Elevated antibody binding to striatal cholinergic interneurons in patients with Pediatric Acute-onset Neuropsychiatric Syndrome

Abstract

Pediatric Acute-onset Neuropsychiatric Syndrome (PANS) is characterized by the abrupt onset of significant obsessive-compulsive symptoms (OCS) and/or severe food restriction, together with other neuropsychiatric manifestations. An autoimmune pathogenesis triggered by infection has been proposed for at least a subset of PANS. The older diagnosis of Pediatric Autoimmune Neuropsychiatric Disorder Associated with Streptococcus (PANDAS) describes rapid onset of OCD and/or tics associated with infection with Group A Streptococcus. The pathophysiology of PANS and PANDAS remains incompletely understood. We recently found serum antibodies from children with rigorously defined PANDAS to selectively bind to cholinergic interneurons (CINs) in the striatum. Here we examine this binding in children with relapsing and remitting PANS, a more heterogeneous condition, collected in a distinct clinical context from those examined in our previous work, from children with a clinical history of Streptococcus infection. IgG from PANS cases showed elevated binding to striatal CINs in both mouse and human brain. Patient plasma collected during symptom flare decreased a molecular marker of CIN activity, phospho-riboprotein S6, in ex vivo brain slices; control plasma did not. Neither elevated antibody binding to CINs nor diminished CIN activity was seen with plasma collected from the same children during remission. These findings replicate what we have seen previously in PANDAS and support the hypothesis that at least a subset of PANS cases have a neuroimmune pathogenesis. Given the critical role of CINs in modulating basal ganglia function, these findings confirm striatal CINs as a locus of interest in the pathophysiology of both PANS and PANDAS.

1. Introduction

Some children develop obsessive-compulsive disorder (OCD) and other neuropsychiatric symptoms extremely rapidly, even overnight. This has been described as ‘pediatric acute-onset neuropsychiatric syndrome’, or PANS (Chang et al., 2015; Murphy et al., 2015; Singer et al., 2012; Swedo et al., 2012). Clinical manifestations of PANS include abrupt onset of significant OCS and/or severely restricted food intake, together with other symptoms such as emotional lability, anxiety, motor abnormalities, cognitive regression, sleep disruption, and urinary urgency. Symptoms typically follow a relapsing and remitting course (Chang et al., 2015; Frankovich et al., 2015a; Murphy et al., 2015; Swedo et al., 2012; Toufexis et al., 2015). PANS is often associated with evidence of immune dysregulation, and it has been proposed that many cases derive from an autoimmune process triggered by infection. While the diagnosis does not require temporal association with a documented infection, it has been associated with various pathogens, including Streptococcus and mycoplasma (Allen et al., 1995; Ercan et al., 2008; Frankovich et al., 2015b; Gerentes et al., 2019; Rhee and Cameron, 2012).

The association between pediatric infection and rapid onset of neuropsychiatric symptoms was first described in conjunction with Streptococcus. A recent population-based study reported that children positive for Streptococcus are at increased risk of developing OCD (Orlovska et al., 2017). Rapid onset of OCD and/or tics in a child with evidence of Streptococcal infection has been named Pediatric Autoimmune Neuropsychiatric Disorder Associated with Streptococcus, or PANDAS (Swedo, 2002; Swedo et al., 1998). An autoimmune pathophysiology based on molecular mimicry has been hypothesized, by analogy to the well-established postinfectious neuropsychiatric condition Sydenham chorea (Swedo and Williams, 2017; Williams and Swedo, 2015). Several autoantibodies to Streptococcal antigens that cross-react with various basal ganglia epitopes have been reported (Ben-Pazi et al., 2013; Dale et al., 2012; Kirvan et al., 2007; Kirvan et al., 2003; Kirvan et al., 2006a). Immunomodulatory therapies have been recommended for for PANS and PANDAS, though controlled trials have been small and have had mixed results (as is the case with SC) (Allen et al., 1995; Frankovich et al., 2017; Garvey et al., 2005; Hajjari et al., 2022; Johnson et al., 2021; Perlmutter et al., 1999; Sigra et al., 2018; Williams et al., 2016).

PANS and PANDAS overlap phenomenologically with pediatric obsessive-compulsive disorder (OCD) and Tourette syndrome (TS); comorbidity is common (Calaprice et al., 2017; Gromark et al., 2019; Murphy et al., 2015; O’Dor et al., 2022). The symptoms of PANS and PANDAS overlap in that OCD symptoms are typically prominent in both, though PANDAS has stricter requirements for clearly documented temporal association with a Streptococcal infection and an episodic course. Cases where the core symptom is eating restriction (PANS) or tics (PANDAS), in the absence of prominent OCD symptoms, may look quite distinct. Inflammation and microstructural differences have been described in the basal ganglia during the acute phase of both PANDAS and PANS (Giedd et al., 2000; Giedd et al., 1996; Kumar et al., 2015; Zheng et al., 2020).

Numerous studies over the past 25 years have sought to characterize the causes of PANDAS, and in particular to identify the hypothesized pathogenic autoantibodies; however, the pathophysiology remains incompletely understood. Binding of PANDAS serum antibodies to basal ganglia tissue has been reported (Frick et al., 2013; Morer et al., 2008; Pavone et al., 2004; Singer et al., 2004; Xu et al., 2021). Several autoantibody targets have also been described, including lysoganglioside GM1, tubulin, and dopamine D1 and D2 receptors (Ben-Pazi et al., 2013; Chain et al., 2020; Cox et al., 2013; Dale et al., 2012; Kirvan et al., 2003; Kirvan et al., 2006b; Singer et al., 2015). However, these findings have not been consistently replicated (Brilot et al., 2011; Morris et al., 2009; Singer et al., 2005; Xu et al., 2021). Few studies have examined the pathophysiology of PANS.

In recent work (Frick et al., 2018; Xu et al., 2021), we focused on the binding of PANDAS-associated antibodies to specific cell types in the basal ganglia; this contrasts with earlier work examining candidate molecular targets. We have described, and replicated, elevated binding of IgG from children with strictly defined PANDAS to striatal cholinergic interneurons (CINs), relative to age- and gender-matched healthy controls. This antibody binding leads to reduced CIN activity in vitro; it is reduced, in parallel with symptom improvement, after IVIG treatment (Frick et al., 2018; Xu et al., 2021).

It is unclear whether this finding extends to PANS, a more heterogeneous diagnostic entity. Here, we addressed this question in a cohort of patients with relapsing/remitting PANS and clinical evidence of Streptococcal infection, treated at the Immune Behavioral Health Clinic at Stanford University. We confirm elevated binding to striatal CINs by IgG in plasma drawn during symptom flare, compared to plasma drawn from matched control subjects. PANS IgG drawn at flare reduces activity of striatal CINs; neither elevated IgG binding to CINs nor the inhibition of CIN activity are seen in plasma drawn from the same children during symptom recovery/remission. These data suggest that IgG binding to striatal CINs contributes to the pathophysiology of PANS.

