Inhibiting proBDNF to mature BDNF conversion leads to autism-like phenotypes in vivo

Autism spectrum disorders (ASD) comprise a range of early age-onset neurodevelopment disorders with genetic heterogeneity. Most ASD related genes are involved in synaptic function, which is oppositely regulated by brain-derived neurotrophic factor (BDNF): the precursor proBDNF inhibits while mature BDNF (mBDNF) potentiates synapses. Here we generated a knock-in mouse line (BDNFmet/leu) in which the conversion of proBDNF to mBDNF is inhibited. Biochemical experiments revealed residual mBDNF but excessive proBDNF in the brain. Similar to other ASD mouse models, the BDNFmet/leu mice showed decreased brain volumes, reduced dendritic arborization, altered spines, and impaired synaptic transmission and plasticity. They also exhibited ASD-like phenotypes, including stereotypical behaviors, deficits in social interaction, hyperactivity, and elevated stress response. Interestingly, the plasma level of proBDNF, but not mBDNF, was significantly elevated in ASD patients. These results suggest that proBDNF level, but not Bdnf gene, is associated with autism-spectrum behaviors, and identify a potential blood marker and therapeutic target for ASD.


60
A key regulator for synaptic function is brain-derived neurotrophic factor (BDNF) ( these in vitro studies have led to the "Yin-Yang" hypothesis that (i) BDNF and proBDNF 95 elicit opposing biological functions and (ii) the conversion of proBDNF to mBDNF, or the 96 ratio of proBDNF and mBDNF may determine the functional outcomes (Lu et al., 2005b). 97 98 A number of key questions remain: what is the physiological role of proBDNF? How 99 important is the cleavage of proBDNF? A cleavage-resistant proBDNF knock-in mouse line 100 was developed by mutating the two key arginine residues ( Figure 1A, 127R and 128R) at the 101 cleavage site of proBDNF (Yang et al., 2014). The homozygous mice were not viable, 102 possibly due to a complete lack of mBDNF, in addition to the strong apoptotic effects of 103 dramatically elevated levels of un-cleavable proBDNF in the brain. The heterozygous 104 (probdnf-HA/+) mice, which have one copy of un-cleavable proBDNF, displayed a reduced 105 dendritic complexity and impaired LTP similar to BDNF heterozygous mutant (BDNF+/-) 106 (Yang et al., 2014). In addition, the probdnf-HA/+ showed a reduced spine density, a 107 decreased basal synaptic transmission, and an elevated LTD, which were not observed in 108 BDNF+/-mice (Woo et al., 2005). Because the probdnf-HA/+ mice expressed a high level of 109 un-cleavable proBDNF but its mBDNF level was reduced by half, it is unclear whether the 110 phenotypes observed were due to an increase in proBDNF or a decrease in mBDNF, or both. 111 Moreover, the behaviors of the the probdnf-HA/+ were not examined.

113
In the present study, we generated a proBDNF knock-in mouse line named BDNF met/leu , in 114 which the arginine residues R125 and R127 at the cleavage site were converted to methionine 115 (R125M) and leucine (R127L) (rs1048220 and rs1048221, respectively), based on previously 116 reported human SNPs (H. Koshimizu et al., 2009). In vitro biochemical experiments 117 demonstrated that these two mutations, either alone or together, markedly inhibit the cleavage 118 of proBDNF (M. Koshimizu et al., 2015). The BDNF met/leu mice expressed high levels of 119 proBDNF but low levels of mBDNF, compared with their wild type (WT) littermates (H. 120 Koshimizu et al., 2009). Unlike the homozygotes of proBDNF knock-in mice (probdnf-121 HA/HA) which died at birth, the homozygous BDNF met/leu mice survived until adulthood 122 (Kojima et al., 2020). This allowed us to perform detailed characterization of the adult 123 BDNF met/leu mice. Unexpectedly, this new line of mutant mice exhibited many morphological, 124 physiological, and behavioral phenotypes often seen in ASD patients. Further, preliminary 125 characterization of ASD patients revealed a significant increase in the plasma level of 126 proBDNF. Our results point to a possible non-genetic mechanism of ASD and suggest a 127 blood marker potentially useful in the clinic. 128 Generation of BDNF met/leu mice predominantly expressing proBDNF 130 We generated a knock-in mouse line in which the endogenous Bdnf allele was replaced with 131 proBDNF containing human SNPs that changed two arginines proximal to the cleavage site 132 to methionine (R125M) and leucine (R127L), respectively ( Figure 1A). We previously 133 reported that these amino acid changes led to inefficient conversion of proBDNF to mature 134 BDNF in cultured neurons (H. Koshimizu et al., 2009). Immunoblot analyses using a 135 polyclonal BDNF antibody (N-20, Santa Cruz) demonstrated that hippocampal lysates from 136 8-week-old homozygous mutant (BDNF met/leu ) mice contained an excess amount of proBDNF 137 and only a residual amount of mBDNF, whereas those from wild-type (WT) littermates (or 138 BDNF +/+ ) contained more mBDNF than proBDNF ( Figure 1B). This opposite ratio of 139 mBDNF and proBDNF was quantitatively confirmed by Western blot analysis using a 140 monoclonal anti-pan-BDNF antibody (clone; 3C11, Icosagen) (Kojima et al., 2020) ( Figure  141 1C). A band with the same size as proBDNF was also observed via immunoblotting using 142 another previously-generated proBDNF-specific antibody ), further 143 validating that this was indeed proBDNF ( Figure 1D). Taken together, we generated a 144 proBDNF knock-in mouse line that expresses a high level of proBDNF with a residual 145 amount of mBDNF.

