, allowing their proximity to blood‐borne antigens. For efficient activation of MZBs, antigens need to be captured and handed over for stimulation. Hence, many cell types and different capturing mechanisms participate in organizing the MZ architecture in a manner compatible for the execution of MZB activation.1

Human MZBs are located in the spleen and in lymph nodes, but MZBs in the mouse are located only in the spleen. Because the MZ in the mouse harbours specific types of cells, it can be discerned by a typical immunohistological analysis. The MZ is delineated by a ring of marginal metallophilic macrophages that express Siglec‐1 (identified by the MOMA‐1 monoclonal antibody, often called the MOMA‐1 protein). This layer of cells is physically touching the marginal sinuses, which express the mucosal vascular addressin cell adhesion molecule 1 (MAdCAM‐1) adhesion molecule.2 At the outskirts of the MZ sinuses are the MZ macrophages, which express the macrophage scavenger receptor with a collagenous structure, termed MARCO.3 All these markers generate ring‐like appearances in an intact MZ.

Marginal zone B cells are motile cells, which patrol the follicle area of the spleen and return to the MZ. Retrieval signals such as sphingosine‐1‐phosphate (S1P), released from the MZ sinusoids and captured by its cognate receptors of MZBs (S1P1 and S1P3), together with integrin/adhesion molecule interactions with the MZ stroma cells ensure the return of MZBs to their original location. Hence, perturbations in homing signals of MZBs result in their specific loss, whereas the follicular B cells that constitute the vast majority of splenic B cells remain intact.4 Importantly, in the absence of B cells, the MZ does not exist, suggesting that cross‐communication with MZ stromal cells is critical to establish this compartment.5 This is especially intriguing in light of the fact that normal MZ architecture can be maintained in the complete absence of MZBs, as evidenced by various animal models.6 Several studies addressed the B‐cell‐specific mechanisms required for MZ formation. For instance, B‐cell‐specific deletion of interleukin‐7 or deletion of components of lymphotoxin signalling lead to the loss of MZBs and the disappearance of the MZ.7, 8 Whether this is a result of lack of MZ formation or impaired maintenance is not known. Moreover, the factors downstream of the signalling cascades that control MZ have not been documented.

The mammalian target of rapamycin (mTOR) is an evolutionarily conserved serine/threonine kinase, present in two cellular complexes that determine its substrate specificity, mTOR‐complex 1 (mTORC1) and mTORC2.9 mTORC1 plays a major role in cellular metabolism by optimizing anabolic activities to cellular energy and oxygen levels, nutrient availability and exposure to various extracellular stimuli. mTORC1 is regulated by various mechanisms, a potent one is its inhibition by the tuberous sclerosis complex (TSC). Deletion of either TSC component, TSC1 or TSC2, renders mTORC1 hyperactive and less sensitive to stressful signals. Most of the studies that investigated the role of mTORC1 in B cells used loss of function strategies, such as rapamycin,10 expression of hypomorphic mTOR11 or conditional deletion of raptor.12 However, a full understanding of the role of mTORC1 in B‐cell development and function requires also a gain‐of‐function approach. We have generated mice in which TSC1 was conditionally deleted in B cells (referred to as TSC1^BKO). These mice exhibited a severe loss in MZBs and accordingly responded poorly to challenges with a T‐independent antigen.13 An independent parallel study that analysed the same mouse model confirmed these observations.14

We report that the lack of MZBs in TSC1^BKO mice is accompanied by an abnormal MZ architecture and impaired MZ function. Expression of lymphotoxin α (LTα), but not LTβ, was reduced in TSC1 knockout (KO) B cells and was elevated by rapamycin treatment. The surface expression of the lymphotoxin receptor, LTβR, was also reduced in the splenic stroma, implicating shedding as a potential mechanism. In support, we found an elevated cathepsin activity in the TSC1 KO B cells that was reduced by rapamycin treatment. Remarkably, in vivo inhibition of cathepsins restored MZ architecture, elevated LTβR expression in the spleen and corrected the reduced LTα expression in B cells. This study underscores for the first time the cathepsin proteases as key players in the maintenance of MZ architecture and their regulation by mTOR.

Materials and methods

Mice

CD19‐Cre/TSC1^f/f and the CD19‐Cre mice were previously described.13 To generate mice in which B cells are fluorescently labelled, we crossed the mice with the ROSA26^LSL‐YFP, as described in ref. 15. The resultant strains CD19‐Cre/TSC1^f/f/YFP and CD19‐Cre/YFP were used for the transfer experiments of B cells into JHT mice, lacking B cells. The RERT strain was provided by Dr Mariano Barbacid (CNIO, Madrid, Spain).