2. Materials and methods

2.1. Specimen collection

These investigations of human plasma samples were approved by the Human Investigations Committees of Yale and Stanford Universities. Plasma was collected at the Stanford Immune Behavioral Health Clinic and Research Program, from local patients and controls (living in the 7 counties surrounding Stanford University). Patients were classified as meeting PANS criteria, with a relapsing and remitting course, by a child psychiatrist (MT or MS). Patients were selected who had a positive test for Streptococcus, though many did not meet full diagnostic criteria for PANDAS. Parents gave written informed consent to participate in the study, and competent subjects gave assent prior to blood collection. Serum samples were collected during symptom flare; in a subset of subjects, samples were also collected during a period of symptom recovery/remission, either before or after the flare (time between symptom onset and remission/recovery draws: 4.15 ± 0.85 years). Mean IgG titers were 956.53 ± 76.59 mg/dL in the healthy control subjects and 756.75 ± 46.63 mg/dL in the flare subjects. In the subset of subjects who had recovery draws, IgG titer were 717.27 ± 55.35 mg/dL and 756.61 ± 91.08 mg/dL during flare and recovery, respectively. All samples were aliquoted prior to storage at the Stanford Biobank, re-aliquoted into smaller volumes upon arrival at Yale, and stored at −80 °C until use. Demographic data for all subjects are shown in Supplementary Table 1 and Table 1. All samples were anonymized before being sent from Stanford to Yale for analysis; all analyses were performed blind to diagnosis and condition. Samples/disease status classification was based on the data collected by the Clinician Encounter Form which clinicians/research staff complete at the end of the clinic visit after reviewing patient/parent questionnaires, psychometrics, and visit notes.

Table 1. Cohort Characteristics - Patients and Healthy Controls.

Demographic and Clinical Characteristics of 25 Patients with Pediatric Acute-onset Neuropsychiatric Syndrome and 25 Healthy Controls.

	Patients (N = 25)	Healthy Controls (N = 25)
Characteristic	No. (%)	No. (%)
Age of First neuropsychiatric symptom onset, mean (SD), years	8.4 (2.9)
Age of meeting PANS criteria	9.9 (3.8)
Age of first IBH PANS clinic visit	11.2 (3.6)
Age at “flare”/Control blood draw chosen for this study	12.2 (4.0)	12.2 (4.2)
Age at “recovery” blood draw chosen for this study	12.1 (4.4)
Biological Male Sex	12 (48.0)
Race and ethnicity, n (%)
Non-Hispanic White	19 (76.0)	11 (44.0)
Other	6 (24.0)	14 (56.0)
PANS symptoms at clinic presentation, n (%)
OCD	25 (100.0)
Eating restriction	15 (60.0)
Anxiety	25 (100.0)
Emotional lability and/or depression	19 (76.0)
Irritability, aggression, and/or severely oppositional behaviors	22 (88.0)
Behavioral/developmental regression	10 (40.0)
Deterioration in school performance	11(44.0)
Sensory dysregulationa	16 (64.0)
Motor abnormalitiesb	12 (48.0)
Sleep disturbance	19 (76.0)
Increased urinary frequency and/or enuresis	10 (40.0)
Homicidal or suicidal ideation/attempts	9 (36.0)
Global Impairment from psychiatric symptoms, mean (SD)c, d	52.8 (18.9)	0.0 (0.2)
Caregiver Burden Inventory, mean (SD)c, e	39.1(14.8)	4.2 (6.5)

^aHyperacusis, photophobia, pain amplification, inability to feel pain, etc.

^bTics, motoric hyperactivity, chorea, etc.

^cAs these measures vary with the relapsing-remitting course of PANS, the score recorded during the visit closest to the blood draw was used for analysis.

^dGlobal impairment score is a parent-rated score of global functioning validated for the PANS population. The score ranges from 0 to 100 (the higher the worse).

^eCaregiver burden inventory is a parent-rated score of caregiver burden, and ranges from 0 to 96 (the higher the worse). A cut-off point of 36 represents a need for respite care. It has been validated in the PANS population.

ADHD, attention-deficit/hyperactivity disorder; IQR, interquartile range; OCD, obsessive compulsive disorder; PANS, Pediatric Acute-onset Neuropsychiatric Syndrome; SD, standard deviation.

2.2. Mouse and human brain tissues

All experimental procedures were approved by the Yale University Institutional Animal Care and Use Committee, in accordance with the NIH Guide for the Care and Use of Laboratory Animals. Male and female C57BL/6J mice (adults 3–6 months old and juveniles 4–5 weeks old) were purchased from the Jackson Laboratory (Bar Harbor, Maine, http://jaxmice.jax.org/strain/013636.html). Mice were anesthetized by intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg) and transcardially perfused with cold 4% paraformaldehyde in 1×PBS (pH 7.4). Brains were fixed overnight in 4% PFA at 4 °C, followed by equilibration in 30% sucrose for 48 h at 4°C. Striatal slices were cut at 20 μm using a Leica CM3050S cryostat (Leica, Buffalo Grove, IL). Slices were stored in a cryoprotectant solution (30% glycerin, 30% ethylene glycol in 1×PBS pH 7.4) at −20°C until use.

Slices of human basal ganglia were collected and prepared as part of an unrelated study, as described previously (Kataoka et al., 2010). Subject demographics are shown in Supplementary Table 2.

2.3. Reagents and antibodies

Ketamine (Ketaset) was obtained from Zoetis (Madison, NJ). Xylazine (Anased) was obtained from Akorn, Inc. (Decatur, IL). Reagents for immunohistochemical staining (Fig. 1–3) were obtained from Sigma (St. Louis, MO: Sudan Black B, Triton X-100, sodium fluoride) or from JT Baker (Phillipsburg, NJ: paraformaldehyde). Normal donkey serum in blocking buffer was obtained from Jackson Immunoresearch (West Grove, PA). Chemicals used for acute brain slice analysis (Fig. 4) were obtained from Sigma (N-methyl-D-glucamine (NMDG), glucose, thiourea, sodium pyruvate, sodium ascorbate and HEPES) or JT Baker (KCl, NaCl, NaH₂PO4, NaHCO₃, MgSO₄ and CaCl₂). Antibodies used in immunohistochemical staining are listed in Supplementary Table 3.

Fig. 1. Elevated binding of PANS IgG to striatal cholinergic interneurons in adult mouse brains.

A. Representative images of immunohistochemical staining for human IgG (green) and choline acetyltransferase (ChAT, red) after incubation of mouse brain slices with diluted PANS flare and control plasmas. Arrowheads indicate human IgG binding to ChAT-positive interneurons in the mouse striatum. B. IgG from PANS “flare” plasmas showed elevated binding to ChAT interneurons relative to IgG from control plasmas (2-tailed independent sample t-test: t[48] = 3.263, p = 0.002; Cohen’s d = 0.922). C. Representative images of immunohistochemical staining of human IgG (green) and parvalbumin (PV, red). Arrowheads indicate human IgG binding to PV-positive interneurons. D. Plasma IgG binding to PV interneurons did not significantly differ between PANS “flare” and control groups (2-tailed independent sample t-test: t[48] = 1.915, p = 0.061). E. Representative images of immunohistochemical staining of human IgG (green) and choline acetyltransferase (ChAT, red) in the medial septal nuclei. Arrowheads indicate human IgG binding to ChAT-positive neurons. F. Plasma IgG binding to septal ChAT neurons did not significantly differ between PANS “flare” and controls (2-tailed independent sample t-test: t[48] = 1.627, p = 0.110). Data are shown as mean ± S.E.M. Each data point represents the mean value obtained from 6 mice treated with a single plasma. **p < 0.01; Numbers in parentheses indicate sample size.. Scale bars: 20 μm.

