Pulmonary B Lymphocytes Ameliorate Alzheimer's Disease-Like Neuropathology


 Increasing evidences reveal that the peripheral immune system is involved in the pathogenesis of Alzheimer's disease (AD). Here, we report that pulmonary B lymphocytes mitigate beta-Amyloid (Aβ) pathology in 5xFAD mice. The proportion of B cells, rather than T cells, increases within the brain, meningeal and lung tissues in 3-month-old 5xFAD mice. Deletion of mature B cells aggravates Aβ load and memory deficits of 5xFAD mice. Mechanistically, pulmonary B cells can migrate to the brain parenchyma and produce interleukin-35, which inhibits neuronal β-site APP-cleaving enzyme 1 expression, and subsequently reduces the production of Aβ. In turn, pulmonary B cell proliferation is associated with activation of the toll-like receptor/nuclear factor kappa-B pathway through elevated Aβ that is drained from the brain parenchyma to the lungs via meningeal lymphatics. Furthermore, promoting pulmonary B cell proliferation via overexpression of B-cell-activating factor ameliorates brain Aβ load and improves cognitive functions of 10-month-old 5xFAD mice. Together, these results highlight the lungs as both immune targets and effector organs in Aβ pathogenesis. Pulmonary B cells could serve as a potential target against AD.

anti-rat, -goat, -rabbit or -mouse IgG antibodies (Thermo Fisher Scientific, 1:1000) were 189 incubated for 1 or 2 h at room temperature in PBS. Then, DAPI was used to label the nucleus 190 following mounting with glass coverslips. Preparations were stored at 4 °C for no more than one 191 week until images were acquired using a confocal microscope (LSM 710 Laser Scanning 192 Confocal Microscope, Zeiss). Quantitative analysis was performed on the acquired images using 193 antibodies. Fluorescence data were collected with a FACS verse Cytometer (BD Bioscience) and 207 analyzed using FlowJo software (Tree Star, Inc.). In brief, singlets were gated using the height, 208 area and the pulse width of the forward and side scatter and viable cells were selected as FVS. 209 Cells were then gated for the appropriate cell-type markers. An aliquot of unstained cells of each 210 sample was counted using Automated Cell Counter (Merck) to provide accurate counts. 211

Sorting of pulmonary B cells 212
To obtain a suspension of pulmonary B cells using MACS, mice were euthanized by i.p. injection 213 of euthasol and transcardially perfused with ice-cold PBS with heparin. The lungs were quickly 214 collected and digested in RPMI-1640 medium with 1.4 U/ml collagenase VIII and 1 mg/ml DNase 215 I for 30 min at 37 °C. Individual samples with erythrocyte lysis were obtained after filtration 216 through a 70-μm nylon-mesh cell strainer. Cell suspensions were then pelleted, resuspended in 217 ice-cold MACS buffer containing anti-B220 microbeads (Miltenyi Biotec, 130-049-501) and 218 incubated for 30 min at 4 °C. Cells were rewashed and resuspended in ice-cold MACS buffer, then 219 passed through MACS column and collected for the following experiments. 220

B cells tracking experiments 221
B cells sorted from the lungs of 3-month-old WT or 5xFAD mice were initially labeled by PKH26 222 Red Fluorescent Cell Linker Kits (Sigma-Aldrich, MINI26) according to the instructions. 223 Afterwards, 3-month-old 5xFAD mice were infused with 2×10 5 cells per mouse via retro-orbital 224 injection. Twenty-four hours after transplantation, IVIS Spectrum In Vivo Imaging System 225 (Perkin-Elmer) was used to detect the distribution of labeled cells. Finally, the mice were 226 sacrificed for the following immunofluorescence analysis. 227

Western blotting 228
For Western blot analyses, homogenized protein samples of brain or lung were loaded onto 229 10-15% Tris/tricine SDS gels and transferred to PVDF membranes. After blocking for 1 h in 5% 230 nonfat milk/TBST, the membranes were incubated at 4 ℃ overnight with primary antibodies 231 (Supplementary Table 2). Horseradish peroxidase-conjugated secondary antibodies (Vector 232 Laboratories, USA) were used, and bands were visualized using ECL plus detection system. 233 GAPDH served as an internal reference for protein loading and transfer efficiency. Four mice per 234 group in duplicate experiments were averaged to provide a mean value for each group. 235 12

