In vivo CHI3L1 (YKL-40) expression in astrocytes in acute and chronic neurological diseases

Ten de-identified CSF samples from each of AD, MS, ALS, stroke and normal control patients were obtained from the Human Brain and Spinal Fluid Resource Center (UCLA, CA) and the Center for ALS Research at the University of Pittsburgh. Cortical tissue samples from three cases of AD and schizophrenia, two cases of infarct and Pick's disease and one case of MS and ALS were obtained from the Human Brain and Spinal Fluid Resource Center (UCLA, CA) and used for ISH and immunohistochemical analyses. All human studies were approved by the corresponding institutional review boards.

Previously banked brains from four pigtailed macaques (Macaca nemestrina) infected with SIVDeltaB670 viral swarm (SIVdB670) that developed encephalitis and three macaques that did not develop encephalitis were used for in situ hybridization (ISH) and immunohistochemistry.

Sense and antisense CHI3L1 DNA templates containing the T7 promoter were generated by PCR from the pUC57 vector (GenScript, Piscataway, NJ) containing the full length YKL40 cDNA (NCBI Reference Sequence: NM_001276.2). ³⁵S-labeled RNA probes were generated using MAXIscript in vitro transcription kit (Ambion, Austin, TX). After removal of paraffin, cortical tissue sections from patients with MS, AD, Pick's disease, ALS, stroke and schizophrenia and SIV-infected pigtailed macaques were processed for in situ hybridization (ISH) and then for immunohistochemistry. For ISH, tissue sections were microwave oven treated and then incubated in hybridization buffer (1× HYB buffer, 0.6 M NaCl, 10% dextran, 50 μg/ml tRNA, 0.1 M DTT) containing radiolabeled CHI3L1 probe (50,000 cpm/μl) in 50°C over night. The following day the tissue sections were washed and processed for immunofluorescence. Tissue sections were exposed to emulsion (Eastman Kodak company, Rochester, NY) for 10 days in 4°C and then ISH signal was visualized using D19 (Sigma-Aldrich, St. Louis, MO) and then fixed in Rapid fix (Sigma-Aldrich, St. Louis, MO).

Formalin-fixed, paraffin-embedded, 6-micrometer thick tissue sections were deparaffinized in Histoclear (National Diagnostics, Atlanta, GA) and rehydrated for 3 minutes in 100%, 95%, and 70% alcohol followed by PBS. Endogenous peroxidase activity was inactivated by immersing the section in 3% H₂O₂ for 20 minutes. Antigen unmasking was performed using antigen retrieval Citra^® solution (BioGenex, San Ramon, CA). Tissue sections were blocked with protein blocking agent (Thermo, Pittsburgh, PA) for 20 minutes. GFAP staining was performed using polyclonal rabbit anti-human GFAP (1:500; DAKO, Carpinteria, CA). Iba1 staining was performed using rabbit anti-human Iba1 (1:500; Waco Chemicals, Richmond, VA) and CHI3L1 staining was performed using goat anti-human chitinase 3-like 1 (1:1000; R&D systems, Minneapolis, MN). These staining were followed by goat anti-rabbit Dylight 488 antibody (1:200; Jackson ImmunoResearch Laboratories, West Grove, PA) or mouse anti-goat biotin-conjugated antibody (1:200; Jackson ImmunoResearch Laboratories, West Grove, PA).

Five fields from each case were captured by confocal microscopy and analyzed for the number of ISH-positive cells or ISH/IHC pixels per field (LSM 510, Carl Zeiss MicroImaging, Inc., Thornwood, NY). Illumination was provided by Argon (458, 477, 488, 514 nm, 30 mW), HeNe (543 nm, 1 mW) lasers. Digital images were captured with LSM 510 version 4.2 (Zeiss).

Infarct presence or absence was established by a neuropathologist (RLH), then infarct age was charachterized as acute, subacute, or chronic based upon neuropathological features seen on hematoxylin staining as previously described [17]. Due to the focal nature of infarct pathology a semi-quantitative scoring was performed by a neuropathologist (CAW) blinded to clinical characteristics in 4 cases of acute infarct, 5 cases of subacute infarcts and 5 cases of chronic infarct. ISH scoring of CHI3L1 was assigned a score from 0 to (++++) based on ISH intensity: no siganl = 0, occasional positive cells = (+/-), weak signal = (+), moderate signal = (++), strong signal = (+++) and intense signal = (++++).