Fig. 3. PANS IgG shows elevated binding to cholinergic interneurons in human brain slices.

A, C. Representative images of immunohistochemical staining of human IgG (green) and choline acetyltransferase (ChAT, red) in human caudate (A) and putamen (C). Arrowheads indicate human IgG binding to CINs. See Figs S6 and S7 for additional images. B, D. PANS “flare” IgG showed elevated binding to CINs in human caudate and putamen. In caudate (B), a significant main effect was found (one-way ANOVA: F(3, 20) = 32.16, p < 0.0001, η² = 0.828). Tukey’s post hoc tests revealed elevated IgG binding in the PANS “flare” group relative to the control group (Control vs Flare, p < 0.0001). This elevation was reversed in the “recovered” group (Flare vs Recovered, p = 0.0002). No difference was found between control and “recovered” samples (Control vs Recovered, p = 0.930). All three group showed higher IgG fluorescence intensity relative to PANS “flare” plasmas from which IgG was depleted (p < 0.01). In putamen (D), a significant main effect was found (one-way ANOVA: F(3, 20) = 14.26, p < 0.0001, η² = 0.681). Tukey’s post hoc tests revealed elevated IgG binding in the PANS “flare” group relative to the control group (Control vs Flare, p = 0.037). This elevation was lost in the “recovered” group (Flare vs Recovered, p = 0.010). No difference was found between the control and “recovered” groups (Control vs Recovered, p = 0.930). All three group showed higher IgG fluorescence intensity relative to PANS “flare” plasmas from which IgG was depleted (p < 0.05). E, G. Representative confocal images of immunohistochemical staining of human IgG (green) and parvalbumin (PV, red) in human caudate (E) and putamen (G). Arrowheads indicate human IgG binding to PV-positive interneurons. F, H. PANS “flare” IgG did not show elevated binding to PV-positive interneurons. In caudate (F), a significant main effect was found (one-way ANOVA: F(3, 20) = 11.42, p = 0.0001, η² = 0.631). IgG binding in the PANS “flare” group did not differ from that of the control group (Control vs Flare, p = 0.987) or the “recovered” group (Flare vs Recovered, p = 0.445). No difference was found between the control and “recovered” groups (Control vs Recovered, p = 0.278). All three group showed higher IgG fluorescence intensity relative to PANS “flare” plasmas from which IgG was depleted (Control vs IgG-dep, p = 0.0002; Flare vs IgG-dep, p = 0.0005; Recovered vs IgG-dep, p = 0.015). In putamen (H), a significant main effect was found (F(3, 20) = 4.440, p = 0.0151, η² = 0.400). IgG binding in the PANS “flare” group did not differ from that of the control group (Control vs Flare, p = 0.975), or the “recovered” group (Flare vs Recovered, p = 0.990). No difference was found between the control and “recovered” groups (Control vs Recovered, p = 0.999). All three group showed higher IgG fluorescence intensity relative to PANS “flare” plasmas from which IgG was depleted (Control vs IgG-dep, p = 0.052; Flare vs IgG-dep, p = 0.022; Recovered vs IgG-dep, p = 0.041). Data are shown as mean ± S.E.M. Each data point represents mean value obtained from 2 post-mortem human brains for each plasma. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ^#p = 0.052; Numbers in parentheses indicate sample size. Scale bars: 20 μm.

Fig. 4. PANS plasma reduces spontaneous activity in striatal cholinergic interneurons.

Acute mouse striatal slices were treated with plasma from control subjects (Con), PANS “flare” (Flare), PANS “recovered” (Recov.), and PANS flare plasmas from which IgG was depleted (IgG-dep.). A. Representative images of immunohistochemical staining for phospho-rpS6 (p-rpS6, green) and choline acetyltransferase (ChAT, red). Arrowheads indicate p-rpS6/ChAT co-labeled ChAT interneurons. B. A significant main effect was found (F[3, 20] = 5.673, p = 0.0056, η² = 0.460). Control plasma did not significantly alter p-rpS6 relative to vehicle (one-sample t-test: t[5] = 1.879, p = 0.119). PANS “flare” plasma reduced phosphorylation levels of rpS6 in striatal CINs compared to control, at trend level (Con vs Flare, p = 0.079). This reduction was not seen in “recovered” plasma from the same subjects (Flare vs Recov, p = 0.046), or after IgG depletion in PANS “flare” plasma (Flare vs IgG-dep, p = 0.004). p-rpS6 levels did not differ between the control, “recovered”, or IgG-depleted groups (Con vs Recov, p = 0.993; Con vs IgG-dep, p = 0.500; Recov vs IgG-dep, p = 0.664). C. p-rpS6 levels correlated negatively with plasma IgG binding to striatal CINs (shown in Fig. 1A,B and Fig. 2A,B) after plasma incubation (R² = 0.547, p = 0.004). D. p-rpS6 levels in CINs correlated negatively with CY-BOCS scores after plasma incubation (R² = 0.355, p = 0.045). E. For each PANS “flare”-“recovered” pair, the difference in IgG binding to striatal CINs (Fig. 2A,B) correlated negatively with difference in p-rpS6 levels after plasma incubation (R²=0.767, p = 0.022). F. Representative images of immunohistochemical staining of phospho-rpS6 (p-rpS6, green) and parvalbumin (PV, red) in striatal slices after plasma treatment. Arrowheads indicate p-rpS6/PV co-staining. G. p-rpS6 levels in striatal PV interneurons remained unchanged after plasma incubation (one-way ANOVA: F(3, 20) = 0.306, p = 0.821). H. p-rpS6 levels did not correlate with plasma IgG binding to PV interneurons (shown in Fig. 1C, D and Fig. 2C, D) after plasma incubation (R² = 0.029, p = 0.499). I. p-rpS6 levels in PV interneurons did not correlate with CY-BOCS scores after plasma incubation (R² = 0.072, p = 0.393). J. For each PANS “flare”-“recovered” pair, the difference in IgG binding to striatal PV interneurons did not correlate with change in p-rpS6 levels after plasma incubation (R² = 0.101, p = 0.314). Data are shown as mean ± S.E.M. Each data point represents mean value obtained from slices from 5 mice treated with a single plasma. Pairwise comparisons were performed using Tukey’s post hoc test. *p < 0.05, **p < 0.01, ^#p = 0.079; Numbers in parentheses indicate sample size. Correlation was determined by linear regression analysis using Pearson’s correlation (C, E, H and J) or Spearman’s correlation (D and I). Dotted lines in correlation analyses (C-E and H-J) indicate 95% confidence intervals. Scale bars: 20 μm.