147
Despite insufficiency in the proBDNF cleavage, the BDNF met/leu homozygous mice were born 148 in a Mendelian manner, and approximately 95% of the homozygous mutant mice survived to 149 adulthood (Kojima et al., 2020). We therefore performed all subsequent experiments using 150 BDNF met/leu homozygous, but not heterozygous, mice. The survival of the BDNF met/leu mice to 151 adulthood was unexpected because the homozygous proBDNF knock-in mice with different 152 mutations at the proBDNF cleavage site (RVRR→RVAA) died shortly after birth, as reported 153 previously (Yang et al., 2014). Unlike the proBDNF knock-in mice, the morphology and 154 TrkB activation of the hearts of BDNF met/leu mice were normal (Kojima et al., 2020). These 155 morphological and immunohistochemical results suggest that the inefficient conversion of 156 proBDNF into mBDNF due to R125M/ R127L mutation may not influence the survival of 157 animals and/or development of the heart. 158 159 Brain volumes and dendritic complexity in BDNF met/leu mice 160 The brains of the BDNF met/leu mice were smaller than their WT counterparts ( Figure 2A). The 161 mean wet weight of the BDNF met/leu brains was 15.1 ± 1.3% lower than that of the BDNF +/+ 162 brains ( Figure 2B). Cavalieri analysis of Nissl-stained sections revealed that the whole brain 163 volume of the BDNF met/leu group was 17.7 ± 1.56% lower than that of the BDNF +/+ group 164 ( Figure 2C). Similarly, the volume of the BDNF met/leu hippocampus and cortex was reduced 165 by 14.0 ± 2.3% and 15.8 ± 1.5%, whereas the reduction in olfactory bulb was not significant, 166 compared with the BDNF +/+ group ( Figure 2C). Thus, inefficient cleavage of proBDNF 167 affects the volume of the adult brain, particularly in the hippocampus and cerebral cortex, two 168 areas known to express high levels of BDNF mRNA (Timmusk et al., 1993). 169 Since proBDNF is highly expressed in dentate gyrus (DG) (Zhou et al., 2004) the dendritic 171 complexity of the DG neurons was analyzed using Golgi staining. The BDNF met/leu 172 homozygous mice had a marked decrease in dendritic arbor complexity compared with their 173 BDNF +/+ littermates ( Figure 2D, left panel). The difference became evident at 70 μm from 174 cell body ( Figure 2D, right panel). Nevertheless, there was no significant difference between 175 the diameters of the cell bodies in the BDNF met/leu and WT animals (data not shown). No 176 apparent apoptotic cells were detected in the DG regions of brains from each group ( Figure  177 2E). Thus, inhibition of proBDNF cleavage may lead to impairment of dendritic complexity 178 but not neuronal survival at least in hippocampal DG region in animals.

180
The reduction in brain volume was also reported in mouse models with autism-like 181 phenotypes ( Dendritic spine changes in BDNF met/leu mice 189 Next, we performed a detailed analysis of the dendritic spine morphologies and density in 190 BDNF met/leu hippocampus. The maximal length, maximum width, and density of spine 191 protrusions in the secondary dendritic segments of hippocampal pyramidal neurons in the 192 stratum radiatum of the CA1 area were determined ( Figure 3A). We found no significant 193 difference in the mean protrusion lengths between the BDNF met/leu and WT groups (P = 0.25) 194 ( Figure 3B). However, the mean protrusion width of the BDNF met/leu mice was significantly 195 smaller than that of the WT mice ( Figure 3C).