CHI3L1 levels were determined, in duplicate, for all of the CSF samples using the MicroVue CHI3L1 ELISA kit from Quidel Corporation (San Diego, CA) according to the manufacturer's protocol. Absorbance was measured using a microplate reader (BioTek Instruments, Inc., Winooski, VT).

Average ± standard error (SEM) of CSF CHI3L1 concentrations from MS, AD, ALS, stroke patients and healthy young and aged controls were graphed. The Mann-Whitney test was used to computes the statistical difference between disease groups and control patients. CHI3L1 ISH and GFAP imuunohistochemistry pixels were captured by confocal microscopy from 5 random fields of each of the SIVE, SIV, infarct, MS, AD, Pick's disease and schizophrenia cases. Average pixels per filed were calculated and a linear regression was used to determine a correlation between CHI3L1 pixel count and GFAP pixel count.

We have previously observed elevated CHI3L1 concentration in the CSF of SIV and HIV encephalitis [11] and therefore wanted to assess its concentration in a broad spectrum of human neurological diseases. CSF CHI3L1 levels of MS patients (age range: 25 to 58) were compared to young healthy controls (age range: <40) and were found to be significantly elevated (p = 0.0007) (Figure 1A, Table 1). CSF samples from healthy aged controls, AD, ALS and stroke showed statistically significant elevated levels of CHI3L1 compared to young controls. However there was no statistical difference between AD (age range: 59 to 86; median age: 75), ALS (age range: 32 to 74; median age: 62.5) and stroke (age range: 60 to 89; median age: 72) when these values were compared to healthy older age-matched controls (age range: 60 to 70, median age: 65). CSF CHI3L1 levels in healthy older people are significantly higher then healthy young controls (p = 0.0126) which implies that there is a modest but significant increase of CSF CHI3L1 with aging (Figure 1B, Table 1).

Since our previously published in vitro studies demonstrated that YKL40 was present in the supernatant of macrophages and not astrocyte or neuronal cultures, we were surprised to observe YKL40 staining localized with astrocytes in regions of microglial nodules and occasional activated macrophages/microglia in SIVE and HIVE. To elucidate the cellular source of CHI3L1 in SIVE encephalitis we performed ISH combined with immunohistochemistry. In order to validate the presence of CHI3L1 protein with its transcript we performed combined CHI3L1 ISH with immunohistochemistry for CHI3L1 on brain tissue from pigtailed macaques that developed SIVE. This confirmed that the antibody against CHI3L1 stained the cells that expressed the CHI3L1 transcript (Figure 2). The sense probe showed no significant hybridization (data not shown). Since the signal was more robust with ISH and we wanted to identify the cellular source of CHI3L1, we performed combined CHI3L1 ISH with immunohistochemistry for GFAP as a marker for astrocytes, NeuN as a marker of neurons and Iba1 as a marker for microglia/macrophages. CHI3L1 transcripts were shown to co-localize with GFAP staining (Figure 3A-C) but not with NeuN (Figure 3D-F) or Iba1 staining (Figure 3G-I) in SIVE.

These results suggest that while macrophages can synthesize YKL40 in vitro, in vivo CHI3L1 transcription is most commonly associated with astrocytes. Since we found elevated CHI3L1 levels in the CSF of a variety of neurological and neurodegenerative diseases, we wanted to study whether CHI3L1 expression is prevalent in astrocytes in those conditions. Combined ISH with GFAP staining was performed on sections from 4 cases of SIVE, 3 cases of SIV without encephalitis, AD and schizophrenia, 2 cases of Pick's disease and infarct and one case of MS and ALS. The ISH signal showed robust CHI3L1 transcription in SIVE and MS, focal ISH signal in chronic infarct cases while AD, ALS, Pick's disease and schizophrenia showed a few scattered positive cells in some areas while other areas were negative (Figure 4A). CHI3L1 ISH signal and immunohistochemistry was co-localized with GFAP staining in all the tested diseases (e.g. AD (Figure 4B) and brain infarction (Figure 4D). Confocal z-stack confirms that CHI3L1 (green) co-localizes with GFAP (red) staining as seen in yellow (Figure 4E & 4F). Pixel counts of CHI3L1 ISH signal correlated with pixel counts of GFAP staining (r² = 0.5637) indicative of a correlation between CHI3L1 transcription and reactive gliosis (Figure 4C).

In summary, CHI3L1 expression in MS, aging and other neurological diseases is mostly associated with astrocytes. There is a correlation between CHI3L1 transcription and GFAP staining which is indicative of reactive gliosis. CHI3L1 transcription was more pronounced in diseases with a more pronounced inflammatory nature like encephalitis and MS.