2.4. Determination of IgG titers in plasma samples

Total IgG titers in plasma samples were determined using IgG (Total) Human ELISA Kit (Thermo Fisher Scientific, Rockland, IL) following manufacturer’s instructions and as described (Xu et al., 2021). More details are described in the Supplementary Information. Total IgG titers were higher in control samples than in PANS samples; there were no differences between genders or between PANS “flare” and “recovery” samples. All plasmas were diluted in 1×PBS + 0.1% BSA (bovine serum albumin) to 500 mg/dL IgG and randomized before testing. An equal amount of diluted plasma, and thus of IgG, was used in all assays.

2.5. Assessing phospho-rpS6 levels in acute mouse brain slices

Acute coronal mouse brain slices containing the striatum were prepared from male and female C57BL/6J mice as previously described (Ting et al., 2014; Xu et al., 2021). Briefly, mouse brains were quickly removed and placed in ice-cold oxygenated NMDG-aCSF (artificial cerebrospinal fluid, in mM: 92 NMDG, 2.5 KCl, 1.25 NaH₂PO₄, 30 NaHCO₃, 20 HEPES, 10 MgSO₄, 0.5 CaCl₂, 25 glucose, 2 thiourea, 3 sodium pyruvate and 5 sodium ascorbate, pH 7.35, saturated with 95% O₂/5% CO₂). Coronal slices (100 μm) through the striatum were cut using a Leica VT1000S vibratome (Leica Microsystems, Bannockburn, IL, USA) in the NMDG-aCSF solution. Slices were recovered in NMDG-aCSF for 10 min at 32°C before being transferred to regular aCSF (in mM: 119 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 24 NaHCO₃, 2 CaCl₂, 2 MgSO₄ and 12.5 glucose) for 1 h at 30 °C under constant oxygenation with 95% O₂/5% CO₂. After recovery, slices were treated with plasma samples (6.25 mg/dL diluted in aCSF) or the aCSF control for 1 h at 30°C. After treatment, slices were fixed in cold 4% paraformaldehyde (PFA) in 1×PBS (pH 7.4) containing 5 mM NaF for 1 h at 4 °C, followed by immunohistochemistry.

2.6. Immunohistochemistry and image quantification

Paraformaldehyde-fixed mouse striatal slices (20 μm coronal sections) and formalin-fixed human brain slices (50 μm coronal sections of caudate and putamen) were used to examine plasma IgG deposition, as described (Xu et al., 2021). Brain sections were blocked in freshly prepared 0.1% Sudan Black B (in 70% ethanol) for 10 min at RT to reduce autofluorescence (Baschong et al., 2001). After washes, sections were incubated in blocking buffer (1×PBS + 0.3% Triton X-100 supplemented with 5% normal donkey serum) for 1h at RT, and then incubated with plasma (1.25 mg/dL each in blocking buffer) overnight at 4°C. The next day, sections were washed in blocking buffer and incubated with an anti-human IgG antibody and specific neuronal markers (ChAT and PV, Supplementary Table 3) overnight at 4 °C, followed by incubation with fluorophore-conjugated secondary antibodies (Supplementary Table 3) for 1 h at RT. After washes, sections were mounted in Vectashield HardSet Mounting Medium (Vector Laboratories), coverslipped, and stored at 4 °C.

For plasma IgG binding to mouse neurons (Figs. 1 and 2), each plasma was tested in brain sections from 6 mice (both male and female), with 4–6 images collected randomly from the dorsal striatum of each mouse, without overlap, using an Axio Scope A1 fluorescent microscope with a 10×/0.45 NA objective (Zeiss, Germany). For plasma IgG binding to human brain slices (Fig. 3), each plasma was tested in brain sections from 2 subjects, with 6 images collected, without overlap, in each section. Images were captured by sequential scanning of sections on an Olympus Fluoview FV-1000 confocal microscope with a 20×/0.85NA objective (Olympus, Japan).

Fig. 2. IgG binding to cholinergic interneurons in adult mouse brain is reduced in plasmas drawn from PANS subjects during symptom recovery.

A. Representative images of immunohistochemical staining of human IgG (green) and choline acetyltransferase (ChAT, red) in the mouse striatum. Arrowheads indicate human IgG binding to ChAT-positive interneurons. B. IgG binding to ChAT interneurons was decreased during symptom recovery, relative to plasma collected during symptom flare (paired t-test: t[12] = 4.534, p = 0.0007; Cohen’s d = 1.258). C. IgG binding to striatal ChAT interneurons correlated with CY-BOCS scores (Spearman’s correlation: R² = 0.547, p < 0.0001) in all the cases shown in B. D. PANS “flare” IgG binding to striatal ChAT interneurons correlated with CY-BOCS scores (Pearson’s correlation: R² = 0.444, p = 0.013). E. Representative images of immunohistochemical staining of human IgG (green) and parvalbumin (PV, red). Arrowheads indicate human IgG binding to PV-positive interneurons. F. IgG binding to PV interneurons did not differ between PANS “flare” and “recovered” groups (Wilcoxon paired signed-rank test: W = −3, p = 0.946). G. IgG binding to striatal PV interneurons did not correlate with CY-BOCS scores (Spearman’s correlation: R² = 0.064, p = 0.202). H. PANS “flare” IgG binding to striatal PV interneurons did not correlate with CY-BOCS scores (Pearson’s correlation: R² = 0.205, p = 0.121). I. Representative images of immunohistochemical staining of human IgG (green) and choline acetyltransferase (ChAT, red). Arrowheads indicate human IgG binding to ChAT-positive neurons in the medial septum (septal ChAT neurons). J. IgG binding to septal ChAT neurons was decreased in the “recovered” samples relative to PANS “flare” (Wilcoxon paired signed-rank test: W = −69, p = 0.013; r(equivalent) = 0.473). K. Across all samples, IgG binding to septal ChAT neurons showed a slight but significant correlation with CY-BOCS scores (Spearman’s correlation: R² = 0.192, p = 0.025). L. PANS “flare” IgG binding to septal ChAT neurons did not correlate with CY-BOCS scores (Pearson’s correlation: R² = 0.004, p = 0.833). Data are shown as mean ± S.E.M. Each data point represents mean value obtained from 6 mice for each plasma. *p < 0.05, ***p < 0.001, ****p < 0.0001; Numbers in parentheses indicate sample size. Dotted lines in correlation analyses (C, D, G, H, K and L) indicate 95% confidence intervals. Scale bars: 20 μm.

To assay the effect of PANS plasma on CIN activity (Fig. 4), acute mouse striatal slices treated with plasma (as above) were immunostained with anti-phospho-rpS6 (see Supplementary Table 3), following the same procedure. Each plasma was tested in brain sections from 5 mice, with 6 images chosen randomly without overlap collected in each section using an Olympus Fluoview FV-1000 confocal microscope with a 20×/0.85NA objective. Acquisition settings were kept constant for all image acquisitions within each experiment.

