Chi3l3: a potential key orchestrator of eosinophil recruitment in meningitis induced by Angiostrongylus cantonensis

Soluble antigens of the 4th stage larvae of AC (AC L4) were collected from Sprague-Dawley rat brains at 21 dpi as previously described [25]. Bone marrow cells were isolated as described previously [26] and cultured with M-CSF (20 ng/mL) in complete medium [DMEM High Glucose, 100 U/ml of penicillin-streptomycin and 10% heat-inactivated (56 °C, 30 min) FBS] for 7 days. The following reagents were used in this study: Recombinant Murine M-CSF (315-02; Peprotech), Recombinant Mouse Chi3l3 (2446-CH-050, R&D Systems), and Recombinant Mouse IL-13 (413-ML-025/CF, R&D Systems).

For establishment of the airway allergy model, female C57 mice were chosen, and Alternaria alternata (Greerlabs, Lenoir, NC, 100 g/mouse in 50 μL) or PBS (50 μL) was administered intranasally on four consecutive days [27]. Four hours before the daily Alternaria alternata administration, 50 μg soluble rat-derived antigens of AC larvae (L4) or an equal volume of PBS treatment was administered via nasal drip. On the day after the last intranasal stimulation, mice were euthanized and BAL was harvested. Then, the cells derived from BAL were stained by Siglec F PE (BD), CD11c PE-Cyanine5 (eBioscience), or CD45 APC-cy7 (Biolegend) and analyzed by flow cytometry [28]. For the AC infection model, each Sprague-Dawley rat and BALB/c mouse were orally infected with 100 and 30 AC larvae (L3), respectively. AC larvae (L3) were collected from the tissue of infected Biomphalaria glabrata as previously described [29]. The mice and Sprague-Dawley rats were euthanized on days 7, 14, 21, and 28 after infection with AC.

Brain and lung hematoxylin and eosin staining and immunofluorescence were performed on 5-mm-thick, 4% paraformaldehyde-fixed, paraffin-embedded slices. Antibodies against mouse Ym1 (60130, StemCell Technologies), CD11b (MAB1124, R&D Systems), and Iba1 (NCNP24, Wako) were used as primary antibodies, with the corresponding secondary antibodies labeled with FITC or Alexa Fluor® 594. In addition, nuclei of cells were displayed by DAPI staining. All tissue slices were examined with laser-scanning confocal microscopy (Zeiss LSM780; Germany).

Flow cytometry analysis was performed on a CytoFLEX S flow cytometer (Beckman Coulter).

For blood and brain macrophage and eosinophil assessment, PBMCs and BMNCs were isolated as previously described [25] and then incubated with Siglec F PE (BD), CD11c PE-Cyanine5 (eBioscience), or CD11b APC-cy7 (eBioscience) at 4 °C for 30 min.

To determine the cell sources of Chi3l3, BMNCs were stained with CD11b PE (eBioscience), F4/80 PE-Cyanine5 (eBioscience), CD45 PE-eFluor 610 (eBioscience), Ly-6C PerCP/Cy5.5 (Biolegend), CX3CR1 Alexa Fluor 488 (eBioscience), or CCR2 Phycoerythrin (R&D) at 4 °C for 30 min.

For measurement of IL-5 and IL-13 levels, spleen cells [30] from normal and AC-infected mice were incubated with CD3 APC (eBioscience), CD4 FITC (eBioscience), and IL-13 PE cy7 (eBioscience) at 4 °C for 30 min and analyzed by flow cytometry.

To determine the possible effects of Chi3l3 and sAg on IL-13 production, spleen cells isolated from normal and AC-infected mice were cultured in 24-well cell culture plates for 72 h, followed by a 12-h incubation with Brefeldin A and 6-h stimulation with PMA and ionomycin. Then, the cells were incubated with CD3 APC (eBioscience), CD4 FITC (eBioscience), or IL-13 PE cy7 (eBioscience) at 4 °C for 30 min and analyzed by flow cytometry.