Immunohistochemistry

Spleens were immersed in 4% paraformaldehyde for 2 hr at 4°; then, transferred to a 30% sucrose solution for overnight incubation at room temperature. Following freezing in OCT (Scigen, Gardena, CA, USA, Cat#4583) at −80°, 5‐ to 7‐μm‐thick sections were prepared on a Leica CM1950 cryostat. Tissue sections were washed twice with PBS‐T (phosphate‐buffered saline with 0·1% Tween‐20) or super‐sensitive buffer (Biogenex, Fremont, CA, USA, Cat# HK583‐5KE) then fixed in 4% paraformaldehyde. Following washing and blocking in Cas block (Life Technologies, Carlsbad, CA; Cat#008120), sections were incubated with primary antibodies (AbCam, Cambridge, UK; rat anti MOMA‐1, Cat# 51814, 1 : 200; BioLegend, San Diego, CA; anti‐mouse MAdCAM‐1 antibody, clone MECA‐367, 1 : 200; Serotec, Hercules, CA; rat anti‐MARCO, Cat# MCA1849, 1 : 200; AbCam, rabbit anti‐LTβR, Cat# ab186847, 1 : 200) for 1 hr at room temperature, washed twice and incubated with the secondary antibody (Jackson ImmunoResearch, West Grove, PA; Alexa647‐donkey anti‐rat, Cat# 712605153, 1 : 500) for 30 min at 4°. Tissue was washed and counterstained with DAPI; after mounting, slides were kept at 4°. Microscopy was performed on an Olympus IX73 inverted microscope at 4× magnification, equipped with DAPI and Cy5 filters.

Flow cytometry of MZB cells

Splenocytes (1 × 10⁶) were stained with CD23‐Alexa Fluor 647 (BioLegend, Cat# 101611, 1 : 100), CD21/35‐PE (BioLegend, Cat# 123409, 1 : 50 dilution) and CD19‐FITC (Miltenyi Biotec, Bergisch Gladbach, Germany, Cat# 130102822, 1 : 50 dilution). When YFP‐positive mice were used, B cells were gated on the YFP‐positive population in lieu of CD19 staining. LTβR was detected by mouse anti‐LTβR antibody (5G11) (AbCam Cat#ab65089, 1 : 500). Cells were analysed by CytoFLEX (Beckman Coulter, Brea, CA). Data were analysed by using cytexpert software (Beckman Coulter, Indianapolis, IN, USA).

Uptake of bioparticles by MZ macrophages

Two hundred micrograms of heat killed Alexa Flour 488‐conjugated Staphylococcus aureus bioparticle (Invitrogen, Carlsbad, CA; Cat# S23371) were injected into the tail vein of mice. Mice were killed 30 min later and splenocytes were isolated. Splenocytes were stained for MARCO using Rat PE‐MARCO antibody (BioRad, Hercules, CA; MCA 1849PE) and analysed by flow cytometry.

MZ formation assay

B cells were isolated from the spleen using the EasySep Mouse B‐cell enrichment kit (STEMCELL, Vancouver, Canada); 20 × 10⁶ splenic B cells were injected into the tail vain of JHT mice at day 1 and day 3. At day 14, mice were killed and spleens were analysed by immunohistochemistry.

Induction of TSC1 deletion in RERT/TSC1^f/f mice

Three milligrams of tamoxifen in corn oil (20 mg/ml) were injected twice subcutaneously with a 1‐day interval into the remaining animals. Mice were analysed for MZ structure at the indicated time after tamoxifen treatment.

Cathepsin activity labelling

B cells were lysed in RIPA buffer [1% Tergitol‐type nonidet P‐40, 0·1% sodium dodecyl sulphate (SDS), 0·5% sodium deoxycholate] on ice for 10 min, and proteins were quantified by Bradford assay. Equal amounts of total protein extracts (50 μg) in acetate buffer (50 mm acetate, 4 mm dithiothreitol, 5 mm MgCl₂, pH 5·5) were labelled with GB123, a fluorescent cathepsin‐activity‐based probe16 (1 μm), for 1 hr at 37°. Reaction was stopped by addition of Laemmli reducing sample buffer and boiling of the samples for 5 min. Fluorescently labelled proteins were then separated by 12·5% SDS–polyacrylamide gel electrophoresis (PAGE) and visualized by Typhoon FLA 9500 scanner at 635/670 nm excitation/emission. The signal intensities of cathepsin B activity band from fluorescent gel scans were quantified using fiji/imagej software. The values obtained for cathepsin activity were normalized to those of total protein quantification from Coomassie stained for each sample, which were similarly quantified.

Cathepsin activity measurement in intact cells

One million B cells or two million splenocytes isolated from the spleens of TSC1 KO or wild‐type (WT) mice were incubated with 2 μm GB123 in medium for 4 hr at 37°. Excess of unbound probe was removed by centrifugation (100g, 5 min, 4°) and cells were washed for another 1 hr by gentle shaking and replacing the medium every 10 min. Cells were analysed by flow cytometry or used for cytospin preparations. For imaging analysis, cells were resuspended in cold 1% bovine serum/PBS, loaded into cytospin cuvette and centrifuged by a Cytospin 2 centrifuge (ThermoScientific, Shandon, UK) at 100 g for 5 min, followed by cold methanol fixation. Cover slips were mounted on the samples with DAPI Fluoromount‐G (Southern Biotech, Birmingham, AL). Fluorescent images were taken with a Nikon LT SMZ25 inverted fluorescence microscope equipped with 60 × oil objective and Cy5 and DAPI filters.