Automated quantitation of mean fluorescence intensity within each cell was achieved using Fiji ImageJ from NIH (https://imagej.net/Fiji/Downloads) with batch processing as described (Xu et al., 2021). Briefly, images of neuronal marker immunostaining (ChAT or PV) were thresholded and used to generate regions of interest (ROIs) corresponding to cell bodies of the selected cell type. These ROIs were then overlayed on the corresponding images of human IgG immunostaining (Figs. 1–3) or phospho-rpS6 immunostaining (Fig. 4), and immunostaining within each cellular ROI was quantified.

2.7. Data analysis

Plasma samples were processed in multiple batches. All plasmas were diluted to 500 mg/dL and randomized before testing. To minimize batch effects, PANS “Flare” plasmas, age- and gender-matched controls, and PANS “Recovered” draws (when available) were balanced within each batch. Batch effects on raw values were small in most cases (Supplementary Fig. 1); nevertheless, to reduce variability, data were normalized to the mean value of the control samples processed in the same batch (Fig. 1–3) or the aCSF-treated control from the same batch (Fig. 4) for analysis. All data are aggregated by plasma sample (i.e. the n for each analysis is the number of plasma samples, not the number of images, slices, or mice) and expressed as mean ± S.E.M. All raw and normalized values are shown in Supplementary Table 4. Statistical analyses were performed using SPSS Statistics 28 (IBM, New York, NY) or Prism 9.4 (GraphPad Software, La Jolla, California). Significance (p < 0.05) was determined by two-tailed t-test (unpaired or paired), one-sample t-test, or one-way analysis of variance (ANOVA) with post-hoc Tukey’s test for normally-distributed data, or by Wilcoxon paired signed-rank test and Kruskal-Wallis H test for non-normally-distributed data (data normality was evaluated using the Shapiro-Wilk test in Prism 9.4). Pearson or Spearman correlation was used to examine relationships between measures. Test-retest repeatability was evaluated using intraclass correlation coefficients, as described (Koo and Li, 2016). Effect sizes were calculated where significant differences were found between groups. Specifically, Cohen’s d and eta squared (η²) were used for parametric data; whereas r(equivalent) was used for non-parametric data (analyzed by Wilcoxon paired signed-rank test) (Rosenthal, 1991; Rosenthal and Rubin, 2003). Chi-square was used to examine categorical (positive or negative) differences between groups, followed by evaluation of Cramer’s V (Cohen, 2013) and odds ratios (Szumilas, 2010) as measures for effect size. The specific test used in each analysis is indicated in the corresponding figure legend. The details of all statistical analyses are shown in Supplementary Table 5.

3. Results

3.1. Plasma antibodies from children with PANS.

Children with PANS had lower total IgG titers in plasma collected during a symptom flare than matched controls (t[48] = 2.228, p = 0.031). In the subset of children for whom paired blood draws during a flare and during a period of recovery were available, IgG titers did not differ between flare and recovery plasma samples (2-tailed paired t-test: t[12] = 0.450, p = 0.661). Titers of PANS flare (N = 25) or PANS flare-recovery pairs (N = 13 each group) correlate with symptom severity as measured by CY-BOCS scores (not shown).

3.2. Elevated binding of antibodies from PANS plasma to mouse striatal cholinergic interneurons

Plasma IgG from children with PANS, collected during symptom flare, showed elevated binding to adult mouse striatal CINs, relative to plasma from age- and gender-matched controls (Fig. 1A–B; Cohen’s d = 0.922). This effect size is large, but somewhat smaller than that seen in our previous study (Xu et al., 2021) of children with strictly defined PANDAS who responded to IVIG (d = 1.502). There were no effects of gender or age at sample draw (not shown); we therefore did not include these covariates in subsequent analyses, for simplicity. IgG titers did not correlate with values of IgG binding to mouse striatal CINs (not shown).

To categorize subjects as ‘positive’ or ‘negative’ for CIN binding, we initially used the 95^th percentile of binding in the control group controls as a cutoff, as is commonly done in other contexts (Agmon-Levin et al., 2014). Chi-square analysis revealed a trend towards more positive cases in the PANS flare group than in the control group (Supplementary Fig. 2B; p = 0.123). Next, we plotted a receiving operating characteristic curve (ROC) to determine an optimal cutoff value (Supplementary Fig. 2C). The area under the curve (AUC = 0.749) indicated an overall acceptable accuracy of CIN binding as a test for PANS (Hosmer Jr et al., 2013) and identified the optimal cutoff value as the 88^th percentile of the control group. Chi-square analysis using this cutoff value showed the PANS flare group had more positive cases relative to the control group (p = 0.0002; Supplementary Fig. 2D).

Next, we re-analyzed individual images, categorizig each individual cell as ‘positive’ or ‘negative’ for IgG binding based on signal-to-background ratio (≥ 1.5) and found the PANS flare group had higher percentage of IgG⁺-ChAT⁺ cells than controls (p = 0.007; Supplementary Fig. 2E), consistent with a previous study in which individual cells were manually categorized (Frick et al., 2018) and with the intensity-based analysis presented here (Fig. 1B). Using this measure, and setting a cutoff value at the 95^th percentile of the control distribution (Supplementary Fig. 2F), chi-square analysis showed increased IgG-positive cells in the PANS flare group (p = 0.017). As before, we obtained the optimal cutoff value based on the ROC (Supplementary Fig. 2G). The AUC (0.757) indicated an overall acceptable accuracy of the test; the optimal cutoff value was at the 88^th percentile of the control group. Chi-square analysis using this cutoff confirmed that the PANS flare group had significantly more positive cases (p = 0.001; Supplementary Fig. 2H).

No statistically significant group differences were found in IgG binding to parvalbumin (PV)-expressing striatal interneurons (Fig. 1C–D) or to cholinergic neurons in the medial septum (Fig. 1E–F), although trends were seen in both cases.

We selected 6 PANS “flare” cases with relatively high plasma antibody binding to mouse striatal CINs, available “recovered” draws, and corresponding age- and gender-matched controls (see Supplementary Table 1) for further analysis. As PANS is a pediatric condition, we repeated the IgG binding assay in this subset using sections from juvenile mouse brains. As in adult brains, PANS flare IgG showed elevated binding to CINs but not to PV interneurons or ChAT neurons in the medial septum (Supplementary Fig. 3). IgG binding to CINs in adult and juvenile mouse tissue was correlated.

The elevated binding of PANS flare IgG to CINs, but not to PV interneurons, was further confirmed in mouse striatal slice cultures (Supplementary Fig. 4A, B, E, F).