RNA extraction and sequencing 236
For total RNA extraction, tissues or cells were immersed in an appropriate volume of Trizol, 237 immediately snap-frozen in liquid nitrogen and stored at −80 °C until further use. After defrosting 238 on ice, samples were mechanically dissociated in extraction buffer and RNA was isolated using 239 the kit components according to the manufacturer's instructions (RNAiso Plus, Takara, 9109). The 240 Takara RNA Library Prep Kit was used for cDNA library preparation from total RNA samples. 241 Relative expression of mRNA for the target genes was performed by the comparative CT (△△CT) 242 method using GAPDH as control reference genes. The primers were listed in Supplementary 243 Table 3. 244 The RNA sequence of pulmonary B cells was performed by Cloud-seq. The raw sequencing reads 245 (FASTQ files) were first chastity filtered, which removed any clusters that exhibited a higher than 246 expected intensity of the called base compared to other bases. The quality of the reads was then 247 evaluated using FastQC, and after passing quality control, the expression of the transcripts was 248 quantified against the UCSC mm 10 genome using Salmon. These transcript abundances were 249 imported into R. EdgeR was used to normalize the raw counts and perform exploratory analysis 250 and differential expression analysis. The P values from differential expression analysis were 251 corrected for multiple hypothesis testing with the Benjamini-Hochberg false-discovery rate 252 procedure (adjusted P value). Functional enrichment of differential expressed genes, using gene 253 sets from Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) or gene-set 254 enrichment analysis (GSEA), was determined with Fisher's exact test as implemented in the 255 cluster Profiler Bioconductor package. Heat maps of the differential expressed genes and enriched 256 13 gene sets were generated with the R package "pheatmap". Normalized counts of selected 257 transcripts were used to calculate the fold change relative to respective controls. 258

ELISA Analysis 259
Frontal cortex samples were homogenized and sonicated in ice-cold TBS buffer containing 0.5 260 mM PMSF, 0.5 mM benzamidine, 1.0 mM DTT and 1.0 mM EDTA, followed by centrifugation at 261 12000 × g for 30 min. Supernatants were set aside for measurements of IL-35. Pellets were 262 re-suspended and further homogenized in 70 % formic acid (equal volume of TBS), then 263 centrifuged at 12000 × g for 1 h. The above indexes were quantified with ELISA kits from 264 Biolegend according the manufacturer's instructions. 265

Luciferase Assay 266
The promoter of hBACE1 gene was constructed into pGL3 plasmids and transfected into 267  Fluorescence was detected using the GloMax® 20/20 Luminometer. 272

DNA pulldown and mass spectrometry 273
Nuclear extracts were prepared as described above for EMSAs. 10 μg of biotinylated products were centrifuged at 10,000 g for 5 min, and a portion was used for protein quantitative 285 experiment with BCA method. 50 μL of cell lysis was used, followed by the addition of reaction 286 buffer and substrate. The fluorescence of each sample was detected using a full-wavelength 287 microplate reader. 288

Statistical analysis and reproducibility 289
The data analysts were blind to the identity of the experimental group. Statistical tests for each 290 figure were justified to be appropriate using Prism 6.0 (GraphPad Software, Inc.). Two-way 291 ANOVA, with Bonferroni's post hoc test was used for analysis of genotype, treatment and their 292 interaction effect. Two-group comparisons were made using two-tailed unpaired student's t-tests. 293 Repeated-measures ANOVA with Bonferroni's post hoc test was used for day versus treatment 294 comparisons with repeated observations. Data presented as mean ± s.e.m. 295

