“A distinct neutrophil population invades the central nervous system during pancreatic cancer”

Weight loss, fatigue, and cognitive dysfunction are common symptoms in cancer patients that occur prior to initiation of cancer therapy. Inflammation in the brain is a driver of these symptoms, yet cellular sources of neuroinflammation during malignancy are unknown. In a mouse model of pancreatic ductal adenocarcinoma (PDAC), we observed early and robust myeloid cell infiltration into the brain. Infiltrating immune cells were predominately neutrophils, which accumulated at a unique central nervous system entry portal called the velum interpositum, where they expressed CCR2. CCR2 knockout mice had significantly decreased brain-infiltrating neutrophils as well as attenuated anorexia and muscle catabolism during PDAC, without any changes in neutrophils in other organs. Lastly, intracerebroventricular blockade of the purinergic receptor P2RX7 during PDAC abolished neutrophil recruitment to the brain and attenuated anorexia. Our data demonstrate a novel function for the CCR2/CCL2 axis in recruiting neutrophils to the brain, which drives anorexia and muscle catabolism.


Circulating myeloid cells infiltrate the brain early in PDAC 126
We first investigated whether circulating immune cells infiltrate the brain 127 throughout the course of PDAC. We utilized a mouse model of PDAC, generated 128 were by far the most numerous invading myeloid cell type, constituting 34% 155 percent of CD45 high CD11b+ cells in sham animals, and increasing to nearly 54% 156 by 10.d.p.i. in tumor animals (Fig. 1F). 157 Since the CD45 high CD11b+ population we defined as invading circulating 158 myeloid cells could also contain activated microglia, we generated GFP+ bone 159 marrow chimera to determine if the majority of these cells were of peripheral 160 origin. Furthermore, the population of CD45 high CD11b+Ly6C low myeloid cells 161 could also be activated microglia. We generated GFP+ bone marrow chimera 162 mice through conditioning WT mice with treosulfan to ablate marrow, then 163 transplanting marrow from pan-GFP mice (Ly5.1 GFP ) (Figure 1 -figure  164 supplement 2A). This system is advantageous because, unlike other alkylating 165 agents, treosulfan does not cross or disrupt the blood brain barrier (Capotondo et 166 al., 2012). On average, mice that underwent bone marrow transplant (GFP BMT 167 mice) exhibited 75% chimerism (Figure 1 -figure supplement 2C). In agreement 168 with results from WT marrow animals, we observed that at 10 d.p.i., thousands of 169 GFP+ myeloid cells infiltrated the brain in tumor animals (Figure 1 -figure  170 supplement 2B). The majority of these cells were neutrophils, with a concurrent supplement 2F), suggesting that the increase in CD45 high CD11b+Ly6C low cells in 176 our WT marrow PDAC mice was a result of microglia activation, rather than 177 infiltrating monocytes. 178 Taken together, these data show that myeloid cells infiltrate the brain 179 during PDAC, which correlates with symptom onset, and that there is a selective 180 neutrophil recruitment. Since the purpose of this study was to investigate 181 infiltrating cells, we chose to focus our subsequent analysis on myeloid cells, with 182 an emphasis on neutrophils. 183 increase in overall CD45 immunoreactivity, these cells appeared ramified rather 245 than globoid (60X inset in Fig. 2D), suggesting microglia activation rather than 246 immune cell infiltration. We did not observe any CD45+ cells in the lateral 247 parabrachial nucleus (data not shown), which was implicated in cancer-248 associated anorexia (Campos et al., 2017). This was perhaps due to its lack of 249 proximity to a circumventricular organ or meninges. Interestingly, we observed animal at 10 d.p.i., with 60X inset shown on the right, along with quantification of 266 MPO+ and total CD45+ cells. Scale bar for 20X images = 100µm. Scale bar for 267 60X insets = 10µm. Data are presented as mean ± s.e.m., n = 5/group, *P < 268 0.05, **P < 0.01 in student's t-test. 269     The peripheral origin of the CD45+ globoid cells in the brain was assessed 299 using our GFP BMT mice. Sham BMT mice showed very few GFP+ cells in the 300 brain, including the cortex and thalamus (Fig. S2A), as well as the meninges 301 (data not shown). In contrast, there was a large increase in GFP+ cells in the 302 brains of KPC mice at 10 d.p.i. We observed a pattern of infiltrating GFP+ cells 303 that was identical to CD45+ globoid cells in our previous experiments, with    Using in situ hybridization we localized robust CCL2 mRNA expression 363 exclusively within the VI during PDAC. There was no observable Ccl2 mRNA in 364 the brains of sham animals (Fig. 3B). We verified these results at the protein 365 level using CCL2 mCherry mice, which showed abundant CCL2 protein expression 366 in the VI in tumor animals at 10 d.p.i., exclusively expressed in Iba1+CD206+ 367 meningeal macrophages. CCL2 protein was not expressed in VI meningeal 368 macrophages in sham mice (Fig. 3C). We did not observe robust CCL2 protein 369 expression in any other locations in the brain.  Representative 10X confocal microscopy image of brain from CCL2 mCherry tumor 395 mouse brain, 10 d.p.i. Scale bar = 100 μm. Inset of VI shows CCL2 protein 396 expression is confined to meningeal macrophages, identified by CD206 labeling. 397 Scale bar = 20 μm. D) Representative 20X confocal microscopy image of brain 398 from CCR2 RFP/WT tumor (10 d.p.i.) and sham mouse brain. Scale bar = 100 μm.

