CNTNAP2 is cleaved by furin and α-secretase

To reveal the proteolytic processing pathway of CNTNAP2, we first investigated whether CNTNAP2 can be cleaved by ADAM10/17-dependent α-secretase. pRK5-CNTNAP2 was co-transfected with an empty vector, pRK5M-ADAM10 or pRK5M-ADAM17, into HEK293 cells. Two bands of CNTNAP2 full-length protein were detected at 170 kDa, the higher glycosylated mature form and the lower immature form (Fig. 1a). Mature CNTNAP2 was reduced by ADAM10 to 33. 11 ± 3.97% (p < 0.0001) and by ADAM17 to 32.48 ± 2.16% (p < 0.0001) compared with the control, while immature CNTNAP2 was not significantly affected (Supplementary Fig. 1a). ADAM10 and ADAM17 cleaved CNTNAP2 to generate two C-terminal fragments (CTFs) at ~15 kDa, a predominant CTFα1 and a weaker CTFα2. Expression of ADAM10 and ADAM17 increased CTFα1 to 2.65 ± 0.36 folds (p = 0.0483) and 3.84 ± 0.68 folds (p = 0.0026), while markedly elevated CTFα2 to 6.56 ± 1.63 (p = 0.0332) folds and 14.83 ± 1.72 folds (p < 0.0001), respectively (Fig. 1b). To validate the cleavage of CNTNAP2 by α-secretase, we generated ADAM10-KO and ADAM17-KO HEK cell lines via the CRISPR/Cas9 technique. Genetic ablation of ADAM10 drastically reduced CTFα1, while ablation of ADAM17 reduced the CTFα2 level without significantly affecting CTFα1 production (Supplementary Fig. 1c). While the CTFα1/mCNTNAP2 ratio was reduced to 35.2 ± 8.72% (p = 0.0017) in ADAM10-KO cells, the ratio was increased to 139.9 ± 19.54% (p = 0.1106) in ADAM17-KO cells (Fig. 1c, d), due to the compensatory effect between ADAM10 and ADAM17. Our data clearly demonstrates that mature CNTNAP2 is cleaved into two CTFα by both ADAM10 and ADAM17: the major α-secretase product CTFα1 is predominantly generated by ADAM10, and the minor α-secretase product CTFα2 is predominantly produced by ADAM17, respectively.

Fig. 1

CNTNAP2 is cleaved by α-secretase and furin. a, b ADAM10 and ADAM17 cleaved mature CNTNAP2 (mCNT) in HEK cells. HEK cells were co-transfected with CNTNAP2 and empty vector, ADAM10 or ADAM17 plasmids, and harvested 24 h after transfection for Western Blot analysis. Non-transfected HEK (NT HEK) was used as a negative control. CTFα1 and CTFα2 were increased/generated by ADAM10 and ADAM17. n = 4 independent experiments, ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test, *p < 0.05; **p < 0.01; ***p < 0.001. c, d ADAM10 is the major α-secretase that generates CTFα1. CNTNAP2 was co-transfected with GFP into HEK, ADAM10-knockout (KO), or ADAM17-KO HEK cells. The CTFα1/mCNT ratio was reduced in ADAM10-KO cells while increased in ADAM17-KO cells. n = 3 independent experiments, unpaired t-test, **p < 0.01. e, f Furin cleaved mature CNTNAP2 in HEK cells. HEK cells were co-transfected with CNTNAP2 and vector or furin. CTFf at 55 kDa was increased by furin. n = 3 independent experiments, unpaired t-test, *p < 0.05. g CNTNAP2 on the cell membrane was cleaved in HEK and N2a cells. CNTNAP2 was co-transfected with vector (myc), ADAM10, ADAM17, or furin plasmids into HEK and N2a cells. Cells were fixed 24 h after transfection, and immunocytochemistry (ICC) was performed. CNTNAP2 was detected by an anti-CNTNAP2 antibody targeting the N-terminal 1001–1042 aa outside the cell membrane. CNTNAP2 was mainly expressed on the cell membrane and was reduced by co-transfected plasmids. Images were acquired using Zeiss Apotome to reflect the morphology. Scale bar represents 50 μm. h Quantification of ICC in HEK and N2a. CNTNAP2 mean intensity was measured using the corresponding conventional fluorescence images in Supplementary Fig. 2. n = 10 cells from 2 independent experiments, ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test, ****p < 0.0001. All the results are expressed as mean ± SEM