296
Compensatory protections of B lymphocytes in the early stage of AD-like pathology 297 15 Adaptive immunity has recently been implicated in AD pathogenesis 10-12 . For example, several 298 studies indicate that the number of peripheral blood B cells is positively correlated with the 299 cognitive scores of subjects, and decreases in patients with AD 13,14 . Loss of IgG produced by B 300 cells impairs microglial phagocytosis, thereby exacerbating Aβ plaque deposition in Rag-5xFAD 301 mice that lack T, B, and natural killer cells 15 . Recently, lymphatic vessels in the meninges provide 302 a crucial route for the connections between the brain and peripheral immune system 16,17 , although 303 their regulation in AD pathogenesis remains unknown. Based on this, we examined changes of T 304 and B lymphocytes in the brain parenchyma and meninges during AD-like disease progression of 305 5xFAD mice. Flow cytometry analysis revealed that the proportion of CD19 + B lymphocytes 306 increased significantly in the cortex and meninges of 5xFAD mice at 3 months , followed by a 307 continuous decline with age ( Fig. 1A-B and Supplementary Fig. 1A). However, the proportion of 308 CD3 + T cells from cortex and meningeal samples was not different between 5xFAD mice and WT 309 mice at age of 3, 5, and 8 months. 310 In order to elucidate the role of B cells in the pathophysiology of this AD model, we crossed 311 5xFAD mice with mature B cells-deficient (μMT -/-) mice to generate μMT -/-/5xFAD mice. 312 Three-month-old μMT -/-/5xFAD mice exhibited mild short-term memory impairments in both the 313 novel object recognition test and Y-maze test, compared with 5xFAD controls (Fig. 1C-D). These 314 mice also displayed impairments in long-term spatial learning and memory performances in the 315 Barnes maze (Fig. 1E). Consistently, pathological analysis revealed that accumulation of both 316 thioflavin S + and 6E10 + plaques was more obvious in the frontal cortex of three-month-old 317 μMT -/-/5xFAD mice than age-matched 5xFAD littermates (Fig. 1F). μMT -/-/5xFAD mice show a 318 more pronounced activation of GFAP + astrocytes and Iba1 + microglia with significant loss of 319 16 synaptic proteins synaptophysin (SYP) and postsynaptic density 95 (PSD95) in the frontal cortex 320 ( Fig. 1G and Supplementary Fig. 1B). These collective results suggest that compensatory 321 increases of brain B cells delays the occurrence of AD-like pathology of 5xFAD mice. 322 The migration of B lymphocytes from the lungs to the brain in the early AD-like stages 323 Recent evidence suggests there is a close relationship between the lungs and brain for immune 324 regulation, termed as the lung-brain axis 18-20 . Moreover, the lungs could potentially serve as the 325 main target organ of central antigens drained by meningeal lymphatics 21 . However, there is a lack 326 of direct evidence for this hypothesis. Therefore, in order to determine the source of increased B 327 cells in the CNS, we first detected B cells in relation to those found in the spleen and lungs. 328 Interestingly, the proportion of CD19 + B lymphocytes increased in the lungs of 3-month-old 329 5xFAD mice, but gradually decreased with age ( Fig. 2A-C). Furthermore, there were significant 330 positive correlations of B cells between the lungs and brain parenchyma or meninges (Fig. 2B). To elucidate how these lung-derived B cells change with AD-like disease progression, we first 337 conducted RNA-sequencing on sorted B cells from the lungs of 3-month-old 5xFAD mice and WT 338 mice. Gene Ontology (GO) analysis showed that differentially expressed genes between the two 339 groups were significantly enriched in cell migration ( Fig. 2D and Supplementary Fig. 2A), which 340 indicated migration ability of pulmonary B cells in 5xFAD mice. We then conducted B cell 341 transplantation experiments to examine whether lung-derived B cells can migrate to the brain. B 342 cells from the lungs of 3-month-old WT or 5xFAD mice were infused into 5xFAD mice of the 343 same age via retro-orbital injection. The in vivo fluorescent imaging demonstrated that more B 344 cells from 5xFAD lungs were detected in the brain at 24 h after transplantation compared with 345 those from WT mice (Fig. 2E). Together, these results indicate that B cells from the lungs migrate 346 efficiently into the brain, contributing to increased B cells in the early stages of amyloidogenesis 347 of 5xFAD mice. 348