Rapamycin and GB111‐NH₂ in vivo treatment

TSC1^BKO mice were treated with GB111‐NH₂ cathepsin inhibitor by intraperitoneal injections (1 mg/mouse in 10% dimethyl sulphoxide, 15% Tween‐80, 75% PBS). GB111‐NH₂ was administered to mice in two rounds of 3 days, followed by 3 days of saline treatment. Following killing, liver, kidney and spleen were harvested. A third of the spleen was analysed by immunohistochemistry. The remains of the spleen and other organs were frozen in liquid nitrogen. Tissues were lysed using an electric TissueRuptor homogenizer (Qiagen, Hilden, Germany) in RIPA buffer on ice (as described above), and proteins were quantified by Bradford assay. Equal amount of total protein extracts (50 μg) in acetate buffer (50 mm acetate, 4 mm dithiothreitol, 5 mm MgCl₂, pH 5·5) were labelled with 1 μm GB123 for 1 hr at 37°, labelling reaction was stopped by adding sample buffer and boiling. Fluorescently labelled proteins were then separated by 12·5% SDS–PAGE and visualized as described above. Rapamycin was administered in vivo as described in ref. 13.

Cleavage of LTβR and co‐culture experiments

CHO‐K1 cells (4·5 × 10⁵) transfected with an LTβR‐GFP fusion vector (TransIT^®‐2020 Transfection Reagent, Mirus, Mirus Bio, Madison, WI) were incubated in a 96‐well plate in acetate buffer (pH 5·5) for 3 hr with recombinant cathepsin B or L (1·67 or 1·65 μm, respectively) or vehicle control at 37° under gentle shaking. Cells were then lysed directly in sample buffer, boiled for 3 min and equal protein amounts were analysed by Western blotting with anti‐GFP antibody. For co‐culture experiments, we used the LTβR‐GFP transfected CHO cells. One day after transfection, 1 × 10⁵ cells were seeded in a six‐well plate. The next day, 10 × 10⁶ B cells were added on top and cultured for additional 24 hr. Lymphocytes were washed out extensively, and the adherent CHO cells were harvested and lysed in cold RIPA buffer. Lysates were analysed by Western blotting on a 12% SDS–PAGE with anti‐GFP.

Quantitative polymerase chain reaction analysis of LTα and LTβ

Total RNA was isolated from splenic B cells by Trizol and cDNA was generated using the iScript kit (BioRad) and subjected to quantitative polymerase chain reaction (PCR) analysis using the following primers: LTα‐F: CCA CCT CTT GAG GGT GCT TG; LTα‐R: CAT GTC GGA GAA AGG CAC GAT; LTβ‐F: TGG CAG GAG CTA CTT CCC T; LTβ‐R: TCC AGT CTT TTC TGA GCC TGT; UBC‐F: CAG CCG TAT ATC TTC CCA GAC T; UBC‐R: CTC AGA GGG ATG CCA GTA ATC TA.

Results

Loss of MZ architecture in TSC1^BKO mice

To examine whether the reduction in MZBs in TSC1^BKO mice is accompanied by histological changes to the MZ architecture, spleens of TSC1^BKO mice were examined for the typical ring structure when stained for the various MZ markers. We observed the dissipation of the typical MZ structures in TSC1^BKO mice. Instead of ring‐shaped structures of MOMA‐1, MAdCAM‐1 and MARCO expressing cells seen in spleens of WT mice, a scattered and punctuated pattern was observed in the TSC1^BKO spleens (Fig. 1 and see Supplementary material, Fig. S1a). In vivo administration of rapamycin for 2 weeks caused the re‐establishment of MZ architecture (Fig. 1, bottom panel). This indicates that exaggerated mTORC1 activity in B cells damages the MZ architecture in a reversible manner.

Figure 1.

Loss of marginal zone (MZ) architecture in TSC1^BKO mice is not due to attenuated AKT activity. A representative immunohistochemical analysis of B cells (green) and Siglec1 (MOMA‐1, red) in wild‐type (WT), TSC1^BKO, PTEN^BKO and PTEN/TSC1^BKO spleens. Treatment of TSC1^BKO mice for 10 days with rapamycin restored MZ structures. Analyses were performed on six individual mice per each group.