415
Based on our data showing that CCR2 is a brain-specific chemotactic 416 receptor for neutrophils during PDAC, we hypothesized that brain-infiltrating 417 neutrophils are unique compared to neutrophils that infiltrate other organs. In 418 order to characterize the phenotype of brain-infiltrating neutrophils during PDAC, 419 we performed RNA sequencing (RNAseq) on FACS-sorted neutrophils from 420 blood, liver, tumor, and brain during PDAC at 10 d.p.i, as well as circulating 421 neutrophils from sham animals (Figure 3 - Figure   Taken together, these data indicate that the CCR2-CCL2 axis is activated 430 in the CNS during PDAC, and that the CNS microenvironment uniquely 431 influences the neutrophil transcriptome during PDAC. 432 Based on our findings that CCR2+ cells infiltrate the brain during PDAC, 477 we hypothesized that CCR2 is required for immune cell recruitment to the brain. We observed that at 11 d.p.i., there was a 37% decrease in total CD45 high 479 myeloid cells in the brains of CCR2KO tumor mice compared to WT tumor mice 480 ( Fig. 4A and B). This difference was primarily driven by a large decrease in 481 brain-infiltrating neutrophils in CCR2KO tumor mice, and to a much lesser extent 482 a decrease brain-infiltrating Ly6C high monocytes in CCR2KO tumor mice 483 compared to WT tumor mice. There was a decrease in neutrophils (and again 484 Ly6C high monocytes, to a lesser extent) as a percentage of the CD45 high cells in 485 the brain in CCR2KO tumor mice, indicating that the differences were not due to 486 a global decrease in infiltrating immune cells (Fig. 4B). This was also supported 487 by the fact that there were no differences in microglia (data not shown), Ly6C low 488 monocytes, or T-cells in the brains of CCR2KO tumor mice compared to WT 489 tumor mice (Fig. 4C). . We observed that CCR2 knockout (CCR2KO) mice had 500 decreased anorexia during PDAC compared to WT tumor mice (Fig. 4A). 501 CCR2KO tumor mice also had attenuated muscle loss compared to WT tumor 502 mice (Fig. 6B). To determine whether the decreased muscle mass loss in 503 CCR2KO mice was due to decreased muscle proteolysis, we assessed levels of 504 transcripts key for muscle proteolysis in the gastrocnemius, including Mafbx, 505 in WT tumor mice, we found the largest decrease to be in the brain and observed 544 a slight increase in circulating neutrophils in CCR2KO tumor mice compared to 545 WT tumor mice (Fig. 4 -Figure Supplement 1F and H), suggesting that the 546 decrease in brain-infiltrating neutrophils was due to a homing defect rather than 547 inability to mobilize from the bone marrow. Therefore, our data show that CCR2 548 is important for neutrophil recruitment specifically to the brain, and that the 549 decrease in brain-infiltrating neutrophils was due to a homing defect rather than 550 inability to mobilize from the bone marrow. 551 Analysis of neutrophils and Ly6C high monocytes in brain, liver, tumor, and blood 577 in CCR2KO tumor mice, normalized to number in WT tumor mice. n = 5-9/group. 578 . Lastly, although there was a 47% decrease in brain-infiltrating Ly6C high 579 monocytes in CCR2KO tumor mice compared to WT tumor mice, we also 580 observed a 59% decrease in circulating Ly6C high monocytes, suggesting that, 581 unlike neutrophils, the decrease in brain-infiltrating Ly6C high monocytes was in 582 fact due to a defect in marrow extravasation. 583 584 Taken together, these data show that the CCR2-CCL2 axis required for 585 myeloid cell recruitment specifically to the CNS during PDAC, and that this axis is 586 important for development of anorexia and muscle catabolism. 587 588