IL-35 mediates pulmonary B lymphocytes against Aβ pathology in 5xFAD mice 349
In order to investigate the mechanism underlying the alleviating effect of lung-derived B cells on 350 AD-like pathology onset, we further compared the differentially expressed genes of pulmonary B 351 cells between 3-month-old WT mice and 5xFAD mice. GO analysis revealed that differentially 352 expressed genes were also significantly enhanced in the categories of immune activation, cytokine 353 production and secretion (Fig. 3A-B and Supplementary Fig. 2B). Notably, we found that Immunostaining demonstrated that IL-12a + Ebi3 + B cells accumulated in lung tissues of 363 3-month-old 5xFAD mice (Fig. 3C). Both mRNA and protein levels of IL-35, increased in 364 pulmonary B cells and the lung tissues of 3-month-old 5xFAD mice, respectively ( Fig. 3D and 365 Supplementary Fig. 2D). Consistent with increased proportion of B cells in the brain parenchyma, 366 Western blot and ELISA analysis confirmed that IL-12a and Ebi3 levels in the frontal cortex were 367 significantly higher in 3-month-old 5xFAD mice than WT mice (Fig. 3E and Supplementary Fig.  368 Fig. 3A-B). 377

IL-35 inhibits BACE1 expression in neurons via SOCS1/STAT1 pathway 378
Aβ accumulation within the brain is attributed to an imbalance between its production and 379 clearance 26 . In order to investigate the potential mechanisms of IL-35 reducing brain Aβ load in 380 5xFAD mice, we examined the expression of several enzymes and proteins involved in brain Aβ insulin-degrading enzyme (IDE) and low-density lipoprotein receptor-related protein 1 (LRP1) 386 ( Supplementary Fig. 3C). However, expression levels of BACE1, a key amyloidogenic enzyme, 387 significantly increased in the 5xFAD frontal cortex after IL-35 neutralization (Fig. 4A). Similar 388 changes were also observed in the cortex of 3-month-old μMT -/-/5xFAD mice when compared with 389 5xFAD controls (Supplementary Fig. 4A). We further investigated whether elevated BACE1 390 affects APP hydrolysis process. IL-35 neutralization or B cell deficiency significantly increased 391 amyloidogenic products soluble amyloid precursor protein  (sAPP) and C-terminal fragment  392 (CTF) in the frontal cortex of 5xFAD mice ( Fig. 4A and Supplementary Fig. 4B). These data 393 suggested that IL-35 inhibits BACE1 expression, subsequently reducing amyloidogenic products. 394 Previous studies reported that IL-12RB2 and GP130 form homodimers or heterodimers act as  Fig. 5C-H). Furthermore, BACE1 enzyme activity was significantly 403 reduced after IL-35 stimulation in primary cortical neurons (Supplementary Fig. 5I). These results 404 indicated that IL-35 inhibits neuronal BACE1 expression via its receptors. 405 Next, we investigated how IL-35 regulates the expression of BACE1. The -2036 to -1 bp region 406 upstream from the transcription start site (TSS) was divided into four segments (P1-P4) to make 407 20 partially overlapping reporter constructs, each 526 to 800 bp in length. Dual-luciferase assay 408 revealed that IL-35 significantly down-regulated hBACE1 promoter segments P4 activity (Fig.  409   4C). We further subdivided P4 into four segments, namely P401 to P404, encompassing only 30 410 bp. The results revealed that IL-35 regulated the BACE1 transcriptional level through the P403 411 (-127bp to -37bp) region (Fig. 4D). We subsequently identified the proteins binding P403 by DNA 412 affinity chromatography pull-down and mass spectrometry assay to elucidate the transcription 413 factors involved in the regulation of BACE1. The proteins showing significant interaction with the 414 proximal promoter are listed in Supplementary Table 1. In addition, these analyses also identified 415 clusters of spliceosome and ribosome proteins (RBPs) that strongly interacted with the 416 aforementioned promoter segment (Fig. 4E). Interestingly, transcription factor signal transducer 417 and activator of transcription 1 (STAT1) was also observed in the binding proteins, and 418 demonstrated interactions with the spliceosome and ribosome clusters (Fig. 4E). Previous studies 419 have revealed that STAT1 is a key transcription factor for BACE1 in neurons 28 . We further 420 confirmed the binding of STAT1 to hBACE1 P403 fragment by Western blot (Supplementary Fig.  421 5J). We also examined another BACE1 transcription factor, c-jun 28 , which was not detected in the 422 binding proteins (Supplementary Fig. 5J). Meanwhile, suppression of STAT1 activation is mainly 423 mediated by suppressor of cytokine signaling (SOCS), which plays an important role in the 424 regulation of intracellular inflammatory pathways 29 . Based on this, we tested the effect of IL-35 425 on SOCS/STAT1 signaling pathway in primary mouse neurons, N2a cells, and SH-SY5Y cells. 426 IL-35 stimulation significantly increased the expression of SOCS1, rather than SOCS3, in the 427 above three types of cultured cells (Supplementary Fig. 5C-H). Consistently, knockdown of 428 SOCS1 reversed the inhibitory effect of IL-35 on BACE1 expression and STAT1 phosphorylation 429 21 in SH-SY5Y cells (Fig. 4F). Additionally, decreased SOCS1 protein levels and increased STAT1 430 phosphorylation were also observed after IL-35 neutralization in the frontal cortex of 3-month-old 431 5xFAD mice (Supplementary Fig. 5K). The above results together indicated that IL-35 inhibited 432 the transcription of BACE1 via SOCS1/STAT1 signaling pathway. 433