The inhibition of mTORC1 with rapamycin causes an induction of AKT activity, known as the mTOR feedback.17 Conversely, induction of mTOR activity through the genetic loss of TSC inhibition, typically causes the repression of AKT in a manner that can be circumvented by the co‐deletion of phosphatase and tensin homologue (PTEN). Because AKT activity positively regulates the development of MZBs,18 it raises the possibility that the phenotype seen in the TSC1^BKO mice is the mere result of an attenuated AKT activity. To address this question, we crossed the TSC1^BKO to the PTEN^f/f mice to obtain the double KO mice (referred to as TSC1/PTEN^BKO). Assessment of AKT and mTOR activities by looking at the phosphorylation status of AKT and ribosomal protein S6 (S6), indicated the expected biochemical phenotype, a concomitant increase in P‐S6 and P‐AKT (see Supplementary material, Fig. S1b). Analysis of MZBs showed that while MZBs were slightly enriched in PTEN^BKO mice,18 the co‐deletion of TSC1 conferred a similar loss of MZBs as was observed in TSC1^BKO (see Supplementary material, Fig. S1c). Furthermore, examination of TSC1/PTEN^BKO spleen architecture did not reveal normal MOMA‐1 structures (Fig. 1). We conclude that the factors that control the architecture of the MZ and the loss of MZBs are downstream to the mTORC1 and are not a result of a feedback for the attenuation of AKT activity.

Impaired uptake by MZ macrophages in TSC1^BKO mice

To assess whether the impaired architecture in the TSC1^BKO mice is also associated with impaired function of the MZ, we applied a capturing assay that was used to demonstrate the MZ functional defect in CD19 KO mice.19 In this assay, mice were challenged with fluorescently labelled (Alexa 488) StaphA, also known as bioparticles. The bioparticles are introduced intravenously and their capturing by splenocytes is analysed within 30–60 min. Cells that engulfed the fluorescent bioparticles are gated, and MARCO expression is measured. For CD19 KO mice, which lack MZBs, this assay demonstrated a 40% reduction in the levels of MARCO among the bioparticle‐positive cells compared with WT.19 To assess the dynamic range of this assay in our hands, we administered the bioparticles to WT and RAG2 KO mice, which lack the adaptive system and have no detectable MZ. The expression of MARCO on the bioparticle‐positive cells was lowered by 50%, similar to what was described for the CD19 KO mice (see Supplementary material, Fig. S2). When applied to TSC1^BKO mice, an approximately 30% reduction in MARCO expression was observed on the bioparticle‐positive cells, confirming a defect in MZ function (Fig. 2).

Figure 2.

Impaired uptake of bioparticles by marginal zone (MZ) macrophages in TSC1^BKO mice. Wild‐type (WT) and TSC1^BKO mice were intravenously administered with fluorescently‐labelled bioparticles. Cells that engulfed the bioparticles were gated and the level of MARCO was assessed by flow cytometry. The mean expression level of MARCO of four mice from each group is presented. A significant reduction was recorded for TSC1^BKO mice (Mann–Whitney U‐test, P < 0·05). Experiments were performed twice on 4 mice per group.

TSC1 KO B cells home to the spleen but do not rescue MZ architecture

B cells, but not T cells, are required for the establishment of MZ architecture.5 However, little is known about the specific roles that B cells play in MZ establishment or maintenance. Importantly, MZBs themselves are not required for the establishment of MZ architecture. To address whether the impaired MZ microenvironment in the TSC1^BKO mice is recapitulated in an MZ formation assay, we performed a series of transfer experiments of B cells into JHT mice that lack B cells. This protocol was originally used to demonstrate the requirement of B cells for MZ formation and the intrinsic defect to restore the MZB subset in CD19 KO B cells.5, 20 Using loxp‐STOP‐loxp‐YFP mice crossed to the CD19‐Cre (WT/YFP) or TSC1^BKO (TSC1/YFP^BKO) mice, the transferred B cells are easily discerned owing to the YFP expression, which was equivalent for both donor mice (Fig. 3a). Two intravenous inoculations of 20 × 10⁶ splenic WT B cells into the B‐cell‐deficient JHT mice were sufficient to restore the typical MOMA‐1 rings (Fig. 3b, top). The MOMA‐1 rings were not observed following the transfer of TSC1 KO B cells, although both B‐cell genotypes homed to the spleen at similar efficiency following intravenous transfer (Fig. 3b, bottom). These data suggest that the lack of TSC1 in the B cells impairs the mechanisms that support the formation of MZ architecture.

Figure 3.

TSC1 knockout (KO) B cells do not reconstitute marginal zone (MZ) architecture. (a) Levels of YFP‐expressing cells in the spleen of wild‐type (WT)/YFP or TSC1/YFP^BKO mice. (b) Isolated splenic B cells of WT/YFP or TSC1/YFP mice were used for transfer into JHT hosts. Homing of donor cells to the spleen is shown, and a representative image of the MOMA‐1 architecture from the same spleen. Experiments were performed twice on four mice per group.