OCR (Oxygen Consumption Rate) and ECAR (Extracelluar Acidification Rate), as indicators of cellular oxidative phosphorylation and glycolysis, respectively, were monitored consecutively with a Seahorse Bioscience extracellular flux analyzer (XF24, Seahorse Bioscience) as described previously [31]. Approximately 15,000–25,000 cells were seeded in 24-well plates cultured with 500 μL complete medium and incubated for 24 h in a 37 °C incubator. Then, cells were immersed in 500 μL specified medium following two wash steps with specified medium and incubated in an incubator without CO₂ for 1 h before the measurements.

The basal levels of OCR were recorded thrice as were the OCR levels after sequential addition of 1 μM oligomycin, 1.0 μM FCCP, and 0.5 μM rotenone + 0.5 μM antimycin A. Similarly, ECAR was measured under basal conditions and after the injection of the following drugs: 10 mM glucose, 1.0 μM oligomycin, and 50 mM 2-deoxyglucose (2-DG).

Mouse and rat brains were dissected into the cerebrum and cerebellum and stored in TRIzol at − 80 °C. Total RNA was extracted using TRIzol (Life Technologies) reagent and reverse-transcribed to cDNA using a RevertAid™ FirstStrand cDNA Synthesis Kit (Thermo Fisher Scientific, USA) according to the manufacturer’s protocol. Specific gene expression was quantified with SYBR® Premix Ex Taq™ (Tli RNaseH Plus) (RR420A) using the Roche LightCycler® 480 real-time PCR platform.

The following amplification primers (Sangon Biotech) were used (5′ to 3′): mouse Chi3l3 (sense, CTGAATGAAGGAGCCACTGA; antisense, AGCCACTGAGCCTTCAACTT), mouse β-actin (sense, GGCATCCTGACCCTGAAGTA; antisense, CTCTCAGCTGTGGTGGTGAA), rat Chi3l3 (sense, AGTACCCTATGCCGTTCAGG; antisense, CAGACCATTGCACCTCCTAA), rat B2m (sense, GTCACCTGGGACCGAGAC; antisense, GAAGATGGTGTGCTCATTGC), and rat Hprt1 (sense, CTGTCATGTCGACCCTCAGT; antisense, GTCCATAATCAGTCCATGAGGA).

The cerebrums and cerebellums of mice and rats were lysed using RIPA Lysis Buffer (Strong) (Cwbiotech), followed by SDS-PAGE. After transfer to nitrocellulose membranes (GE Healthcare) using the Semi-Dry Transfer with Trans-Blot® SD Semi-Dry Transfer Cell (Bio-Rad), target proteins were verified with the corresponding antibodies. The following primary antibodies were used: mouse JMJD3 (ab85392, Abcam), CREB1 (A11989, ABclonal), CEBPB (A0711, ABclonal), KLF4 (A11853, ABclonal), Y641 phosphorylated-STAT6 (ab54461, Abcam), STAT6 (5397, Cell Signaling Technology), Chi3l3 (#60130, STEMCELL Technology), PPARγ (81B8, Cell Signaling Technology), and β-actin (12262, Cell Signaling Technology).

Our group generated RNA-seq data of AC infection in mouse brain tissue [32]. The raw data were processed using a standard pipeline as follows: we first filtered the low-quality tags and trimmed adaptors. Next, we applied TopHat to map the clean reads to the Mus musculus 9 genome; we then used cufflinks and Cuffdiff to calculate the expression levels of transcripts and to analyze the differential expression, respectively.

The functional protein association network for the genes in each cluster was constructed using the STRING database (https://​string-db.​org/). In this study, a medium confidence level (0.4) was selected to construct the network. We investigated the enriched functions of genes in the constructed network by performing functional enrichment using the DAVID webserver (https://​david.​ncifcrf.​gov/). The adjusted P values were examined to select the significantly enriched biological processes.