197
A previous report using heterozygous probdnf-HA/+ mice showed a marked decrease in total 198 spine density, without detailed classification of different types of spines (Yang et al., 2014). It 199 was also unclear whether the decrease in spine density was due to an increase in proBDNF or 200 a decrease in mBDNF. In contrast, we observed a significant increase in total spine density 201 ( Figure 3D). According to previous studies, the spine protrusions could be classified as thin (a 202 protrusion with length to width ratio >2) or mushroom-like ones (Bourne & Harris, 2008;203 Harris, Jensen, & Tsao, 1992; Matsuzaki et al., 2001). Detailed quantitative analyses of CA1 204 neuron dendrites revealed a small but significant decrease in mushroom-like spines ( Figure  205 3E), but a more than 2-fold increase in thin spines ( Figure 3F), in the BDNF met/leu mice. Recent studies suggest that some of the biological effects of proBDNF require its interaction 217 with p75 NTR and the sortilin family proteins such as SorCS2 and SorCS3 (Glerup et al., 218 2016). SorCS3 is expressed at a high level in the hippocampal CA1 region and is localized to 219 the postsynaptic density (PSD), and the loss of SorCS3 in mice leads to the impairment of 220 hippocampal LTD (Breiderhoff et al., 2013). We therefore investigated whether the proBDNF 221 signaling machinery is intact in the BDNF met/leu hippocampus. We found that the expression 222 levels of p75 NTR , SorCS2, and SorCS3 in the hippocampus were not affected by the 223 BDNF met/leu mutation ( Figure 3-supplement 1). Thus, proBDNF signaling through the p75 NTR -224 SorCS2 and/or SorCS3 receptor complex occurs normally in the BDNF met/leu mice.

226
The deficits in postsynaptic scaffold proteins have been reported in several ASD mouse of PSD-95 and SynGAP were not affected by the BDNF met/leu mutation ( Figure 3G). 232 However, immunoprecipitation studies revealed that the interaction of PSD-95 and SynGAP 233 was attenuated markedly in the lysates prepared from 9-week-old BDNF met/leu mice, 234 compared with the WT control ( Figure 3H).

236
proBDNF has been shown to promote hippocampal LTD by regulating the expression of the 237 GluN2B subunit of the N-methyl-D-aspartate receptor (NMDAR) (Woo et al., 2005). The 238 expression of hippocampal GluN2B were comparable between WT and BDNF met/leu animals 239 ( Figure 3I). Interestingly, the level of GluN2B was significantly higher in the crude 240 synaptosomal fraction (P2) prepared from BDNF met/leu hippocampi ( Figure 3J and K). In 241 contrast, its interaction with PSD-95 was slightly decreased ( Figure 3L). However, the levels 242 of PSD-95, synapsin I, or beta-actin in the same P2 fraction were not altered ( Figure 3J  as ASD models..

250
Impaired synaptic transmission and plasticity 251 We further examined whether deficits in dendritic spines would alter synaptic function in 252 Schaffer collateral-CA1 pyramidal neurons (Mizui et al., 2015). The slope of the input/output 253 (I/O) curve was significantly reduced, suggesting an impairment in basal synaptic function in 254 BDNF met/leu synapses ( Figure 4A). However, paired-pulse facilitation (PPF) was not altered in 255 CA1 synapses of the BDNF met/leu mice, indicating normal presynaptic function ( Figure 4B). 256 The decrease in mushroom spines but normal PPF suggest that the reduction in basal synaptic 257 transmission could be mediated by post-synaptic rather than pre-synaptic mechanisms.

259
We next examined long-term synaptic plasticity in BDNF met/leu mice. Previous studies have 260 shown that the application of exogenous proBDNF enhanced LTD in juvenile (5-week-old) 261 hippocampal slices (Woo et al., 2005). LTD was elevated in 3-week-old hippocampal slices 262 from the previously reported in the heterozygous probdnf-HA/+ mice, which have half of its 263 mBDNF but elevated proBDNF (Yang et al., 2014). We tested LTD in hippocampal slices 264 from 3-week-old BDNF met/leu mice which predominantly express proBDNF with residual 265 mBDNF ( Figure 1C). The Low-frequency stimulation (LFS; 1 Hz, 900 pulses, 15 min) was 266 applied to Schaffer collaterals of hippocampal slices. In marked contrast to what was 267 observed in probdnf-HA/+ mice, both induction and expression of LTD was severely 268 impaired in BDNF met/leu mice ( Figure 4D). We further examined whether hippocampal LTP 269 was altered in BDNF met/leu mice. Two-month-old hippocampal slices were used, field EPSP 270 slopes at CA1 synapses were recorded, and high-frequency stimulation (HFS, 100Hz, 1 271 second) was applied. Although control mice expressed normal LTP ( Figure 4C), the 272 BDNF met/leu mice failed to express LTP ( Figure 4C). Taken together, it appears that 273 BDNF met/leu mice, similar to most ASD models ( While it is difficult to investigate language problems in animals, several mouse ASD models 285 display common autistic behaviors such as hyperactivity, repetitive and stereotyped 286 behaviors, and anxiety (De La Torre-Ubieta et al., 2016). We therefore investigate whether 287 BDNF met/leu mice have some of the ASD-related behavioral phenotypes.