Blockade of P2RX7 in the CNS prevents immune cell infiltration into the 589 brain and attenuates cachexia during PDAC 590
To evaluate the effects of CNS inflammatory responses during PDAC 591 independent of potential systemic effects, we treated mice with 592 intracerebroventricular (ICV) oxidized ATP (oATP). This potently blocks 593 purinergic receptor P2RX7 signaling on brain resident macrophages. Signaling 594 through this receptor is key for neutrophil recruitment to the brain during 595 neuroinflammation (Roth et al., 2014). Animals were surgically implanted with 596 indwelling lateral ventricle cannulas, then inoculated with KPC cells one week 597 later. Mice received daily ICV injections of either 500 ng oATP or vehicle (aCSF), 598 starting 3 d.p.i. (Fig. 5A). oATP treatment completely prevented both neutrophils 599 and total CD45 high myeloid cells from infiltrating the brain (Fig. 5B). were completely absent in oATP-treated tumor animals, compared to large 605 infiltrates in aCSF-treated tumor animals (Fig. 5C). While we did observe sparse 606 CD45+ cells in the VI in oATP-treated tumor animals, they were not globoid and 607 resembled meningeal macrophages. We also observed that oATP treatment 608 attenuated anorexia in tumor mice (Fig. 5D). There was trend toward increased 609 gastrocnemius mass (P = 0.09) in oATP-treated tumor mice compared to aCSF-610 treated tumor bearing mice (Fig. 5E), which corresponded to a trend toward 611 decreased induction of genes associated with proteolysis in gastrocnemius 612 muscle (Fig. 5F), demonstrating that muscle catabolism was moderately 613 attenuated by oATP administration directly into the brain. Tumor size in oATP-614 treated tumor mice was identical to that of aCSF-treated tumor mice (   (Zenaro et al., 2015). Therefore, our results, along with 714 previous studies, implicate brain-infiltrating myeloid cells as key players in driving 715

CNS-mediated signs and symptoms during inflammatory disease. 716
We observed a decrease in total number of lymphocytes in the brain 717 We showed that CCR2KO mice exhibited significantly attenuated myeloid 742 cell infiltration into the brain, as well as decreased anorexia and muscle 743 catabolism, during PDAC. As discussed above, these results are in agreement 744 with previous studies investigating sickness behaviors during inflammatory liver 745 disease, which showed that CCR2KO mice exhibited attenuated monocyte 746 infiltration into the brain, along with decreased sickness behaviors (D'Mello et al., indicating that CCR2 is important for neutrophil recruitment specifically to the 766 brain. Circulating neutrophils in sham animals did not express CCR2, but a small 767 percentage of circulating neutrophils expressed CCR2 in tumor animals. There 768 was no increase in CCR2-expressing neutrophils in the liver during PDAC. 769 Alternatively, a significant percentage of neutrophils in the brain expressed CCR2 770 during PDAC, meaning a distinct population of neutrophils is recruited to the 771 brain from the circulation. In addition, there was actually a small increase in 772 circulating neutrophils in CCR2KO tumor mice, ruling out the possibility that 773 neutrophils were unable to extravasate out of the marrow. These results, along 774 with our RNASeq data (discussed below), suggest that the population recruited 775 to the brain has a distinct function from those recruited to other organs. 776 We administered oxidized ATP, a purinergic receptor antagonist, directly 777 into the brain and observed complete abrogation of circulating myeloid cell 778 recruitment to the brain in tumor animals, as well as anorexia attenuation. These 779 results provide key mechanistic insights to show that brain inflammation is key for 780 PDAC-associated anorexia. While there was no change in microglia morphology 781 Roth et al., 2014), we cannot rule out the possibility that the difference in 783 anorexia we observed were due to changes in microglia phenotype. The 784 presence of an indwelling lateral ventricle cannula may have also induced 785 microglia activation and influenced morphology quantification. However, we did 786 take care to acquire images from the contralateral hemisphere. Furthermore, we 787 observed an increased Ly6C high monocyte infiltrate in our aCSF-treated tumor 788 animals compared to non-cannulated tumor animals, suggesting the indwelling 789 lateral ventricle cannula did affect the inflammatory response in the brain to at 790 least a small degree. Nevertheless, oATP completely prevented myeloid cells 791 from infiltrating the brain during PDAC, strongly implicating these cells as 792