We next investigated whether CNTNAP2, like the Notch-1 protein, could also be cleaved by furin. Furin cleaved mature CNTNAP2 to 52.87 ± 5.86% (p = 0.0013) (Supplementary Fig. 1b) and enhanced a CTFf at 55 kDa to 1.22 ± 0.07 folds (p = 0.0413) in HEK cells (Fig. 1e, f). The results were further confirmed by furin inhibitor treatment (Fig. 3a–d). Consistently, full-length CNTNAP2 was reduced to 39.42 ± 6.65% (p = 0.0028) by ADAM10, 21.13 ± 3.73% (p = 0.0005) by ADAM17, and 51.53 ± 15.48% (p = 0.0104) by furin, while CTFα2 was increased to 3.04 ± 0.27 and 8.09 ± 1.89 (p = 0.002) folds, and CTFf was elevated to 4.87 ± 1.81 (p = 0.041) folds by corresponding proteases in mouse neuroblastoma Neuro-2a (N2a) cells (Supplementary Fig. 1d, e). CNTNAP2 is mainly expressed on the cell membrane of both HEK and N2a cells (Fig. 1g). Consistent with Western blot results, overexpression of ADAM10, ADAM17, or furin drastically reduced CNTNAP2 level on the membrane to 25.27 ± 2.02%, 21.06 ± 1.86%, or 65.82 ± 3.74% in HEK cells, and 24.41 ± 1.74%, 16.13 ± 0.84%, or 47.91 ± 4.54% in N2a cells (p < 0.0001 in all the comparisons with controls) (Fig. 1h, Supplementary Fig. 2). Collectively, our results demonstrated that CNTNAP2 undergoes proteolytic processing by furin and ADAM10/17-dependent α-secretase.

Identification of α-cleavage sites on CNTNAP2

To identify the α-secretase cleavage sites, plasmids expressing the last 178, 108, and 91 aa of CNTNAP2 C-terminus were co-transfected with ADAM10 or ADAM17. CTFα1 and CTFα2 were generated from both 178 and 108 constructs while only CTFα1 was produced from the 91 construct, indicating that the α-secretase site generating CTFα2 is within the last 108 aa, and the CTFα1 is within the last 90 aa (Supplementary Fig. 3a). Next, Mass Spectrometry and N-terminal sequencing were performed to identify the cleavage sites. Synthetic peptide 1 (C-terminal 108-84 aa of CNTNAP2) and peptide 2 (C-terminal 96-72 aa) were treated with the recombinant human ADAM17 protein, and Mass Spectrometry results showed that ADAM17 cleaved peptide 1 between H97 and L96 and peptide 2 between A80 and I79 (Fig. 2a). N-terminal sequencing also identified that CTFα1 started from IRNGV, and CTFα2 started from LDSAS (Fig. 2b, Supplementary Fig. 3b). In addition, we generated truncated CNTNAP2 C-terminal protein ladders C79 and C96, which perfectly matched the size of CTFα1 and CTFα2 (Fig. 2c). These results clearly demonstrated that CNTNAP2 is cleaved by α-secretase primarily at I79 to generate C79 (CTFα1), and less at L96 to yield C96 (CTFα2), respectively.