Induced deletion of TSC1 confers the disintegration of pre‐existing MZ

To further explore the possibility that mTOR induction by TSC1 deletion can lead to an active dissipation of MZ architecture, we crossed the floxed TSC1 mice to the RERT strain. The latter strain expresses a CRE‐ER fusion knocked into the heavy subunit of the RNA polymerase II subunit.21 Hence, administration of tamoxifen in vivo to RERT/TSC1^f/f induces the ubiquitous irreversible deletion of TSC1 in all cell types. Performing kinetic analyses revealed that MZ B cells and MZ architecture were maintained 15 days after tamoxifen administration. However, 35 days after the tamoxifen challenge both MZ B cells and the MOMA‐1 ring were no longer detected (see Supplementary material, Fig. S3).

mTORC1 activity suppresses lymphotoxin expression

Lymphotoxins play a major role in the generation of peripheral lymphoid tissues. Germline deletion of either LTα, LTβ or LTβR results in major defects in the generation of peripheral lymph nodes and spleen architecture in mice.22 MZ development is completely dependent on LT‐mediated signals during early organ development, as it is completely distorted in either of the aforementioned lymphotoxin KO mouse strains. Moreover, even a partial reduction in lymphotoxin expression in the B cells is sufficient to distort MZ architecture in the spleen.23 To our knowledge, most of the commercial reagents for LTβR were not validated for specificity. Following the testing of a few, we present data of immunofluorescence staining of spleen cryosections for LTβR (Abcam, #ab186847), which according to the company is reacting against the antigen in immunohistochemical applications. LTβR is expressed by multiple cell types of the spleen, including dendritic cells and innate lymphoid cells, but not in B or T cells. In WT mice, we observed a punctate pattern of staining with a few cells that stain bright, providing confidence that the antibody indeed identifies LTβR. As expected, there was no overlap of LTβR bright cells with the B cells, providing further credibility to the staining. In TSC1^BKO spleens, these brightly stained cells were not observed. A treatment with rapamycin restored this pattern of staining (we will refer to the panel that shows GB111‐NH₂ treatment later; Fig. 4a). These data suggest that upon mTORC1 activation in B cells, the surface expression of LTβR in the stroma cells is reduced. In addition to this, analysis of LTα and LTβ mRNA levels indicated a reduction only in LTα levels in TSC1 KO B cells. This reduction was completely reversed by in vivo rapamycin treatment, connecting the mTORC1 activity to the control of LTα transcripts (Fig. 4b).

Figure 4.

Lymphotoxin β receptor (LTβR) and LTα expression are reduced in TSC1^BKO mice. (a) A representative immunofluorescence analysis of LTβR expression in spleens of wild‐type (WT), TSC1^BKO and TSC1^BKO mice treated with rapamycin and the pan‐cathepsin inhibitor GB111‐NH₂. Spleen sections taken from two different mice for each group are shown (n = 4). (b) WT or TSC1^BKO mice were treated with rapamycin or vehicle for 10 days, then RNA was extracted from B cells, converted to cDNA and used for quantitative PCR analyses for LTα and LTβ relative to UBC, individual measurements are shown. A significant reduction in LTα transcript level was observed for TSC1 knockout (KO) B cells relative to WT, which was significantly elevated by the rapamycin treatment (Mann–Whitney U‐test, P < 0·05). Experiments were performed twice in four and five age‐matched mice.

Cathepsin activity is elevated in TSC1 KO B cells

Although not having the total LTβR KO mice to validate the specificity of the stainings, the data suggest that LTβR may undergo a proteolytic shedding, a phenomenon that has been described for human cells, but has not been reported in mice.24 Cathepsins are proteases that mainly reside in the lysosomes and have been shown to be involved in several normal and pathological processes. Cathepsins can translocate to the plasma membrane25 and can also be secreted upon stimulation and following transformation.26, 27 At first, we assessed the sensitivity of LTβR to cathepsin cleavage. To be able to capture the cell‐bound portion of LTβR following cleavage, we tagged the C‐terminus of LTβR with GFP and expressed the fusion protein in CHO cells, then exposed the cells to recombinant cathepsin B or L. After 3 hr with the cathepsins, a 45 000 molecular weight fragment of LTβR‐GFP was detected, consistent with the cleavage of the entire LTβR ectodomain (see Supplementary material, Fig. S4). Both enzymes displayed a similar proteolytic pattern. Next, we tested whether incubation with TSC1 KO B cells can promote the cleavage. CHO cells were transfected with LTβR‐GFP and divided into three equal portions. One portion was left untreated, one portion was incubated with WT B cells and one with TSC1 KO B cells. We clearly observed that the co‐culturing of the LTβR‐GFP‐expressing CHO cells with the B cells caused the reduction in the expression levels of the total LTβR‐GFP. Nonetheless, the incubation with TSC1 KO B cells, less with WT cells, resulted in the generation of a cleaved intracellular portion of LTβR‐GFP (see Supplementary material, Fig. S4). This fragment, however, was smaller than was observed with the recombinant cathepsin, implicating additional proteolytic processing promoted by TSC1 KO cells. These data encouraged us to measure cathepsin activity in WT and TSC1 KO B cells using our fluorescently label activity‐based probe. Activity‐based probes covalently bind and label the active site cysteine residue in a manner that is relative to the protease activity. Hence, epitope‐tagged or fluorescently labelled probes enable a direct assessment of the protease activity by SDS–PAGE and detection of the tagged protease. Using the Cy5 activity‐based probe GB123,16 we observed a strong elevation in the activity of cell‐associated cathepsin B, but no significant difference for cathepsin L (Fig. 5a). The same probe can also be used for fluorescence microscopy using cyto‐spun cells and flow cytometry. In all assays, deletion of TSC1 conferred a strong increase in the activation of total cathepsin activity (Fig. 5b,c). It was important for us to test whether the increase in cathepsin B activity in the TSC1 KO B cells is regulated by the mTORC1 activity. We therefore treated TSC1^BKO mice with rapamycin for 1 week, isolated B cells from the spleen and compared cathepsin B activity to vehicle‐treated mice. This analysis indicated a significant decrease in cathepsin B activity that was comparable to the difference seen between WT and TSC1 KO B cells (Fig. 6a,b). We also applied the GB123 labelling method to total splenocytes isolated from WT, TSC1^BKO and rapamycin‐treated TSC1^BKO mice. Following labelling, when gating on B cells, we observed a subset of cells that had an enhanced cathepsin activity in the TSC1 KO background. This completely normalized when the mice were treated with rapamycin (Fig. 6c). Taken together, our data unravel a novel connection between mTORC1 signalling and cathepsin regulation.