We have previously reported that CHI3L1 serves as a CSF biomarker for the development of encephalitis after SIV or HIV infection [11]. Our data showed a correlation between the time-course of CHI3L1 concentration and viral load in the CSF. In addition, we found a paradox that CHI3L1 is released from macrophages in vitro but in vivo CHI3L1 protein was in the soma of astrocytes in the vicinity of microglial nodules. In this study we show that CHI3L1 mRNA is expressed by reactive astrocytes surrounding microglial nodules suggesting that macrophages possibly through released inflammatory mediators, can induce CHI3L1 expression in surrounding astrocytes but not neurons. These findings suggest that CHI3L1 found in the CSF in SIVE/HIVE could be derived from reactive astrocytes. It has been shown by others that differentiated macrophages do express CHI3L1 in vitro and in vivo in peritumoral macrophages in small cell lung cancer, in atherosclerotic plaques or in an asthma model in mice [12, 18, 19]. Therefore, it is not clear why CHI3L1 transcription in the brain is less evident in activated tissue macrophages/microglia. We hypothesize that YKL40 transcription by macrophages is inhibited soon after they enter the brain and this may account for the differences seen in other tissue pathologies. We are attempting to further dissect this paradox with in vitro studies.

Previously, CHI3L1 expression was shown to be up-regulated in inflammatory conditions, but apart from one report of increased CHI3L1 mRNA in AD brain, there were no data in the literature characterizing the cellular source of CHI3L1 in neurodegenerative diseases. In this study we found high CHI3L1 transcription in two highly inflammatory diseases SIVE and MS. MS is a demyelinating disease associated with waxing and waning inflammation, gliosis and variable axon loss. CSF CHI3L1 levels were significantly elevated in MS patients, however, HIV-infected individual with high viral load showed higher CSF concentration consistent with the observation that neuroinflammation in encephalitis is more robust than in unselected MS. Comparing CSF concentration of CHI3L1 in patients at different stages of MS might be helpful in distinguishing active disease.

CHI3L1 transcript was also found in other neurodegenerative diseases like AD, ALS and Pick's disease as well as stroke, although to a lesser extent than SIVE and MS. In all cases, CHI3L1 was localized to astrocytes from which we would conclude that CHI3L1 expression is a general phenomena of reactive gliosis in response to neuroinflammation. Neuropathological astrocytosis not associated with neuroinflammation (e.g. spontaneous spongiform encephalopathy in rodents) is not associated with YKL40 expression (data not shown). This further emphasizes the use of CHI3L1 as a potential marker for CNS inflammatory conditions with its concentration reflecting a process of reactive gliosis. CSF samples from these diseases show increased CHI3L1 CSF levels compared to healthy young adults. Perhaps not surprisingly, older healthy adults have a very modest but significant elevation in CHI3L1 levels consistent with the hypothesis that low-grade inflammatory processes are induced in the aging brain.

Recent studies showed increased CHI3L1 mRNA in the hippocampus and prefrontal cortex of schizophrenic patients [13, 18]. Moreover, two genetic studies claimed that variants in the promoter of CHI3L1 are associated with susceptibility to schizophrenia [19, 20]. However, a more recent study failed to confirm a genetic association between CHI3L1 and schizophrenia in Japanese and Chinese populations [21]. Overall, our results showed only occasional CHI3L1 positive cells in schizophrenia. That the younger schizophrenic case (age 32) had almost no CHI3L1 positive astrocytes while the other two cases (ages 52 and 56) had more CHI3L1 positive astrocytes raises the question whether there is a relationship between CHI3L1 expression and schizophrenia or it is just age-related reactive gliosis.

Comparison of CHI3L1 transcription in acute, subacute and chronic infarcts showed that there is focal and temporal expression of CHI3L1 in astrocytes at the site of injury. Maximal CHI3L1 induction starts at the acute stages (3-5 days; ischemic changes in neurons visible, prominent endothelium and early neutrophilic infiltrate; no reactive astrocytes, no macrophages), continues at the subacute stages (between 5 days to 4 weeks; numerous foamy macrophages, some reactive astrocytes) and then diminished at the chronic stages (more than 4 weeks; few or no macrophages, abundant reactive astrocytes). These differences in YKL-40 transcription in different stages of pathology implies that acute inflammation induces CHI3L1 expression in astrocytes proximal to the lesion and as inflammation resolves CHI3L1 expression is diminished.