We built a mathematical model to describe the dynamics of expression levels of Chi3l3 and IL-13 using ODEs as follows:where [Chi3l3] and [IL13] represent the normalized expression levels of Chi3l3 and IL-13, respectively. a and b are the basic transcript rates of Chi3l3 and IL-13. V₁ is the maximal activation rates of Chi3l3 promoted by IL-13, and V ₂ is that of IL-13 promoted by Chi3l3. K₁ and K₂ are the Michealis constants. n₁ and n₂ are Hill coefficients. d₁ and d₂ are degradation rates of Chi3l3 and IL-13, respectively.

The above ODEs were numerically solved using the 4th order Runge-Kutta method. The parameters in the above model was estimated using the non-linear least square method by fitting the model simulation to the experimental data. The experimental data were re-sampled using the 3rd order Hermite polynomial interpolation scheme. The genetic algorithm [33, 34] was employed to optimize the objective function. The estimated parameter values are as follows: V₁ = 0.2588, V₂ = 0.9686, K₁ = 0.6980, K₂ = 0.5216, n₁ = 2, n₁ = 1, d₁ = 0.1216, d₂ = 0.4824, a = 0.0039, and b = 0.0235. The programming was performed in MATLAB R2007b (MathWorks, USA). The bifurcation analysis was performed using Oscill8 (http://​oscill8.​sourceforge.​net).

AC induced serious eosinophilic meningitis or meningoencephalitis when mice were infected with the third-stage larvae (L3) of AC. To elucidate the underlying mechanism of eosinophil infiltration in the brain, we analyzed the dynamic gene expression profiles of mouse brains using RNA-seq data [32]. We designed an algorithm for selecting significant temporally changing genes (TCGs), and 336 genes were identified as TCGs. The expression profiles of these TCGs are shown in a heatmap (Fig. 1a). TCGs were then clustered into six groups (Fig. 1b); each group had a different dynamic expression pattern. The cluster 1 genes showed a sustained increasing expression pattern after AC infection, indicating possible important functions. Therefore, we then analyzed the dynamic properties of cluster 1 genes by calculating their changing rates as defined in Eq. (1) in the “Methods” section. The maximal values of the changing rates of these genes at all time points were calculated and ranked from the largest to the smallest. We confirmed LOC547349 and Chi3l3 as the top two genes (Table 1), and Chi3l3 showed higher expression than LOC547349 (Additional file 1: Figure S1). Therefore, we hypothesized that Chi3l3 was an important molecule in AC infection. We further investigated the differences in Chi3l3 expression between non-permissive hosts (mice) and permissive hosts (rats) infected with AC and were surprised to find that the mRNA and protein levels of Chi3l3 in the brain varied greatly between mice and rats during infection (Fig. 1c, d). Mouse brain Chi3l3 mRNA levels increased from 1000 at 14 dpi to 10,000 at 21 dpi (Fig. 1d), and the protein level experienced a fold increase from 14 dpi to 21 dpi (Fig. 1e). Notably, AC infection time-dependently upregulated chitinases and chitinase-like proteins (CLPs) at 0, 2, 7, 14, and 21 dpi, with Chi3l3 being the most significant one (Fig. 1f), in accordance with our gene microarray data [35].

Collectively, these results indicated that a sharp increase of Chi3l3 was observed in non-permissive host mice but not in permissive host rats during AC infection, and thus, Chi3l3 is likely to be an important element closely correlated with eosinophilic meningitis [24, 25, 35].