Fig. 2

Identification of α-secretase cleavage sites on CNTNAP2. a Peptide in vitro cleavage. Two peptides were incubated in the assay buffer with or without ADAM17 at 37 °C for 24 h and then were sent to Mass Spectrometry analysis. The extracted ion chromatogram showed two intact peptides (top) and cleaved complementary fragments (bottom). Peptide 1 (M108-G84) was cleaved by ADAM17 at 97H/L96. Peptide 2 (L96-N72) was cleaved at 80 A/79I. b N-terminal sequencing results identified the first 5 amino acids (aa) of CTFα1 as IRNGV and the first 5 aa of CTFα2 as LDSAS. The bottom sequence shows the last 108 to 72 aa of CNTNAP2 from the C-terminus and α-secretase cleavage sites at L96 and I79. c Protein ladders of the C-terminal CNTNAP2. C79 corresponded to the size of CTFα1, and C96 showed the same size as CTFα2. d, e Mutations at the α-secretase cleavage sites L96 and I79 affected CNTNAP2’s cleavage. Wildtype (WT) or mutant plasmids were co-transfected with empty vector (EV), ADAM10 (AD10), or ADAM17 (AD17) into HEK cells. Alterations in the migration rate were observed in constructs containing D98R. Two upper/lower blots show samples from the same experiment. The parallel blots were processed in the same electrophoresis chamber and scanned together simultaneously. n = 3 independent experiments, two-way ANOVA followed by Dunnett’s multiple comparisons test (compare ADAM10 and ADAM17 with Control for each plasmid), row factor = mutation, column factor = ADAM10/17 overexpression. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. All the results are expressed as mean ± SEM

Next, plasmids expressing the 108aa protein containing mutations at α-secretase sites were co-transfected with ADAM10 or ADAM17 into HEK cells. ADAM10 and ADAM17 cleaved the wildtype (WT) construct and markedly enhanced the C96 levels to 25.63 ± 2.21 and 29.19 ± 3.02 folds, respectively (Fig. 2d, e). However, C96 levels were only increased to 1.98 ± 0.86 and 1.41 ± 0.45 folds in H97E, and to 2.07 ± 1.66 and 2.11 ± 1.10 folds in L96P, respectively, by ADAM10 and ADAM17. The C96 generation was also blocked in DH98/97RE and HL97/96EP mutant constructs (Fig. 2d, e). C79 levels in WT were increased by ADAM10 to 5.40 ± 0.46 folds and by ADAM17 to 11.76 ± 1.10 folds, respectively (Fig. 2d, e). However, C79 generation was barely changed to 3.11 ± 1.29 and 1.69 ± 0.39 folds in I79P, and to 1.37 ± 0.84 and 0.76 ± 0.29 folds in AI80/79PP by overexpressed ADAM10 and ADAM17 (Fig. 2d, e). These results demonstrated that H97E and L96P mutations reduced C96 generation, whereas I79P and AI80/79PP blocked C79 generation. Notably, I79T corresponding to an ASD-associated I1253T mutation was found in a family with two autism patients.2 I79T significantly inhibited C79 generation, only reaching 1.29 ± 0.32 and 4.55 ± 1.21 folds upon ADAM10/17 overexpression (Fig. 2d, e). These results demonstrated that mutations at the α-cleavage sites L96 and I79 significantly affect CNTNAP2’s cleavage (Supplementary Table 1). In summary, ADAM10/17 cleaves CNTNAP2 at L96 and I79, and the ASD-associated pathogenic mutation I79T inhibits the α-cleavage-mediated C79 generation.