289
An obvious repetitive and stereotypical behavior seen in the BDNF met/leu mice was 290 "stargazing." In home cages, BDNF met/leu mice displayed repeated head-tossing very similar 291 to that seen in the classic "stargazer" mouse ( Figure 5A, Movie#1) (L. Chen et al., 2000).

292
Quantitative analyses showed that the mutant mice underwent stargazing over 50 times per 293 min ( Figure 5B). Such behavior was hardly seen in WT mice.

295
Next, we performed two tests to assess social interaction by BDNF met/leu mice. First, a three-296 chamber test was performed to examine whether a BDNF met/leu mouse placed in the middle 297 chamber prefers a stranger mouse (in one side chamber) to a novel object (in the opposite 298 side chamber). The WT mice spent twice more time with a mouse than with an object ( Figure  299 5B). In contrast, the BDNF met/leu mice showed no preference to social partner, spending equal 300 amount of time interacting with the mouse partner and the object ( Figure 5B). Second, we 301 measured the interaction of a BDNF met/leu mouse with their littermates in the home cage. A 302 mouse generally interacts less with an aggressive littermate than with a non-aggressive one. 303 We therefore measured the interactions with the two types of littermates separately.

304
BDNF met/leu mice displayed dramatically reduced interaction time with both aggressive and 305 non-aggressive littermates, compared with the WT mice ( Figure 5C). These results indicated 306 that social interaction was impaired in BDNF met/leu mice. 307 308 Hyperactivity and elevated stress response 309 As most ASD mice models are reportedly hyperactive (De La Torre-Ubieta et al., 2016), we 310 tested locomotion and motor ability of BDNF met/leu mice. In an open field test, the BDNF met/leu 311 mice traveled much longer distances, approximately 1000cm/min, than their WT littermates 312 ( Figure 5D, left). They also traveled mostly in the center of the field, spending approximately 313 40sec/min there ( Figure 5D, center). There was no difference in their vertical activities 314 between the BDNF met/leu mice and their WT littermates ( Figure 5D, right). Interestingly, the 315 hyperactivity phenotype appeared to be more apparent during the nighttime when the mice 316 are mostly awake ( Figure 5-supplement 1).