Excess Chi3l3 in the brain prompted us to investigate the possible cell source of this molecule. Flow cytometric analysis revealed that approximately 8% of brain mononuclear cells (BMNCs) of mice infected with AC at 21 dpi were Chi3l3-positive (Chi3l3⁺) cells. Further analysis showed that the majority (81.8%) of Chi3l3⁺ cells possess a CD45^hiF4/80⁺CD11b⁺ phenotype (Fig. 2a). Monocytes/macrophages are currently divided into two classes, “inflammatory” type (CCR2⁺Ly-6C^hiCD62L⁺CX3CR1^lo) and “resident” type (CCR2⁻Ly-6C^loCD62L⁻CX3CR1^hi) cells [36]. The above observation suggested that bone marrow-derived myeloid cells may be recruited to the inflamed CNS during AC infection, similar to other brain inflammatory diseases, such as multiple sclerosis and experimental autoimmune encephalomyelitis [37]. To test this hypothesis, we assessed CCR2, Ly-6C, and CX3CR1 as previously described [36]. Chi3l3⁺ BMNCs are primarily characterized by Ly-6C⁺CCR2⁺CX3CR1^lo and a low side scatter profile (Fig. 2b), which indicate an “inflammatory” subset of monocytes/macrophages, not a “resident” subset [36]. In addition, kinetics of CD45⁺ F4/80⁺ cell infiltration in the brain was evaluated; Chi3l3⁺ cells were first observed at 14 dpi, and the cell percentage increased sharply with eosinophilic meningitis progression (Fig. 2e), accompanied by enhanced expansion of CD45^hiF4/80⁺ cells compared to CD45^loF4/80⁺ cells (Fig. 2d). Further evaluation of the macrophage-specific markers Iba1 [38] and CD11b by immunofluorescence revealed that infiltrating cells were differentiated mature macrophages after AC infection (Fig. 2c).

Intriguingly, we found that the eosinophil percentage synchronized with Chi3l3 and IL-13 levels in brains of AC-infected mice, with a percentage of approximately 2% at 7 dpi. The eosinophil percentage rose sharply over time, exceeded 20% at 14 dpi, and finally reached 40% at 21 dpi (Fig. 2e). In contrast to the CNS symptoms of AC-infected mice, we did not observe obvious changes in the eosinophil percentage in the blood of AC-infected mice at 14 dpi and 21 dpi compared with the uninfected healthy mouse group, although there was a slight increase in eosinophil percentage at 7 dpi (P < 0.05) (Fig. 2f).

To define the potential relationship between macrophage infiltration and eosinophilic meningitis in AC infection, we administered liposomal clodronate [39] prior to and during the infection to effectively deplete macrophages (Fig. 2g). As expected, treatment with liposomal clodronate (Fig. 2g) blocked the accumulation of macrophages in the brain after AC infection, decreased Chi3l3 (Fig. 2h), and relieved CNS pathology (Fig. 2i) at 22 dpi.

A previous study showed that cellular infiltration around dead worms was more prominent than that around living ones [1], and these findings prompted us to investigate the function of sAg in eosinophil recruitment. Alternaria extract induces a pulmonary allergic reaction in mice, thus providing an ideal model for studies concerning the effect of helminths and their products on innate and adaptive immune responses [40, 41]. After 4 days of in vivo exposure to Alternaria and sAg, we observed an increase in eosinophils of bronchoalveolar lavage (BAL) fluid in C57BL/6 mice from the sAg administration group 2 h after Alternaria treatment (Fig. 3a, b).

These results demonstrate that the sAg stimulus may directly or indirectly affect the brain microenvironment during AC-induced eosinophilic meningitis.

To investigate the possible interactions of Chi3l3 with other genes, we constructed a functional interaction network of Chi3l3. We first calculated the Pearson correlation coefficients of Chi3l3 with all the other genes. The genes that were highly correlated with Chi3l3 (correlation coefficient greater than 0.8) were identified and input into the STRING database (https://​string-db.​org/). In this study, a medium confidence level (0.4) was selected to construct a functional gene association network (Fig. 4a). Then, a sub-network containing all genes interacting with Chi3l3 was extracted (Fig. 4b). IL-4, IL-13, and CCL2 were predicted to be functionally associated with Chi3l3.

The functional enrichment of genes in the Chi3l3 network was performed as shown in Fig. 4c. The significantly enriched biological processes include response to organic substance, response to stress, immune system process, response to chemical, regulation of response to stimulus, single-multicellular organism process, positive regulation of biological process, multicellular organismal process, regulation of localization, and regulation of immune system process.