Figure 5.

TSC1 knockout (KO) B cells have an elevated cathepsin B activity. (a) B cells isolated from the spleens of five wild‐type (WT) and five TSC1^BKO mice were lysed and immediately subjected to labelling with GB123. Samples were analysed by SDS–PAGE and following a fluorescence scan, then gel was stained with Coomassie blue. Graphs show the quantification of cathepsin B activity relative to total protein load (Mann–Whitney U‐test, P < 0·05). (b) Isolated intact B cells were treated with GB123, then a portion of the cells was cytospun on cover glasses and fluorescently imaged, two typical images are presented. (c) Another portion of the cells was analysed by flow cytometry, quantification of different preparations of the flow analyses (n = 4). Red, cathepsins, blue, nucleus. The results in (a) and (b,c) were carried out in different mice from two individual experiments.

Figure 6.

Inhibition of mammalian target of rapamycin complex 1 (mTORC1) by rapamycin reduces cathepsin B activity. (a) TSC1^BKO mice were treated with rapamycin or vehicle for 10 days before B‐cell isolation. Following isolation, B cells were lysed and immediately subjected to labelling with GB123. Samples were analysed by SDS–PAGE and following a fluorescence scan, gel was stained with Coomassie blue. Activity of rapamycin was assessed by P‐S6 immunoblotting. (b) Graphs show the quantification of cathepsin B activities relative to total protein load. Cathepsin B activity was significantly reduced (Mann–Whitney U‐test, P < 0·05). The experiment was performed twice on cohorts of four mice per group. (c) Splenocytes were isolated from wild‐type (WT), TSC1^BKO and TSC1^BKO treated with rapamycin. Cells were labelled with GB123, extensively washed and stained with FITC‐conjugated anti‐B220 antibody. B cells were gated according to the forward/side scatter and B220 staining and Cy5 labelling is shown by the histograms. Two representative WT mice, two representative TSC1^BKO mice and three TSC1^BKO mice treated with rapamycin are shown out of cohorts of five age‐matched mice per group. Rapamycin activity was ascertained by P‐S6 immunoblotting.

The possibility of LTβR shedding is supported by the fact that based on the YFP reporter, the efficiency of KO is not complete. Between 70 and 80% of the total B cells were YFP positive in the TSC1 KO background. A subset of the YFP‐positive cells was B220^low to B220^null, probably representing the enhanced differentiation into plasma cells. A similar percentage of YFP reporter‐positive cells were observed in WT/YFP mice (see Supplementary material, Fig. S5). If this indeed represents the coexistence of WT and TSC1 KO B cells, it supports a dominant effect of the TSC1 KO B cells in perturbing MZ structure and function; perhaps by releasing factors detrimental to MZ maintenance, such as proteases.

In vivo inhibition of cathepsin activity replenishes MZ architecture in TSC1^BKO mice

In the last set of experiments, we assessed whether inhibition of cathepsins in vivo can restore MZ architecture in TSC1^BKO mice. To this end, we administered GB111‐NH_2, 28 a non‐fluorescent cathepsin inhibitor, to TSC1^BKO mice for 2 weeks. This dosing regimen was highly efficient in blocking all cathepsin activity in vivo within the spleen as assayed by ex vivo measurements of the residual activity using the fluorescently labelled probe GB123 (Fig. 7a). Although flow cytometry analyses did not detect the recovery of MZBs (not shown), probably owing to the short duration of treatment, remarkably, analysis of MZ structure revealed intact MOMA‐1 rings in the GB111‐NH₂‐treated mice as well as intact MAdCAM1 structures (Fig. 7b). We conclude that systemic inhibition of cathepsin activity suppresses the MZ structure phenotype of TSC1^BKO mice. This was also associated with the elevation of LTα expression (Fig. 7c) and the appearance of LTβR‐positive cells in the spleens of GB111‐NH₂‐treated TSC1^BKO mice (Fig. 4a).