Sequential cleavage of CNTNAP2 by furin, α- and γ-secretase

To further examine the processing pathway of CNTNAP2 by furin and α-secretase, ADAM17 overexpressing cells were treated with ADAM17 inhibitor TAPI-1 and furin inhibitor-2 FI-2. TAPI-1 significantly elevated mature CNTNAP2 to 2.77 ± 0.41 folds (p = 0.0021) and reduced C79 level to 0.57 ± 0.09 fold (p = 0.0215), while it did not affect mature CNTNAP2 or C79 in the ADAM10-transfected cells (Fig. 3a, b). TAPI-1 also increased CTFf to 1.95 ± 0.40 folds, and FI-2 decreased CTFf to 0.47 ± 0.11 fold (p = 0.0148 compared with TAPI-1) in the ADAM17-transfected cells, while the combination of TAPI-1 and FI-2 brought the CTFf to 0.80 ± 0.30 fold, a level between TAPI-1 and FI-2 treatment alone (Fig. 3a, b, CNT + ADAM17 graphs), suggesting that furin-generated CTFf is further cleaved by ADAM17. The downstream CTF patterns also support this sequential cleavage. TAPI-1 reduced C79 level to 0.57 ± 0.09 fold (p = 0.0215) by inhibiting ADAM17 activity. FI-2 marginally decreased C79 to 0.72 ± 0.08 fold by reducing CTFf for the subsequent ADAM17 cleavage. As expected, TAPI-1 + FI-2 significantly reduced the C79 level to 0.49 ± 0.11 fold (p = 0.0086) in an additive manner. While TAPI-1 did not affect the C96 level (0.86 ± 0.08 fold), FI-2 significantly reduced C96 with (0.54 ± 0.07 fold, p = 0.0028) or without TAPI-1 (0.32 ± 0.06 fold, p = 0.0002) (Fig. 3a, b, CNT + ADAM17 graphs), indicating that C96 was also sequentially cleaved from CTFf by ADAM17. These results demonstrated that mature CNTNAP2 is sequentially cleaved by furin to generate CTFf, and subsequently, CTFf is cleaved by ADAM17 to generate C96 and C79.

Fig. 3

Sequential cleavages of furin, α- and γ-secretase. a Western blot of control (Vector), ADAM10 overexpressing, and ADAM17 overexpressing HEK cells. Exposure 1 was used to analyze CTFf in the ADAM17 overexpressing cells and C96 in the ADAM10 overexpressing cells; exposure 2 was used for assessing C79 and C96 in the ADAM17 overexpressing cells. TAPI-1 (10 μM) is an ADAM17-specific inhibitor; FI-2 (20 μM) stands for Furin inhibitor-2. b Quantification of the dynamic changes of mature CNTNAP2 and CTF levels in the ADAM10/ADAM17 transfected cells upon inhibitor treatments in (a). n = 3 independent experiments, ordinary one-way ANOVA followed by Tukey’s multiple comparisons test, *p < 0.05; **p < 0.01; ***p < 0.001. P values stand for comparisons with the Control group unless noted by the brackets. c, d Dynamic changes of CTFs in the furin transfected cells, showing the furin-ADAM17 cleavages. n = 3 independent experiments, ordinary one-way ANOVA followed by Tukey’s multiple comparisons test, *p < 0.05; **p < 0.01; ***p < 0.001. P values stand for comparisons with the Control group unless noted by the brackets. e Western Blot of control (Vector), ADAM10 overexpressing, and ADAM17 overexpressing HEK cells. GSI is γ-secretase inhibitor L-685,458 (20 nM). f Quantification of the sequential cleavage by α- and γ-secretase in (e). n = 3 independent experiments, unpaired t-test, *p < 0.05, **p < 0.01. g Schematic of CNTNAP2 processing. Mature CNTNAP2 undergoes sequential cleavages by furin and α-, γ-secretase. Furin cleaves CNTNAP2 to generate CTFf, which is further processed by α-secretase. α-secretase ADAM10 and ADAM17 share the same cutting sites but have different site preferences. ADAM10 primarily cleaves at I79 to produce the predominant C79, and ADAM17 preferably cleaves at L96 to produce the weaker C96. C96 is subsequently processed by α-secretase into C79. γ-secretase further cleaves C79 at L53 within the transmembrane domain to generate CICD. All the results are expressed as mean ± SEM