mediators of anorexia. 793
A few limitations should be considered when interpreting results of this 794 study. First, our data were produced in a single model of pancreatic cancer. 795 While our model is extensively characterized and reliably recapitulates many of 796 the CNS-mediated symptoms observed in humans, other malignancies should 797 also be considered. Second, it is possible, even likely, that circulating immune 798 cells infiltrate and influence function in other organs dysfunctional during cancer 799 (skeletal muscle, adipose tissue, etc.). However, the purpose of this study was to 800 investigate and characterize interactions between circulating immune cells and 801 the brain during PDAC. Therefore, we chose to focus specifically on the brain so 802 as to not overcomplicate analysis. Third, we did observe a small increase in 803 Ly6C hi monocytes in the brains of animals during PDAC, which was attenuated 804 by CCR2 deletion. We cannot rule out that these cells did not contribute to 805 anorexia and muscle catabolism. However, there were far fewer Ly6C hi 806 monocytes (≈2,000) in the brain than neutrophils (≈9,000) during PDAC, and 807 these cells only constituted about 15% of brain CD45 high myeloid cells (vs. 808 about 50% for neutrophils). Lastly, despite our extensive analysis, we cannot rule 809 out with absolute certainty that the differences we observed in CCR2KO mice 810 were not due to differences in tumor response. However, both the CCR2/CCL2 811 axis and neutrophils are reported to be "pro-tumor" ( In summary, we demonstrated that myeloid cells infiltrate the CNS 820 throughout the course of PDAC and that preventing myeloid cells from infiltrating 821 the brain attenuates anorexia and muscle catabolism. We showed there are 822 distinct mechanisms for immune cell recruitment to the brain during systemic 823 inflammation, and demonstrate a novel role for CCR2 in neutrophil recruitment to 824 the brain, providing key insights into mechanisms of neuroinflammation and 825 associated symptoms. hrs at 4°C before being stored at −80°C until used for immunohistochemistry. 926 Immunofluorescence immunohistochemistry was performed as described below. The following primary anti-mouse antibodies were used, with company, 937 clone, host species, and concentration indicated in parentheses: CD11b 938 (eBioscience, rat, M1/70, 1:1000), CD45 (BD, rat, 30-F11, 1:1000), 939 myeloperoxidase (R&D, goat, polyclonal, 1:1000), Ly6G (Biolegend, 1A8, rat, 940 1:250), Iba-1 (Wako, Rabbit, NCNP24, 1:1000), CD206 (Bio-rad, rat, MR5D3, 941 1:1000), ER-TR7 (Abcam, rat, ER-TR7, 1:1000), and citrillunated histone H3 942 (Abcam, rat, polyclonal, 1:1000). We also used a chicken anti-mCherry antibody 943 (Novus Biologicals, polyclonal, 1:20,000), to amplify mCherry signal in sections 944 from CCL2 fl/fl mice and a rabbit anti-RFP antibody (Abcam, polyclonal, 1:1000) to 945 amplify RFP signal in sections from CCR2 RFP/WT mice. 946 The following secondary antibodies were used, all derived from donkey 947 and purchased from Invitrogen, with dilution in parentheses: anti-goat AF488 948 (1:500), anti-rabbit AF555 (1:1000), anti-rat AF555 (1:1000), anti-rat AF633 949 (1:500), and anti-chicken AF555 (1:1000). 950 951

Image acquisition and analysis 952
All images were acquired using a Nikon confocal microscope. Cell quantification 953 was performed on 20X images using the Fiji Cell Counter plugin by a blinded 954 researcher. CD45+ cells were defined as CD45 bright globoid cells, and 955 neutrophils were defined as CD45+ MPO+ cells. The velum interpositum (VI) was 956 defined as the layer of meninges (identified by appearance of staining 957 background) between the hippocampus and thalamus, from bregma -1.7 to -2.6 958 mm. At least 8 VI images were quantified from each animal. The median 959 eminence was defined as the base of the mediobasal hypothalamus (far ventral 960 part of the brain), adjacent to the third ventricle from bregma -1.95 to -2.5 mm. 961 Four ME images were quantified from each animal. The area postrema was 962 defined as the region in from bregma -7.2 to -7.75 mm. Four area postrema 963 images were quantified from each animal. 964 965

Microglia morphology analysis 966
Microglia activation in the hippocampus was quantified using Fiji (ImageJ, 967 NIH). Five images of the dentate gyrus were acquired from each animal. Images 968 were 2048 x 2048 pixels, with a pixel size of 0.315 µm. Images were uploaded to 969 Fiji by a blinded reviewer (KGB) and converted to 8-bit greyscale images. After 970 thresholding, microglia were identified using the "analyze particle" function, which 971 measured mean Iba-1 fluorescent intensity per cell, cell area, and percent area 972