318
While hyperactive, ASD patients often become restrictive and immobile when challenged or 319 stressed (De La Torre-Ubieta et al., 2016). Tail-suspension test (TST) is often used to test 320 stress-induced immobile behavior in mice. The BDNF met/leu mice (10-12-week old) showed 321 significantly longer immobility time in TST than their WT littermates ( Figure 6A). However, 322 unlike in WTs, the immobility time was not shortened by fluoxetine (20 mg/kg, 30 min, i.p.), 323 a selective serotonin reuptake inhibitor commonly used as an antidepressant (Crowley,324 Blendy, & Lucki, 2005)( Figure 6A). To determine whether prolonged immobility in TST is 325 associated with stress, we determined blood concentrations of corticosterone, an adrenal 326 steroid known to increase in response to stress. Blood samples were collected from the tail 327 vein of BDNF met/leu and WT littermates reared in the same cage for 2 weeks. As expected, the 328 blood corticosterone concentration in resting time in WT mice was comparable to that 329 reported previously (Dugovic, Maccari, Weibel, Turek, & Van Reeth, 1999). In contrast, the 330 BDNF met/leu mice demonstrated a 3-fold higher blood corticosterone concentration ( Figure  331 6B, Before stress), suggesting that BDNF met/leu animals are severely stressed compared with 332 WT littermates reared in the same cage. However, 60 min after the exposure to 333 immobilization stress, both BDNF met/leu and WT mice exhibited similar, elevated levels of 334 corticosterone (Dugovic et al., 1999), ( Figure 6B, right, After stress). Consistent with these 335 results, in an unstressed home cage environment, the net weight of the adrenal glands, which 336 secrete corticosterone (Jankord & Herman, 2008), was significantly higher in BDNF met/leu 337 than that in WT mice ( Figure 6C, Adrenal gland/Body). As a negative control, the weight of 338 the kidney did not differ between the BDNF met/leu and WT animals (( Figure 6C, 339 Kidney/Body). These results suggest that mice deficient in proBDNF-processing are sensitive 340 to stress and exhibit activation of the HPA axis under normal conditions. 341 342 343 Plasma proBDNF and BDNF levels in ASD patients and controls 344 A number of previous studies have suggested that blood (serum) level of BDNF is 345 proportional to its level in the brain (Klein et al., 2011). Given that the BDNF met/leu mice 346 expressed excessive proBDNF in the brain, we determined whether blood proBDNF 347 concentration was also elevated. However, blood platelets contain a large quantity of BDNF We therefore turned to examine the levels of mBDNF and proBDNF in the plasma of human 362 ASD patients and healthy controls. We obtained 1 ml of blood from human volunteers 363 (controls), 100 times the amount from mice. In the initial, pilot experiments, 9 ASD patients 364 and 10 healthy volunteers were enrolled, with well-documented consent forms and the 365 approval of the Institutional Ethics Board. The age of participants ranged from 3~13 years: 7 366 boys and 3 girls in the ASD group; 1 boy and 10 girls in the control group (Table 1). Among 367 those 10 ASD patients, all had delayed development of speech, and most showed repetitive 368 behaviors, a common feature in ASD diagnosis. We also noticed that the intellectual abilities 369 of at least 8 out of 10 ASD children were also impaired, suggesting that these ASD patients 370 likely belonged to the traditional ASD, but not some special groups of ASD with normal 371 intellectual ability, such as Asperger's syndrome. 372 373 Next, peripheral venous blood, from patients and controls. was collected in EDTA-coated 374 tubes and centrifuged within 5 min to obtain plasma (see Methods). Using the sandwich 375 ELISA specific for proBDNF and mBDNF, respectively, we found a selective increase in 376 proBDNF, but not mBDNF, in the ASD patients ( Figure 7, and Table 6). The concentration of 377 proBDNF was 1316.17 pg/ml, whereas that of mBDNF was 946.27 pg/ml in ASD patients. In 378 healthy controls, the concentration of proBDNF was 688.39 pg/ml, nearly half of that in the 379 ASD group. However, the concentration of mBDNF is 713.12 pg/ml, which was comparable 380 to that in the ASD group. Thus, the increase in plasma proBDNF, but not mBDNF, is 381 associated with ASD. 382 383 Furthermore, we observed, evaluated, and documented all related symptoms of these ASD 384 patients (Table 2~5). 8 out of 10 patients were firmly diagnosed with ASD, while in the 385 remaining two, autistic features were clearly found, but these features did not match the 386 criteria of ASD diagnosis (Table 2). All patients exhibited conventional autistic features, 387 including the delay of speech development (Table 2), intellectual disabilities (Table 2), 388 repetitive behavior (Table 4), and ADHD (Table 4). However, in some of these ASD patients, 389 we also observed unconventional symptoms, including delays in motor development ( Table  390 2), sleep disturbances (Table 3), anxiety (Table 4), self-injury (Table 4), obsessive behavior 391 (Table 4), and aggressive behavior (Table 4). Interestingly, we found that the proBDNF data 392 of ASD patients was associated with a few of these unconventional symptoms. reductions in synaptic density, synaptic protein synthesis, or synaptic plasticity. In this study, 408 we took an unconventional approach, focusing not on genetic mutations but on the processing 409 of BDNF, a neurotrophic protein known to play a critical role in synaptic function. We have 410 developed an ASD-like mouse model in which synaptic function is decreased, through 411 elevating the proBDNF level but reducing mBDNF level. To the best of our knowledge, this 412 is the first ASD mouse model not based on genetic mutations, but on protein processing.

414
Due to the lack of tools to non-invasively detect cellular changes in human brain, we do not 415 know the neuronal phenotypes in the brains of ASD patients. However, studies using 416 postmortem ASD brain tissues have revealed increased spine density in many brain regions, 417 such as the frontal, parietal and temporal lobes ( behavior observed in some ASD models but not in human ASD. Bead-burying, a repetitive 445 behavior seen in some ASD models but not in ASD children, is also absent in our BDNF met/leu 446 mice. Thus, the BDNF met/leu mice could be a different and perhaps better ASD model, with 447 regards to 'self-stimulating' behaviors. Regardless, the robust behavioral phenotype of the 448 BDNF met/leu mice provides a unique opportunity for mechanistic studies and drug testing.