TLR/NF-κB pathway 435
Typical lymphatic vessels have recently been identified in the human and mouse dura 6,7 , and serve 436 as an important drainage route for CNS macromolecules and antigens 3,16 . There is also a 437 glymphatic pathway that transports interstitial metabolites from the brain parenchyma to the 438 subarachnoid space 8,30 . Blocking meningeal lymphatic vessels or ligating deep cervical lymph 439 nodes (dcLNs) impairs glymphatic transport, resulting in an increased Aβ load 3,31 . These findings 440 suggests that the glymphatic and meningeal lymphatic systems constitute a functional 441 macromolecular drainage pathway. However, the peripheral target organs and biological effects of 442 these CNS antigens have yet to be fully determined. Therefore, we injected Evans blue into the 443 cistern magna and found the existence of the dye in the lungs (Fig. 5A). Furthermore, the dye 444 came into the lungs as soon as 8 min after injection. However, there was less Evans blue content 445 in the spleen than lungs at 8 min after injection (Fig. 5B). We further examined whether ligation of 446 the dcLNs would affect the ability of brain Aβ to reach the lungs. As expected, dcLNs ligation 447 reduced the content of cortical exogenous injected AF555-labeled Aβ in the lungs (Fig. 5C-D). 448 Furthermore, prior ligation of the dcLNs for one month aggravated deposition of Aβ plaques and 449 activation of microglia and astrocytes in 3-month-old 5xFAD mice ( Supplementary Fig. 6A-C). 450 We likewise found that polarization of aquaporin-4 (AQP4) was impaired after dcLN ligation 451 22 (Supplementary Fig. 6C), which may subsequently hamper glymphatic clearance of Aβ 32, 33 . 452 Consistently, endogenous oligo-A levels were significantly reduced in lung tissues of 453 3-month-old 5xFAD mice 1 month after dcLN ligation, without changes in the APP content (Fig.  454 5E-G). These results have revealed that macromolecules, including A drained by meningeal 455 lymphatic vessels, could be transported from brain parenchyma to dcLNs, enter the superior vena 456 cava, and finally reach the lungs via the pulmonary circulation. We also found that ligation of the 457 dcLNs significantly decreased the content of other specific central antigens, including 458 microtubule-associated protein 2 (MAP2) in the lung tissues ( Supplementary Fig. 6D). nuclear factor-kappa B (NF-ĸB) pathways were up-regulated significantly in 5xFAD mice when 466 compared to age-matched WT littermates (Fig. 5J). These findings were consistent with previous 467 reports demonstrating that activation of TLR and BCR participates in differentiation of B cell and 468 IL-35 secretion 35 . Interestingly, A has been reported to activate TLR and NF-ĸB pathways 36 , 469 suggesting that A may contribute to IL-35 secretion from lung-derived B cells in the early 470 AD-like stage of 5xFAD mice. In order to further verify this hypothesis, hA1-42 peptides were 471 injected intratracheally into the lungs of three-month-old WT mice. The results show that mRNA 472 levels of IL-12a and Ebi3 increased significantly in pulmonary B cells 5 days after injection (Fig.  473 23 5K). Consistently, TLR and NF-ĸB pathways associated genes were markedly up-regulated as 474 well ( Supplementary Fig. 6E). Together, these results suggests that the lungs not only receive CNS 475 lymphatic draining A, but also serve as an immune effector organ of A. short-term and long-term memory of 3-month-old 5xFAD mice, accompanied with more severe 495 Aβ pathology. We further discovered lung-derived B cells can migrate to the