3.3. Decreased plasma IgG binding to CINs in “recovered” draws

For a subset of these relapsing and remitting PANS cases, plasma was available both during symptom flare and during a period of symptomatic recovery (see Methods for details). We examined the difference in plasma IgG binding to adult mouse CINs bewteen symptom flare and recovery samples in these subjects (N = 13). Antibody binding to CINs was reduced in recovery samples and paralleled symptom improvement (Fig. 2A–B; d = 1.258). IgG binding to CINs correlated with symptom severity, as measured using the Child Yale-Brown Obsessive-Compulsive Scale (CY-BOCS (Scahill et al., 1997); Fig. 2C–D). There was no similar change in plasma IgG binding to PV interneurons (Fig. 2E–F) or correlation between IgG binding to PV interneurons in adult mouse brains and CY-BOCS scores (Fig. 2G–H). Unexpectedly, there was a reduction in plasma IgG binding to adult septal cholinergic neurons in conjunction with symptom improvement (Fig. 2I–J). A weak but statistically significant correlation was found between IgG binding to septal cholinergic neurons and CY-BOCS scores (Fig. 2K–L). Total IgG titer did not correlate with IgG binding to CINs in the flare-recovery samples (not shown).

We repeated this flare-recovered comparison in juvenile mouse sections in the six selected samples (Supplementary Table 1). We found the same binding patterns in juvenile mouse brains, i.e. the “recovered” samples showed decreased binding to CINs and septal cholinergic neurons, but not to PV interneurons (Supplementary Fig. 3 IgG binding in adult and juvenile mouse brain sections were moderately correlated in CINs and PV interneurons, with a trend-level correlation (based on Cohen’s guidelines (Cohen, 2013)) in septal cholinergic neurons (Supplementary Fig. 3D, H, L). We evaluated repeatability using the intraclass correlation coefficient (ICC) and found moderate consistency between results from adult and juvenile mouse brain slices (Koo and Li, 2016).

Finally, we examined the flare-recovered comparison in slice cultures. CIN binding was again normalized in matched plasma samples at recovery draws (Supplementary Fig. 4A, C). Both Pearson’s correlation and ICC indicated moderate consistency between results from slice cultures and brain sections (Supplementary Fig. 4D). No significant difference was found in IgG binding to PV interneurons in slice cultures. The values obtained in the different assays showed a trend towards a positive correlation (Supplementary Fig. 4E, G, H).

3.4. PANS IgG shows elevated binding to CINs, but not to PV interneurons, in normal human brains

We replicated the IgG binding analysis in the selected 6 PANS “flare” cases, matched recovery samples, and healthy controls (Supplementary Table 1) on post mortem human brain tissue. We additionally tested PANS “flare” plasmas from which IgG was depleted using Protein A/G agarose beads; IgG depletion was verified using an anti-human IgG antibody on a dot blot (Supplementary Fig. 5).

In the caudate, one-way ANOVA revealed a significant main effect of group (η² = 0.828), with significantly elevated IgG binding to CINs in the PANS “flare” group but not in “recovery” samples, relative to controls (Fig. 3A–B). All plasma-treated brain sections showed higher IgG immunostaining than the IgG-depleted group (Fig. 3A–B). Similar results were seen with IgG binding to CINs in the putamen (η² = 0.681; Fig. 3C–D). In PV interneurons, in contrast, significant main effects of group were driven by the IgG-depleted group; there were no differences between control and PANS plasmas, during “flare” or “recovered”, in caudate (Fig. 3E–F) or putamen (Fig. 3G–H).

To assess the reproducibility and reliability of our assays, we examined the relationship of IgG binding to interneurons in the mouse (Fig. 1) and human brain (Fig. 3); these values correlated strongly (Supplementary Fig. 6). Good to excellent repeatibility was also revealed by ICCs. To assess the possibility that our results are confounded by autofluorescence, which can be a problem in immuofluorescent studies in human post-mortem tissue despite the use of autofluorescence quenching, we replicated binding to human caudate using an enyzyme-based colorimetric approach. This again showed significantly elevated IgG binding to CINs in the PANS “flare” group relative to controls, which was reversed in in “recovery” samples (Supplementary Fig. 7).

3.5. PANS plasma IgG reduces CIN activity

We next tested the effect of PANS IgG on interneuron function, using phosphorylation of ribosomal protein S6 (p-rpS6) as a marker of spontaneous neural activity (Bertran-Gonzalez et al., 2012; Knight et al., 2012; Pirbhoy et al., 2016; Xu et al., 2021). Acute mouse striatal slices were incubated with plasma from the 6 selected PANS subjects and age- and gender-matched controls described above. In CINs, there was a significant main effect of pretreatment (one-way ANOVA, p = 0.006, η² = 0.460). Control plasma did not significantly alter p-rpS6 levels, relative to vehicle. PANS “flare” plasma treatment significantly lowered p-rpS6 levels relative to control; this was not seen in “recovered” plasma or after IgG depletion. Neither “recovered” plasma nor IgG-depleted PANS “flare” plasma reduced p-rpS6 in CINs relative to ACSF (Fig. 4A–B).

Across the control, PANS “flare”, and “recovered” groups, plasma IgG binding to CINs correlated negatively with p-rpS6 levels after plasma pretreatment (Fig. 4C). p-rpS6 levels also showed a weak correlation with CY-BOCS scores in PANS “recovered” samples (Fig. 4D). Importantly, the difference between PANS “flare” and “recovered” IgG binding to CINs in the same subjects correlated with the difference in p-rpS6 levels after plasma incubation (Fig. 4E). These results indicate that PANS “flare” IgG binding negatively regulates the activity of striatal CINs.

No group difference was found in the activity of PV-expressing interneurons by one-way ANOVA (Fig. 4F–G). No correlations were found between PV p-rpS6 levels and other measures (Fig. 4H–J).

We used whole-cell patch recording in acute mouse brain slices from ChAT^BAC-eGFP transgenic mice (Gamage et al., 2020) to confirm the effect of PANS plasma on CIN activity. Because this experiment uses a large amount of serum it was performed with a single selected PANS sample that showed the highest IgG binding to mouse striatal CINs and the matched control (Fig. 1 andSupplementary Table 1), as a confirmatory test. CINs were readily identified by their size and morphology and confirmed by their electrophysiological properties (Supplementary Fig. 8A–C). PANS and control plasmas did not significantly affect intrinsic membrane properties of CINs, although there was a trend towards hyperpolarized resting membrane protetial (RMP) after treatment with the PANS plasma (not shown).

CINs receive various synaptic inputs, including glutamate and dopamine (DA) (Lim et al., 2014), and we have previously shown PANDAS serum to alter response to these transmitters (Xu et al., 2021). We examined CIN response to bath application of agonists for these receptors. As predicted, the ionotropic glutamate agonist AMPA produced a robust increase in action potential frequency in control plasma-pretreated CINs; this was significantly attenuated after pretreatment with PANS plasma (Supplementary Fig. 8D–E). Similarly, bath application of DA increased action potential frequency in control plasma-pretreated, but not PANS plasma-pretreated, CINs (Supplementary Fig. 8F–G).