450
Finally, our preliminary data indicates that the plasma level of proBDNF could be used as a 451 potential ASD biomarker that may help in monitoring disease progression and drug efficacy. 452 Specifically, in the pilot experiment with 9 ASD patients and 10 healthy volunteers, we found 453 that an increase in plasma proBDNF level, but not mBDNF level, is associated with ASD. 454 However, a biomarker for clinical use requires both good sensitivity and specificity (Brower,455 2011). Our data (Figure 7 and Table 6) indicates that the concentration of proBDNF in the 456 ASD group is twice of that in healthy controls, suggesting that the sensitivity of the current 457 ELISA might be in the right range. Technologies exist to further improve the sensitivity 458 greatly. Given the small sample size, we cannot claim the specific association of elevated 459 proBDNF with ASD. Further studies with much larger sample sizes are necessary to validate 460 the specificity of this potential biomarker and to investigate whether it is associated with a 461 distinct ASD subtype. Previous studies have suggested that the plasma level of BDNF could 462 reflect BDNF concentration in the brain (Klein et al., 2011). It is unclear whether the plasma 463 level of proBDNF also correlates with proBDNF concentration in the brain. The source and 464 the functional implication of plasma proBDNF should also be investigated in the future.

466
Although most of the phenotypes found in the BDNF met/leu mice coincide with the symptoms 467 of ASD patients or neuronal features of the other ASD models, the brain volume changes may 468 not be the same as those found in human studies. In contrast to the decrease of brain volume 469 in the BDNF met/leu mice, previous studies with large cohorts have shown brain overgrowth in 470 ASD children (De La Torre-Ubieta et al., 2016). However, no such brain overgrowth was 471 reliably detected in the adult ASD patients. In addition, neuroimaging analysis of ASD brains 472 also showed decreased volume of the corpus callosum, cerebellum and brainstem (Penn,473 2006). These clinical observations suggest that brain volume may not be used as a general 474 pathological criterion for ASD. Interestingly, mutations of the X-linked genes encoding 475 neuroligins, which are known for their association with ASD, have been shown to lead to 476 decreased brain volume (Jamain et al., 2003). Moreover, both neuroligins-3 and -4 knock-out 477 mice, which could display all ASD-like phenotypes, have reduced brain volume ( and thus results in a decrease rather than an increase in LTD. This may explain the difference 496 in LTD phenotype between our BDNF met/leu mice and the probndfHA heterozygous mice.

498
In summary, we have developed a new ASD model, not based on genetic mutations, but on 499 protein processing, and provided a link between ASD and BDNF, a key factor for synaptic 500 regulation. With its robust phenotypes, the BDNF met/leu mouse line may serve as a new ASD 501 model with more comprehensive behavioral deficits resembling human autism patients. 502 Finally, our preliminary study using human blood samples raises the possibility of using 503 plasma proBDNF level as an ASD biomarker. 504

Animals 506
All animal experiments were performed in strict accordance with protocols that were 507 approved by the Institutional Animal Care and Use Committee of AIST. All efforts were 508 made to minimize animal suffering during the experiments. Animals 509 All animal experiments were performed in strict accordance with protocols that were 510 approved by the Institutional Animal Care and Use Committee of AIST (number.2019-084, 511 approval date 6 June 2015). All efforts were made to minimize animal suffering during the 512 experiments. Mice were housed with access to food and water ad libitum at constant room 513 temperature (24 ± 2°C) and were exposed to a 12 h light/dark cycle (lights on between 7 a.m. 514 and 7 p.m.). All mice were housed in groups of 3-4 per cage. 515 516 Generation of proBDNF knock-in mouse line 517 Generation of proBDNF knock-in mouse line was described in the report of Kojima et al.

518
(2020) (Kojima et al., 2020). Briefly, the proBDNF knock-in mouse line was generated by 519 replacing the endogenous Bdnf allele with a cDNA encoding cleavage-resistant form of 520 BDNF that contained two amino acid substitutions proximal to the cleavage site of proBDNF 521 ( Figure 1A, RVRR to MVLR). A 6.0 kb long arm fragment (NCBI accession number: 522 AY057907; 49015-54460) and a 3.4 kb short arm fragment (55207-58754) flanking the 5' 523 and 3' ends of mouse Bdnf exon 5, respectively, were amplified from 129SV mouse genomic 524 DNA using PCR. The long and short arm fragments were introduced into the ClaI-NotI and 525 PacI-AscI sites of the pMulti-ND 1.0 vector, respectively. A DNA fragment encoding mutant 526 proBDNF was then inserted into the ClaI-NotI site of the targeting vector. A pGK-thymidine 527 kinase gene was used as a negative selectable marker ( Figure 1A, Vector). Linearized 528 targeting vector was electroporated into embryonic stem (ES) cells of the D3 line (strain 129 529 SV). DNA derived from G418-resistant ES clones was screened using PCR and positive ES 530 clones were injected into blastocysts obtained from C57BL/6 mice. The injected blastocysts 531 were then introduced into the uteri of pseudo-pregnant females. Chimeric mice were mated 532 with C57BL/6 mice to produce heterozygotes, and these mice were subsequently crossed 533 with mice expressing Cre recombinase in germ cells to excise the neo cassette.