Both RNA-seq data and qPCR measurements from in vivo experiments (Fig. 4d, e) indicated that Chi3l3 and IL-13 showed a high positive correlation and burst expression pattern. Therefore, we hypothesized that a positive feedback loop might exist between Chi3l3 and IL-13. To test this hypothesis, we developed an ordinary differential equation (ODE) model describing the positive feedback between Chi3l3 and IL-13 (Eqs. (2–3) in the “Methods” section] to quantitatively study the dynamics of Chi3l3 and IL-13 expression. We fitted the mode to the in vivo experimental data (Fig. 4d, e). A good agreement between the theoretical and experimental results (the mean squared error is 0.0849) supported the hypothesized positive feedback loop between Chi3l3 and IL-13. Moreover, bifurcation analysis of the model revealed a bistable switch of Chi3l3 and IL-13 with respect to the parameter K₂ (Fig. 4f). As the value of K₂ increased above ~ 2.63, the steady states of Chi3l3 and IL-13 then switched from high levels to low levels. The bistability of this system indicated the state transition of infection progression and heterogeneity of infection outcome between individuals. The switches induced by the increase in the value of K₂ suggested that therapeutically inhibiting the IL-13 signaling pathway, including the IL-13 receptor, might ameliorate the AC infection.

Flow cytometry analysis was carried out, which indicated IL-13 and IL-5 expression of CD3⁺CD4⁺ lymphocytes (Fig. 5a, b) in spleen, the most important peripheral immune organ. We observed a significant increase from 0.4 to 0.6% in IL-13⁺ expression of spleen-derived CD3⁺CD4⁺ cells, but not IL-5, which is consistent with the qPCR results of brains (Fig. 4e) (Additional file 2: Figure S2e).

To clarify whether Chi3l3 acts as a positive regulator of IL-13 cytokine production in our model, we cultured splenocytes of normal or AC-infected mice for 72 h in the presence of sAg or Chi3l3 followed by a 6-h PMA and ionomycin stimulation to increase cytokine production (Fig. 5c, d). We found that Chi3l3 enhanced spleen CD3⁺CD4⁺ T cell IL-13 secretion, with a percentage increase from 0.8 to 1.5% (Fig. 5c) and 0.8 to 1.9% (Fig. 5d) in normal mice and AC-infected mouse spleens, respectively.

Consistent with our previous study using RAW 264.7 cells [25], we found that BMDM Chi3l3 initiated by IL-13 was also promoted by sAg (Additional file 2: Figure S2a) (Fig. 5e). Thus, we hypothesized that uncontrolled Chi3l3 is achieved by an interlaced net of signaling pathways and transcription factors vital to the alternative activation of macrophages, including STAT6 phosphorylation [42], Histone H3 modification [43], PPARγ [44], C/EBPβ [45], and KLF4 [46] activation. We found that only BMDM PPARγ expression (Additional file 2: Figure S2b) (Fig. 5e) increased when stimulated with IL-13 and sAg plus IL-13, and sAg and IL-13 co-stimulation further enhanced Chi3l3. These data indicate that PPARγ is a strong activator of BMDM Chi3l3 expression in the sAg and IL-13 co-stimulation model. PPARγ is a known director of the oxidative phosphorylation process, which regulates M2 activation by controlling gene expression critical for metabolic reprogramming [44]. To gain insight into the metabolism of BMDMs in modulation of Chi3l3 expression, we measured the basal OCR, max OCR, and SRC OCR (spare respiratory capacity; the difference between basal OCR and max OCR) [31] level of BMDMs (Fig. 5g–i). Consistent with the increase in PPARγ, elevated OCR indicated that BMDMs displayed an enhanced oxidative phosphorylation metabolic phenotype (Fig. 5f).

Based on the results above, we verified the skewed cytokine pattern in brains of AC-infected mice, especially Chi3l3, which may be due to the positive feedback loop between IL-13 and Chi3l3 of inflammatory macrophages. This may be mediated by activating the PPARγ signaling pathway, which is associated with oxidative phosphorylation.