4. Discussion

PANS, like PANDAS, has been proposed to derive from the production of autoantibodies that cross-react with targets in the brain, including in the basal ganglia, following infection in a susceptible individual, leading to neuroinflammation and subsequent motor and behavioral abnormalities (Chan and Frankovich, 2019; Swedo and Williams, 2017; Williams and Swedo, 2015). In support of this post-infectious autoimmune hypothesis, high rates of autoimmune disorders and inflammatory conditions have been reported in children with PANS and PANDAS, and in their first-degree relatives (Calaprice et al., 2017; Frankovich et al., 2015a; Gromark et al., 2019; O’Dor et al., 2022). Meta-analysis documents an increase in proinflammatory cytokines (most clearly TNFα, but probably also IL-17 and IL-1β) in children with OCD and movement disorder symptoms (Fabricius et al., 2022), though negative findings have also been reported (Cosco et al., 2019). Serum/plasma immunoglobulin levels have been reported to be altered in PANS/PANDAS subjects, with lower levels of IgG, IgA, and IgM, and higher levels of IgE (Calaprice et al., 2017; Frankovich et al., 2015a; Gromark et al., 2019; Leonardi et al., 2023; Murphy et al., 2015). Neuroimaging suggests that patients with PANS and PANDAS display changes to the basal ganglia region consistent with neuroinflammation; one PET study suggested microglial activation in the basal ganglia during the acute phase of illness (Cabrera et al., 2019; Giedd et al., 2000; Giedd et al., 1996; Kumar et al., 2015; Zheng et al., 2020).

We have previously described binding of PANDAS antibodies to striatal CINs, using sera collected over several years at the National Institute of Mental Health (Frick et al., 2018; Xu et al., 2021). To test whether this binding pattern generalizes to PANS, a more heterogeneous diagnosis, and to samples collected at other clincial sites, we here performed similar analyses in plasma from children with relapsing and remitting PANS, assessed at the Immune Behavioral Health Clinic at Stanford University. We find elevated binding of antibodies from children with PANS with evidence of Streptococcal infection, collected during symptom flare, to striatal CINs but not to two other neuron types, in both adult mouse and human brain (Figs. 1 and 3; Supplementary Fig. 7) as well as in juvenile mouse brain sections and adult mouse brain slice cultures (Supplementary Figs. 2 and 3). These results replicate our previous findings in patients with strictly-defined, IVIG-responsive PANDAS (Frick et al., 2018; Williams et al., 2016; Xu et al., 2021). The consistency and specificity of elevated antibody binding to CINs, despite the differences in diagnosis, geography, and sample collection (Stanford PANS clinic vs NIMH; PANS plasma vs PANDAS serum), increases confidence in the finding. Unlike our previous work with PANDAS, we find a trend towards increased antibody binding to PV interneurons and cholinergic septal neurons, suggesting that antibody binding to these other cell types might be relevant in a subset of the more heterogeneous PANS population.

Elevated IgG binding is not seen in samples collected from a subset of these subjects during a period of symptom recovery; binding correlates with symptom severity (as measured by CY-BOCS scores; Fig. 2). This is similar to what we have reported previously after IVIG treatment in PANDAS (Frick et al., 2018; Xu et al., 2021); however, in the PANS cases analyzed here, patients were treated clinically using a variety of interventions, and the order of and time between “flare” and “recovery” blood draws varies. Of note, we also find a significant reduction in antibody binding to cholinergic neurons in the medial septum at recovery in these subjects (Fig. 2), which was not observed in our previous study of PANDAS (Xu et al., 2021); this may reflect differences in the patient populations or in the treatments administered.

The molecular targets of these IgG antibodies remain to be elucidated. A number of potential antibody targets (including D1R and D2R) have been described in Sydenham chorea (SC), Tourette syndrome, and PANDAS. Some autoantibody targets may be shared among these disorders; others are likely to be distinct. Antibodies binding to D1R and D2R have been more consistently reported in SC (Ben-Pazi et al., 2013; Cox et al., 2013; Kirvan et al., 2007; Kirvan et al., 2003; Kirvan et al., 2006b) than in PANDAS (Brilot et al., 2011; Dale et al., 2012; Morris et al., 2009; Singer et al., 2005). We previously examined binding of antibodies to D1- and D2-expressing medium spiny neurons in the striatum and found no difference between PANDAS and control sera (Xu et al., 2021). It remains possible that PANS and PANDAS-associated antibodies bind to D2 receptors on CINs; D2Rs are known to be expressed at the postsynaptic sites in the majority of CINs (Dawson et al., 1988; Yan et al., 1997).

Here, as in our previous work (Xu et al., 2021), we find that PANS antibody binding to CINs reduces their activity, indexed using an activity-dependent molecular marker, p-rpS6 (Fig. 4) and confirmed using single-cell patch clamp electrophysiology (Supplementary Fig. 8). This effect on CIN activity is not seen with plasma collected during recovery, or in flare plasma from which IgG has been depleted. This supports the conclusion that IgG binding to CINs, as opposed to other activities in the flare plasma, is responsible for CIN inhibition.

Precisely how PANS antibody binding affects CINs, and how binding leads to reduced activity, remains to be determined. Previous studies in other contexts have shown various PANDAS and SC antibodies to decrease cyclic adenosine monophosphate (cAMP) levels and to induce calcium calmodulin protein kinase II (CaMKII) activity and dopamine release (Cox et al., 2013; Cunningham and Cox, 2016; Kirvan et al., 2006b). The relationship of these findings to the alteration of CIN activity shown here merit further study.

We have used phosphorylation of rpS6 to index neuronal activity (Biever et al., 2015; Meyuhas, 2008, 2015), but rpS6 is an important regulator of a range of cellular processes and may directly participate in pathophysiologically relevant events. rpS6 is regulated by phosphorylation at 2 pairs of sites: Ser235/Ser236 and Ser240/Ser244 (Bertran-Gonzalez et al., 2012) and is regulated by a number of cell signaling pathways, including S6 kinases (S6Ks) protein phosphatase 1 (PP1); these molecules are in turn regulated by the Akt-mammalian target of rapamycin (mTOR)-S6K signaling cascade (Biever et al., 2015; Meyuhas, 2008, 2015) and by cAMP-PKA through dopamine- and cAMP-regulated phosphoprotein, 32 kDa (DARPP-32) (Greengard, 2001; Hemmings et al., 1984). This network of signaling pathways is regulated by dopamine as well as by neural activity. It will be important to further characterize the effects of PANS and PANDAS autoantibody binding in CINs on these and related cellular signaling mechanisms.

Although few in number (about 1% of the total striatal cell population), CINs are key modulators of striatal function, releasing acetylcholine (ACh) across broad striatal territories (Apicella, 2007; Burguiere et al., 2015; Lim et al., 2014; Poppi et al., 2021; Rapanelli et al., 2017a; Tepper and Bolam, 2004). Several ACh receptors (AChRs) are expressed in GABAergic SPNs and interneurons in the striatum (Goldberg et al., 2012; Lim et al., 2014). Interestingly, inhibitory M4 mAChRs are expressed at higher levels in SPNs of the direct pathway (dSPNs) than in Spns of the indirect pathway (Hersch et al., 1994; Kreitzer, 2009; Yan et al., 2001). Thus, inhibition of CINs may result in an imbalance between the direct and indirect pathways (Lanciego et al., 2012; Nelson and Kreitzer, 2014). CINs are known to co-release glutamate (Higley et al., 2011) and GABA (Granger et al., 2016; Nelson et al., 2014), and to regulate dopamine release (Threlfell et al., 2012; Zhou et al., 2001). They may also modulate neuroinflammation: activation of α7 nAChRs and M3 mAChRs decreases microglial activation and cytokine (TNFα and IL-6) secretion in animal models (Binning et al., 2020; Dash et al., 2016; Gamage et al., 2020; Pannell et al., 2016). Thus, CIN dysregulation may alter basal ganglia information processing by multiple mechanisms.