534
A genetic backcross to the 129/SvEv background were performed over 10 generations before 535 the animals were used in behavioral analyses and all other analyses. 536 537 Genotyping 538 Genotyping of the proBDNF knock-in mouse line was performed according to our recent 539 report (Kojima et al., 2020). using the following primers: 5'-TGCACCACCAACTGCTTAG-540 3' and 5'-GGATGCAGGGATGATGTTC-3'. PCR analyses using these primers generated 550 541 bp and 320 bp DNA fragments from wild-type and mutant alleles, respectively. Genotyping 542 of the p75NTR knock-out mice (Ngfrtm1Jae) from the Jackson Laboratory (Bar Harbor, ME) 543 was performed using the following primers: 5'-GCTCAGGACTCGTGTTCTCC-3', 5'-544 CCAAAGAAGGAATTGGTGGA-3', and 5'-TGGATGTGGAATGTGTGCGAG-3'. PCR 545 analyses using these primers generated 386 bp and 193 bp DNA fragments from wild-type 546 and mutant alleles, respectively. Genotyping of the BDNF knock-out mice (kindly provided 547 by Prof. Nakamura, Tokyo University of Agriculture and Technology, Tokyo) was performed 548 using the following primers: 5'-ATGAAAGAAGTAAACGTCCAC-3', 5'-549 CCAGCAGAAAGAGTAGAGGAG-3', and 5'-GGGAACTTCCTGACTAGGGG-3'. PCR 550 analyses using these primers generated 275 bp and 340 bp DNA fragments from wild-type 551 and mutant alleles, respectively. Wild-type mice with the background of 129S6/SvEv and 552 C57BL6/J were obtained from Taconic Biosciences (Hudson, NY) and Crea Japan (Shizuoka, 553 Japan), respectively. 554 555 Golgi impregnation of granule neurons in the hippocampal dentate gyrus 556 Golgi impregnation of mouse brains was performed using the FD Rapid GolgiStain Kit. Cavalieri analysis of Nissl-stained sections were used for hippocampal volume estimation. 571 Steroinvestigator software was applied to measure the entire volume of the hippocampus at 572 4X objective magnification. The external capsule, alveus of hippocampus, and white matter 573 were used as boundary landmarks. All sections throughout each hippocampus were traced 574 and reconstructed. The Cavalieri estimator function was used to calculate the volume of each 575 hippocampus. Following total hippocampal measurements, the cellular layer of each 576 subregion of the hippocampus (DG, CA1, CA2/3) was traced separately and analyzed in the 577 same manner.

602
In situ hybridization 603 Brains from 8-week-old mice were fixed by transcardial perfusion with PBS followed by 4% 604 paraformaldehyde in PBS at 4°C. The brains were immersed in the fixative overnight at 4°C, 605 treated with 20% sucrose in PBS, embedded in Tissue-Tek Optimal Cutting Temperature 606 Compound (Sakura Finetek, Tokyo, Japan), and then frozen on a block of dry-ice. In situ 607 hybridization was performed on 20 µm coronal sections. The hybridization and post-608 hybridization wash steps were performed at 65°C to avoid nonspecific binding. Stained 609 sections were photographed under the bright field system of a BioRevo microscope 610 (Keyence, Osaka, Japan).

612
Immunoblotting 613 Immunoblotting was performed according to our previous report (Suzuki et al., 2007). Briefly, 614 hippocampal tissues were homogenized in cold lysis buffer comprising 50 mM Tris-HCl (pH 615 7.4), 1 mM EDTA, 150 mM NaCl, 10 mM NaF, 1 mM Na3VO4, 1% Triton X-100, 10 mM 616 Na2P2O7, 100 µM phenylarsine oxide, and 1% protease inhibitor cocktail (Complete Mini, 617 Roche Diagnostics, Hertforshire, UK). The lysed tissues were incubated at 4°C for 20 min 618 and then centrifuged at 15,000 rpm for 15 min. The protein levels in the supernatants were 619 determined using the BCA Protein Assay Kit (Pierce/Thermo Scientific, Rockford, IL). To 620 detect BDNF, the lysates were boiled for 5 min at 100°C, separated on SDS-polyacrylamide 621 gels, and then transferred to polyvinylidene fluoride membranes (Immobilon P; Millipore, 622 Billerica, MA). The membranes were blocked in Tris-buffered saline containing 0.2% Tween-623 20 (TBST) and 5% BSA or Block Ace (Dainippon Pharmaceutical, Osaka, Japan), incubated 624 with a rabbit polyclonal anti-BDNF antibody (N-20, Santa Cruz Biotechnology, San Diego, 625 CA) in TBST containing 0.5% BSA or 1% Block Ace at room temperature for 90 min, and 626 then washed three times with TBST. Subsequently, the membranes were incubated at room 627 temperature for 30 min with peroxidase-conjugated secondary antibodies in TBST containing 628 5% BSA and washed three times with TBST. The signal was detected using ImmunoStar 629 Reagents (Wako, Tokyo, Japan) or SuperSignal WestFemto Maximum Sensitivity Substrate 630 (Pierce). The exposure time was adjusted such that the intensity of the bands was within the 631 linear range. To quantify the amount of proBDNF relative to the amount of total BDNF, the 632 blots were scanned and the images were converted to TIFF files and quantified using the NIH 633 ImageJ software (version 1.37v). After subtracting the background in the same lane, the 634 signals for proBDNF and mature BDNF were summed to obtain the total amount of BDNF.