Next, we examined the cleavage in ADAM10 overexpressing cells. While TAPI-1 alone did not affect the C79 level (1.03 ± 0.06 folds), TAPI-1 and FI-2 together significantly reduced the C79 level to 0.70 ± 0.08 fold, compared with control (p = 0.0103) and TAPI-1 treatment (p = 0.0058), suggesting that C79 was subsequently cleaved from CTFf. C96 was increased by TAPI-1 treatment to 1.81 ± 0.32 folds (p = 0.1436) though not significant, suggesting that ADAM10-generated C96 was further cleaved by ADAM17 (Fig. 3a, b, CNT + ADAM10 graphs). Consistently, TAPI-1 marginally increased CTFf to 1.22 ± 0.10 folds (p = 0.1251) in furin-transfected cells, while FI-2 significantly decreased CTFf with (0.49 ± 0.05 fold, p = 0.0017) and without TAPI-1 (0.55 ± 0.04 fold, p = 0.0036) (Fig. 3c, d), supporting that CTFf is generated by furin (Figs. 3c, d and 1e, f). In addition, TAPI-1 and FI-2 decreased C79 to 0.55 ± 0.09 fold (p = 0.0119) and 0.73 ± 0.11 fold (p = 0.1326) respectively, and the combination of TAPI-1 and FI-2 treatment further reduced C79 level to 0.44 ± 0.06 fold (p = 0.0034), suggesting the furin-ADAM17 sequential cleavage of CNTNAP2 (Fig. 3c, d). GI254023X is a potent ADAM10 inhibitor with a higher selectivity for ADAM10 over ADAM17. We expressed ADAM10 or ADAM17 in ADAM17-KO or ADAM10-KO cells, and then treated the cells with 0.1–10 μM GI254023X. 5 μM GI254023X inhibited ADAM10 cleavage (reduced C79 to 21.75 ± 7.58%, p < 0.0001; reduced C96 to 8.19 ± 3.44%, p < 0.0001, increased CNT-FL to 169.9 ± 9.20%, p = 0.0238), but not ADAM17 cleavage (brought C79 to 70.11 ± 10.25%, C96 to 106 ± 18.21%, CNT-FL to 112.5 ± 4.79%) of CNTNAP2 (Supplementary Fig. 4a, b). ADAM17-generated C96 was increased by 5 μM GI254023X treatment to 144 ± 22.68% (p = 0.1576), suggesting that C96 is further cleaved by ADAM10 (Supplementary Fig. 4c, d).

Taken together, these data indicated that CNTNAP2 undergoes complex sequential cleavages by furin and two α-secretases: ADAM10 and ADAM17. ADAM10 and ADAM17 share the same α-cleavage sites at L96 and I79 but show different site preferences, with ADAM17 predominantly at L96 and ADAM10 primarily at I79, resulting in C79 as the main CTF cleavage product of CNTNAP2 through sequential cleavages (Fig. 3g).

Next, we determined the relationship between α- and γ-secretase cleavages. In a separate study, we discovered that CNTNAP2 is cleaved by γ-secretase to generate the CNTNAP2 intracellular domain (CICD), and CNTNAP2-CICD expression improves ASD-related behaviors.31 L-685,458, a potent γ-secretase inhibitor (GSI), was applied to treat HEK cells co-transfected with CNTNAP2 and empty vector, ADAM10 or ADAM17. L-685,458 treatment significantly increased C79 levels in the empty vector (1.34 ± 0.06 folds, p = 0.0043), ADAM10 (1.97 ± 0.12 folds, p = 0.0013), and ADAM17 (1.78 ± 0.22 folds, p = 0.0227) overexpressing cells, indicating that the main α-cleavage product C79 is further cleaved by γ-secretase (Fig. 3e, f). Therefore, α-secretase cleavage is necessary for the subsequent γ-secretase cleavage to generate CNTNAP2-CICD (Fig. 3g).