brain parenchyma 496 and produce IL-35 that inhibits neuronal BACE1 expression through the SOCS1/STAT1 pathway, 497 subsequently reducing the production of Aβ. In turn, the proliferation of pulmonary B cells is 498 associated with activation of the TLR/NF-κB pathway by elevated Aβ, which is drained from the 499 brain parenchyma to the lungs via meningeal lymphatics. Blocking meningeal lymphatic drainage 500 markedly reduces Aβ levels and B cell number in the lungs. Furthermore, promoting pulmonary B 501 cell proliferation and IL-35 secretion via overexpression of BAFF ameliorates AD-like pathology 502 in 5xFAD mice. These results collectively suggest that the lungs serve as sites for B cell activation 503 in response to brain Aβ accumulation and migration in order to inhibit Aβ production via the 504 IL-35/SOCS1/BACE1 pathway ( Supplementary Fig. 7). 505 Anatomical evidence has been found in our understanding of the relationship of brain and 506 peripheral tissues. Dural lymphatic vessels absorb CSF from the adjacent subarachnoid space and 507 brain interstitial fluid (ISF) via the glymphatic system. Dural lymphatic vessels transport CSF into 508 dcLNs via foramina at the base of the skull 39 , and also CSF drains directly from the subarachnoid 509 space into nasal lymphatics 40 , which contributes to macromolecules such as antigens drained into 510 the peripheral tissues. Moreover, connections between the lungs and brain have been documented 511 recently 18-20 . For example, activated T cells in the EAE model are observed to be enriched in the 512 lungs before invading the brain parenchyma, where they up-regulate the expression of adhesion 513 molecules and chemokines 18 . Additionally, disruption of meningeal lymphatic drainage diminishes 514 pathology and reduces the inflammatory response of brain-reactive T cells in an animal model of 515 multiple sclerosis (MS) 16 . However, the anatomical basis for the brain-lung axis remains elusive. 516 25 The present results have revealed this "black box". Aβ and a variety of central antigens can reach 517 the lungs through meningeal lymphatic vessels. This study further indicates that decreased brain 518 infiltration by pulmonary B cells also partially contributes to increased Aβ load following the 519 blockage of meningeal lymphatic drainage. Inhibition of BACE1 expression and activity to reduce Aβ load in the brain has been considered as 555 a key target for AD therapy. However, the safety and side effects of small molecule inhibitors of 556 BACE1 are of concern 50 . Therefore, regulation of pulmonary B cells to produce IL-35 may 557 provide a new avenue against AD, but its clinical efficacy and safety need to be further defined. In 558 addition, the mechanism of reduced production of lymphocyte B and IL-35 in the late stages of 559 AD-like disease progression still needs further elucidation. 560 27

561
The present findings highlight the nvolvement of the "lung-brain" axis in the Aβ related pathology. 562 Brain derived Aβ was transported to the lungs, subsequently activating B cells and promoting the 563 production of IL-35. These activated B cells had ability of migrating to the brain and then 564 down-regulated neuronal BACE1 expression, consequently inhibiting Aβ production. Furthermore, 565 promoting pulmonary B cell proliferation and IL-35 secretion via overexpression of BAFF 566 ameliorates AD-like pathology in 5xFAD mice. These results collectively suggest that the lungs 567 serve as sites for B cell activation in response to brain Aβ accumulation and targeting pulmonary 568