Previous work has highlighted the contribution of CINs to symptomatology in OCD and TS. CINs are reduced in number in post-mortem tissue from individuals with TS (Kataoka et al., 2010; Lennington et al., 2016), as well as in TS-derived basal ganglia organoids (Brady et al., 2022). In animal models, selective depletion of striatal CINs results in repetitive behavioral pathology (Rapanelli et al., 2017b; Xu et al., 2015) and increased functional connectivity between frontal cortical areas and the motor region of the striatum (Martos et al., 2017). Thus, reduced activity by PANS antibody binding to CINs may have a range of effects across relevant brain circuitries.

There are several limitations to this study. First, although we extend our previous findings (obtained in strictly-defined PANDAS cases) to a more heterogeneous population (relapsing and remitting PANS, from a different clinical setting), and confirmed the binding patterns in both adult and juvenile mouse brain sections and adult mouse brain slice cultures, these PANS cases were selected for a history of Streptococcal infections. Unselected PANS cases may be still more heterogeneous; this will be examined in future work. Importantly, an autoimmune pathophysiology may not be restricted to PANS and PANDAS, but may occur in a broader subset of OCD cases. Anti-basal ganglia antibodies have been reported in idiopathic OCD (Bhattacharyya et al., 2009; Cox et al., 2015; Dale et al., 2005; Pearlman et al., 2014). Further work is required to examine whether elevated antibody binding to CINs is also present in non-PANS/PANDAS OCD or in other diagnostic groups.

Second, our immunofluorescence-based ex vivo assay is semi-quantitative and has significant background. Background fluorescence may derive from autoantibodies in control plasma, from nonspecific interactions between IgG and the target tissue, or from IgG-independent sources of fluorescence. Of note, IgG-depleted plasmas retains 10–30% of the signal of the control groups in the binding to human brain (Fig. 3). Autofluorescence may be caused by components other than IgG species in human serum/plasma samples (Wolfbeis and Leiner, 1985). This issue was overcomed by using enzyme-based double staining (Supplementary Fig. 6). However, while enzymatic IHC (e.g. HRP-DAB reaction) is widely used in the literature, there are concerns about the usage of DAB staining for quantification in light microscopy. It has been shown that the brown product from HRP-DAB reaction is actually a scatterer rather than an absorber of light, and thus the optical density does not follow a linear relationship with its concentration (van der Loos, 2008). Our confirmation of elevated PANS IgG binding to CINs in both mouse and human brain tissue despite these limitations is encouraging.

Third, we have focused on three neuron types (CINs, PV interneurons, and septal cholinergic neurons). We cannot exclude the possibility that PANDAS/PANS IgG may bind to other cells and in other brain regions – indeed, we think this is likely to be the case.

Fourth, the samples sizes were small in some experiments due to limited supplies of clinical plasma samples, human brain tissue, and the time-consuming nature for some of the work. A subset of control, PANS flare, and recovery samples was tested in the IgG binding to juvenile mouse brain, adult brain slice cultures, human brain, and spontaneous neural activity using p-rpS6 as a molecular marker (Figs. 3 and 4; Supplementary Figs. 3, 4 and 7). Whole-cell recording of CINs was only conducted on a single control-PANS pair (Supplementary Fig. 8). Nevertheless, results in these experiments were in agreement with our hypothesis and supportive of findings in the full cohort.

The identity of antibody targets and key signaling pathways affected by PANDAS/PANS antibody binding to CINs remain to be elucidated. The recent development of high-throughput screening technology (Wang et al., 2022; Wang et al., 2021) may facilitate the discovery of molecular targets of these autoantibodies. Given the etiological and phenomenological heterogeneity of these conditions, we expect that distinct autoantibodies will be present in different PANDAS/PANS cases. The current work and our previous findings point to CINs as a plausible locus of pathology in PANDAS/PANS, providing a framework to test candidate pathogenic autoantibodies.

In conclusion, this work in PANS replicates the antibody binding pattern and functional consequences on cellular activity previously seen in PANDAS (Frick et al., 2018; Xu et al., 2021), with elevated IgG binding to striatal CINs in both mouse and human brain, and consequent decreased CIN activity in an ex vivo assay. Antibody binding to CINs (but not to PV interneurons) and reduction of CIN activity parallels symptom improvement during recovery. These finding are in agreement with the hypothesis that at least a subset of PANS cases have an immunological pathogenesis. Striatal CINs may be a locus of interest to investigate the pathophysiology in PANS.

Highlights.

• PANS IgG drawn at flare shows elevated binding to CINs in both mouse and human brain.

• PANS IgG binding to CINs reduces their activity using p-rpS6 as a marker.

• Elevated IgG binding to CINs is resolved in the same subject during symptom recovery.

• Reduction of PANS IgG binding to CINs parallels symptom improvement during recovery.

• Striatal CINs may be a locus of interest to investigate the pathophysiology in PANS.

Abbreviations

ACh

acetylcholine

AChRs

acetylcholine receptors

aCSF

artificial cerebrospinal fluid

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

ANOVA

analysis of variance

AP

alkaline phosphatase

AUC

area under the curve

cAMP

cyclic adenosine monophosphate

CaMKII

calcium calmodulin protein kinase II

ChAT

Choline Acetyltransferase

CINs

cholinergic interneurons

CY-BOCS

Child Yale-Brown Obsessive-Compulsive Scale

DA

dopamine

DAB

diaminobenzidine

eGFP

enhanced green fluorescent protein

ELISA

enzyme-linked immunosorbent assay

HRP

horseradish peroxidase

ICC

intraclass correlation coefficient

IgG

Immunoglobulin G

IL-6

interleukin 6

IVIG

intravenous immunoglobulin

mAChRs

muscarinic AChRs

mTOR

mammalian target of rapamycin

OCD

obsessive-compulsive disorder

PANDAS

Pediatric Autoimmune Neuropsychiatric Disorder Associated with Streptococcus

PANS

Pediatric Acute-onset Neuropsychiatric Syndrome

PV

Parvalbumin

ROC

receiver operating characteristic curve

rpS6

ribosomal protein S6

SC

Sydenham chorea

S.E.M.

standard error of the mean

TH

tyrosine hydroxylase

TMB

3,3’,5,5’-tetramethylbenzidine

TNFα

tumor necrosis factor alpha

TS

Tourette syndrome

Data availability

Data will be made available on request.