ASD-associated CNTNAP2 mutations impair the α-secretase processing of CNTNAP2

Point mutations in CNTNAP2 have been identified in ASD patients (Fig. 4a),2 however, the mechanism underlying the role of CNTNAP2 mutations in ASD pathogenesis is not clear. We discovered that I1253T at the α-secretase site I1253 (C-terminal I79 residue) inhibited the α-cleavage of CNTNAP2 to generate C79 (Fig. 2d, e, I79T corresponding to I1253T). Consistent with a previous report,18 we found that the mature/immature ratio of several mutant CNTNAP2 proteins differed from that of WT (WT, 100%; Y716C, 58.3 ± 5.6%, p = 0.0027; I869T, 54.86 ± 8.29%, p = 0.0012; R1119H, 36.95 ± 2.87%, p < 0.0001; D1129H, 13.78 ± 3.3%, p < 0.0001) (Fig. 4b, c). These mutations, particularly I869T (57.48 ± 7.35%, p = 0.0023), R906H (29.96 ± 4.66%, p < 0.0001), R1119H (31.47 ± 6.85%, p < 0.0001), and D1129H (13.34 ± 0.66%, p < 0.0001), reduced C79 levels (Fig. 4b, c). In the WT, C79 was increased to 2.64 ± 0.17 folds by ADAM10 or 3.88 ± 0.72 folds by ADAM17, and C96 was elevated to 11.87 ± 2.43 folds by ADAM10 or 26.92 ± 2.94 folds by ADAM17 (Fig. 4d, e). These pathogenic mutations generally affected α-secretase cleavage, especially the ADAM10 cleavage, with I869T, R1119H, and D1129H showing stronger inhibition on α-secretase cleavage. In R1119H, for example, C79 level was changed from 0.57 ± 0.20 in control to 0.43 ± 0.10 fold by ADAM10 and to 0.57 ± 0.17 fold by ADAM17, and C96 level was altered from 0.74 ± 0.49 to 0.39 ± 0.06 fold by ADAM10 and to 2.49 ± 0.86 folds by ADAM17, compared with the level of control in WT (Fig. 4d, e). Our results clearly demonstrate that the ASD-associated pathogenic mutations impair the α-secretase processing of CNTNAP2, leading to the reduction of CNTNAP2 CTF products.

Fig. 4

Pathogenic mutations affect CNTNAP2 processing. a Schematic of CNTNAP2 protein and locations of pathogenic mutations found in ASD patients. Mutations predicted deleterious or at conserved sites are noted in red. ADAM10/17-mediated α-cleavage affected by the mutations was summarized in the brackets (10 = ADAM10, 17 = ADAM17; slight affections are noted in purple, while severe impacts are marked in red). SP signal peptide, DISC discoidin-like domain, L1-4 four laminin A-like G domains, FBG fibrinogen-like domain, E1-2 two epidermal growth factor (EGF)-like domains, TM transmembrane domain; 4.1b, 4.1 binding domain; PDZb, PSD-95/Discs large/zona occludens-1 (PDZ) binding domain. Braces represent three lobes of CNTNAP2 protein. The figure is to scale. b, c CNTNAP2 mutations affected expression patterns of CNTNAP2 full-length and CTFs. CNTNAP2 plasmids were co-transfected with GFP into HEK cells. n = 3 independent experiments, ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test, *p < 0.05; **p < 0.01; ****p < 0.0001. d, e Altered α-secretase cleavage in CNTNAP2 mutations. CNTNAP2 plasmids were co-transfected with empty vector (EV), ADAM10 (AD10), or ADAM17 (AD17) plasmids into HEK cells. Mutations showed variant cleavage patterns. Three blots show samples from the same experiment. The blots were processed in parallel and scanned together simultaneously. n = 3 independent experiments, two-way ANOVA followed by Dunnett’s multiple comparisons test (compare ADAM10 and ADAM17 with Control for each plasmid), row factor = mutation, column factor = ADAM10/17 overexpression. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. All the results are expressed as mean ± SEM

Cntnap2 -I1254T mice exhibit autism-like phenotypes and the C79 improves the phenotypes in vivo

To examine the effect of the essential α-secretase cleavage of CNTNAP2 on autism-like phenotypic behaviors, we generated mutant I1254T knock-in mice (Cntnap2-I1254T) using the CRISPR/Cas9 technique to express the I1254T mutant CNTNAP2 protein, corresponding to the pathogenic I1253T mutation in ASD patients (Fig. 5a). All mice were bred through crossing heterozygous Cntnap2-I1254T and further confirmed by genotype sequencing. The pups were born at a Mendelian ratio, and there was no gender difference of pups (Supplementary Table 2 and Supplementary Fig. 5). Cntnap2-I1254T mice grew as well as control mice during the 8 weeks of study period. Body weight was monitored from 3 to 8 weeks of age, and there was no difference in body weight between the mutant Cntnap2-I1254T mice and control WT mice (Fig. 5b).