Figure 7.

In vivo inhibition of cathepsins replenishes marginal zone (MZ) architecture and lymphotoxin α (LTα) expression in TSC1^BKO mice. Mice were treated with GB111‐NH₂ for 2 weeks. (a) Liver, kidney and spleen were harvested and extracts were reacted with GB123 to fluorescently image the residual cathepsin activity. Fifty micrograms of protein were analysed by SDS–PAGE following fluorescence scanning. A representative analysis indicates the in vivo inhibition of cathepsins by GB111‐NH₂ out of four treated mice. (b) A representative analysis of MOMA‐1 structures out of four independent TSC1^BKO mice treated with GB111‐NH₂ and of MAdCAM1 in the different mice_. (c) Individual values of quantitative PCR analysis for LTα in B cells isolated from four TSC1^BKO mice treated with vehicle and four treated with GB111‐NH₂. Expression of LTα was significantly increased (Mann–Whitney U‐test, P < 0·05). The experiment was performed twice.

Discussion

The function of the MZBs is regulated by multiple cell types that ensure the position of the cells and their ability to respond to stimulation. This includes at least neutrophils,29 innate lymphoid type 2 cells,30 follicular dendritic cells,31 macrophages and dendritic cells. However, all these interactions are contingent on generating the MZ architecture scaffolds. Because B and T cells constitute the vast majority of the spleen white cells, it is surprising that only B cells are critical to generate the MZ architecture. It is even more surprising that the MZBs, which are able to homeostatically proliferate, are not obligatory for this process. For example, in CD19 KO mice, although MZ macrophage maturation is impaired, the MOMA rings that signify the siglec1‐expressing metallophilic macrophage, are normal.20 A similar phenotype was observed in B cells that over‐express BCL3. While MZBs were completely absent, MOMA rings were intact.6 In this regard, the phenotype of TSC1^BKO is unusual with respect to the distorted MZ architecture (Fig. 1).

MZBs develop from transitional B cells in a manner that depends on B‐cell receptor signalling, nuclear factor‐κB and Notch2.32, 33 Although MZBs possess a strong mitogenic response to various stimulations, in vivo and ex vivo, particularly to Toll‐like receptor stimulation, whether they undergo in situ homeostatic proliferation is unknown. Bromodeoxyuridine labelling experiment indicate a turnover in the range of 1–2 weeks; however, following this period, half of the bromodeoxyuridine‐positive MZBs lingered for a much longer time, suggesting the existence of at least two pools of MZBs with respect to their turnover properties.34 This illustrates a gap in the understanding of the underlying mechanisms that regulate MZ formation and maintenance, which can be addressed by lineage tracing within the MZB pools, experiments that have not been performed.

We addressed the question of formation and stability of the MZ architecture and observed that the lack of TSC1 is associated with the inability to form de novo MZ structures upon B‐cell transfer and also a defect in its maintenance and function (Figs 2 and 3). These B‐cell intrinsic defects that affect the entire architecture of the MZ have been described for only a few molecules, primarily the lymphotoxin family of proteins. While lymphotoxins are expressed by several types of cells within the spleen,30 the role of lymphotoxins in B cells as critical for MZ formation was documented two decades ago.35, 36 CD19‐Cre‐mediated B‐cell‐specific knockout of LTβ is sufficient to confer a pronounced loss of MZ microarchitecture on all its parameters. Remarkably, this striking phenotype was observed notwithstanding a partial efficiency in the deletion.8 This suggests that a partial compromise in lymphotoxin signalling would be sufficient to confer a similar phenotype. LTβR is expressed in haematopoietic and stromal cells and both are required to mount a protective interferon response against viruses.37 The reduction in LTβR expression in the spleens of TSC1^BKO was unexpected and suggests communication between the B cells and their environment (Fig. 4), a result that prompted us to investigate further the possibility of proteolytic shedding.