Fig. 5

Impairment of α-secretase cleavage resulted in repetitive and social behavior abnormalities in mice. a Schematic representation of CRISPR/Cas9 knock-in method used to generate mutant I1254T mice by mutating “ATA” (Ile) to “ACG” (Thr) at the major α-cleavage site of CNTNAP2. b Body weight record of WT and I1254T mice. The weight of both mice was recorded daily, and no significant difference was found between mutants and WT mice (WT, n = 15; I1254T, n = 15). c Isolation-induced ultrasonic vocalizations (UsVs) test. The numbers of distress calls from infants of WT and mutants were detected at the age of P3, P6 and P12 (WT, n = 13; I1254T, n = 13). d Juvenile play test. Time involved in social interaction, as well as repetitive behaviors like grooming in WT and I1254T mutant mice were measured at the age of P21 when interacting with unfamiliar mice (n = 10 per group: male = 5, female = 5). e, f T maze and three chamber sociability tests performed at 7-week-old WT and mutants. I1254T mutation at the α-cleavage site increased repetitive behavior abnormalities in T maze (e) and decreased social interactions (f) in I1254T mutants, and C79 overexpression rescued the ASD-like behaviors (n = 12 per group: male = 6, female = 6). g–i Social and repetitive behavior tests in Cntnap2−/− mice. C79 overexpression rescued the social interactions (g, i) and repetitive behaviors (h) in Cntnap2−/− mice (n = 11: male = 6, female = 5 in WT and KO groups; n = 12: male = 6, female = 6 in KO + C79 group). Statistical significance was assessed by either one- or two-way ANOVA followed by Turkey or Bonferroni’s test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. All the results are expressed as mean ± SEM

The mice were tested with the isolation-induced ultrasonic vocalizations (UsVs) at the ages of 3, 6 and 12 days, in which the distress calls emitted by separated pups and the infant-mother vocal communications were considered as ASD-like behavior.3,32Cntnap2-I1254T mice emitted a significantly lower number of distress calls than WT mice at different ages (Fig. 5c). In the Juvenile Play test where the social interaction and repetitive behaviors between different pairs of unfamiliar mice was recorded, the Cntnap2-I1254T mice showed increased repetitive behaviors like grooming (Fig. 5d).

To confirm the autism-like behaviors was due to the impaired α-cleavage of CNTNAP2 at ADAM10 site, the adeno-associated virus (AAV) vector expressing CNTNAP2-C79 protein (AAV-C79) and AAV-EGFP (as the control) were generated and injected into the medial prefrontal cortex (mPFC) of the mutant (Cntnap2-I1254T) and CNTNAP2-KO mice (Cntnap2−/−) at the age of 3 weeks (Supplementary Fig. 6a). AAV-C79 injection significantly increased the C79 protein level in mPFC (Supplementary Fig. 6b). In an open-field test, Cntnap2-I1254T mice displayed locomotor activity similar to that of the WT mice (Supplementary Fig. 7a), while the Cntnap2−/− mice exhibited hyper locomotion compared with the WT mice (Supplementary Fig. 7b). In addition, Cntnap2-I1254T mice injected with AAV-EGFP displayed an increased number of no alterations, representing an ASD-like repetitive behavior (Fig. 5e). In the three-chamber social interaction test, all mice showed no preference for each chamber in the habituation phase (Supplementary Fig. 8a, b), and Cntnap2-I1254T mice did not show a significant social preference for strangers when compared to the highly social-interactive WT mice (Fig. 5f). However, the expression of C79 significantly reduced the number of no alterations of Cntnap2-I1254T mice in T maze test (Fig. 5e) and increased their social interactions with unfamiliar mice (Fig. 5f) at the age of 7 weeks, indicating that C79 expression improved autism-like phenotypes in the mutant mice. Similarly, C79 expression also rescued the deficiency in repetitive and social behaviors of Cntnap2−/− mice (Fig. 5g–i). Taken together, these data demonstrate that the mutation of α-cleavage site in CNTNAP2 leads to autism-like repetitive and social behavior abnormalities, and restoring α-cleavage product C79 expression improves autism-like phenotypes.