Cathepsin proteases are mostly found in the lysosome, where the acidic conditions enhance their activity and allow proteolysis. However, cathepsins are not only confined to the lysosome and were shown to leak out to the cytoplasm and nucleus or be secreted following various conditions.26, 27 We have observed a greater than twofold increase in the mRNA levels of cathepsins L, D and B in TSC1 KO versus WT B cells (not shown), suggesting that in B cells, mTOR activity is connected to expression of several cathepsin genes. However, mRNA levels are not always correlated with proteolytic activity, particularly for zymogens, which require a processing step for activation. We therefore conducted activity assays to establish a positive connection between mTORC1 activity and cathepsin activity in B cells (Figs 5 and 6). Interestingly, when we analysed the effect of TSC1 deletion on cathepsin activity in other cell types, such as T cells and bone‐marrow‐derived macrophages, we did not observe significant differences (not shown). Furthermore, a study performed in mouse embryonic fibroblasts and HEK293T cells has shown that the inhibition of mTORC1 by TOR kinase inhibitors, and as a consequence the induction of autophagy, promoted lysosomal functions including that of cathepsin B as measured by fluorescent substrate.38 This infers the opposite from what we have found for B cells and although in that study the activity of cathepsins was not visualized directly, it suggests a B‐cell‐specific cross talk between cathepsin activity and mTOR. Indeed, increase in cathepsin expression levels was observed in various types of non‐Hodgkin's B‐cell lymphomas, such as diffuse large B‐cell lymphoma and chromic lymphocytic leukaemia,39, 40 and is associated with poor prognosis. Whether this is in a correlation with mTOR activity has not been investigated.

Cathepsins, particularly cathepsins B and G, can be secreted or displayed at the cell surface.25, 41 We have not been able to detect cathepsin activity released to the medium, when B cells were cultured for 24 hr in the presence of lipopolysaccharide. Whether cathepsins are released in vivo from the TSC1 KO B cells to mediate their action in the MZ environment is plausible, but difficult to directly ascertain. Prolonged treatment with GB111‐NH₂ was not tolerated by the mice. We therefore limited the treatment to 1 week, a time frame that may be too short for testing the recovery of the MZBs themselves. For rapamycin treatment, a period of 17 days was needed to see the reappearance of MZBs.13 However, it was clear that a week‐long in vivo treatment with the pan‐cathepsin inhibitor was sufficient to restore the MZ architecture in a manner indistinguishable from normal mice (Fig. 7) supported by a possible elevation in LTβR expression (Fig. 4).

Interestingly, the mRNA levels of LTα were also elevated by the cathepsin inhibitor, connecting the cathepsin to gene transcription (Fig. 7). Although a connection between cathepsins and gene transcription is mechanistically difficult to imagine owing to the compartmentalization of cathepsin to the lysosomes, multiple studies demonstrate that such connection indeed occurs. Nevertheless, the presence of cathepsins in the nucleus where they can cleave or interfere with the action of transcription factors was reported.42, 43 Interestingly, a connection between cathepsin B and extracellular‐signal regulated kinase (ERK) activity was reported using both inhibitors and knockdown experiments.44 Because ERK activity promotes the production of lymphotoxins,45 it is conceivable that this mechanism also operates in B cells to connect cathepsin B and LTα transcription. This possibility is under investigation with ramifications broader than only for lymphotoxin transcription.

The model that emerged from our study suggests that hyper‐mTORC1 in the B cells, through yet to be identified actions on cathepsin proteases, leads to global perturbations in MZ architecture, possibly by curtailing lymphotoxin signalling in the microenvironment. An alternative interpretation may be that the TSC1 deletion arrests the B cells in a premature transition state that does not support the generation of the MZ. Several evidences stand against this argument. First, TSC1‐deficient B cells yield normal numbers of follicular cells, undergo class switching, and differentiate into plasma cells upon stimulation.13, 14, 15 Thus, there is no maturation arrest. Second, the fact that incomplete deletion of TSC1, as assessed by the YFP reporter, was sufficient to abrogate MZ architecture, infers a dominant effect of mTORC1. Lastly, other mutations in B cells that block maturation should result in the loss of MZ architecture. Of note, the invariant chain KO mouse strain exhibits a severe block in maturation, almost complete loss of follicular B cells, yet displays normal architecture of the MZ.46 Together, the data strongly suggest that the observed phenotype relates to the activation of mTORC1 and is not an indirect consequence of an impaired maturation.

The mTOR pathway has been reported to promote cell invasion and metastasis.47 For several types of malignancies, primarily carcinomas, mTOR activity is higher in metastases than in the primary tumour site and is typically associated with a poorer prognosis.48, 49, 50 The connection between mTOR and metastatic potential has been mostly attributed to promoting endothelial to mesenchymal transition and allowing the cells to better sustain various stress conditions, such as hypoxia or nutrient starvation.51, 52 Our study implicates a novel signalling axis between mTOR and proteolytic activity with implications for the microarchitecture of the immune system.

Disclosures

The authors declare no conflict of interests.

Supporting information

Figure S1. Co‐deletion of PTEN does not restore marginal zone abnormality.

Figure S2. Impaired uptake of bioparticles by marginal zone macrophages in RAG2 knockout (KO) mice. Wild‐type and RAG2 KO mice were intravenously administered with fluorescently labelled bioparticles.

Figure S3. Induced deletion of TSC1 confers the disintegration of pre‐existing marginal zone and loss of marginal zone B cells.

Figure S4. Lymphotoxin β receptor is susceptible to proteolytic shedding by TSC1 knockout B cells.

Figure S5. Efficiency of deletion in CD19‐Cre mice.