636
To detect other molecules, the lysates were separated on SDS-polyacrylamide gels and 637 transferred to polyvinylidene fluoride membranes (Immobilon P membrane, Millipore). The 638 membranes were blocked at room temperature for 30 min in TBST containing 5% skimmed 639 milk (Nacalai Tesque, Kyoto, Japan), and then incubated with the primary antibodies in 640 TBST containing 3% BSA at room temperature overnight. The following primary antibodies 641 were HRP substrate (Millipore) and images of the immunoblots were captured using an 650 ImageQuant LAS500 (GE Healthcare) imaging system. Quantitative analyses of the band 651 intensities were performed using ImageQuant software (Image Analysis Software v8.1, GE 652 Healthcare). Hippocampal tissues were isolated from 9-week-old mice and rapidly 653 homogenized in lysis buffer comprising 50 mM Tris-HCl (pH 7.4), 1 mM EDTA, 150 mM 654 NaCl, 10 mM NaF, 1 mM Na3VO4, 1% Triton X-100, 10 mM Na2P2O7, 100 µM phenylarsine 655 oxide, 1% protease inhibitor cocktail, and 1% protein phosphatase inhibitor cocktail (Sigma). 656 Lysed tissues were incubated at 4°C for 20 min and centrifuged at 15,000 rpm for 15 min. 657 The protein levels in the supernatants were determined using the BCA Protein Assay kit 658 (Pierce). Protein G Sepharose (50 µl, GE Healthcare) was added to the hippocampal lysates 659 and the mixture was rotated at 4°C for 60 min. After removing the Protein G Sepharose, 1 µg whole brains were silver impregnated for 2 weeks, cryoprotected for 1 week, and then 673 sectioned (80 m) on a sliding microtome (REM-700, Yamato Kohki. Industrial Co., Ltd., 674 Saitama, Japan). The sections were developed, clarified, and then coverslipped with resinous 675 medium. During staining, image acquisition, and image analysis, the operators were blinded 676 to the genotype of each animal. Images of secondary dendritic segments of hippocampal 677 pyramidal neurons in the stratum radiatum in the CA1 region were obtained using an 678 AxioObserver Z1 microscope (Zeiss, Oberkochen, Germany) with a 63×/0.75 numerical 679 aperture objective lens (LD PlanNeo; Zeiss) for high magnification imaging or a 20×/0.8 680 numerical aperture objective lens (PlanApo; Zeiss) for low magnification imaging, and a 681 cooled CCD camera (CoolSnap HQ2; Photometrics, Tucson, AZ). For morphological 682 analyses of dendritic spines, a z-series with 0.5 µm intervals was created to visualize the 683 entire extent of each dendritic process and to capture spine details in all focal planes. The 684 density, maximum length, and maximum width of the dendritic protrusions was quantified 685 using the Region Measurements tool of MetaMorph Software (Universal Imaging 686 Corporation, West Chester, PA, USA) (10). The form factor of each protrusion, defined as the 687 ratio of its maximum length to its maximum width, was calculated (Takahashi et al., 2003).

688
The age, genotype, and number of mice used for morphological analyses are described in the 689 figure legends. The morphologies of dendritic protrusions were quantified using 33-36 690 segments of the secondary dendrites. These segments represented a total dendritic length of 691 1329-1683 µm per group of mice.

693
Electrophysiology 694 Field excitatory synaptic potentials were recorded as follows. Transverse slice was placed on 695 custom-made recording chamber according to a previous report (Tominaga,Tominaga,& 696 Ichikawa, 2002)with some modifications. Briefly, both sides of slice were perfused with 697 carbogenize aCSF containing (in mM) 118 NaCl, 3 KCl, 2.5 CaCl2, 1.2 MgCl2, 11 D-glucose, 698 10 HEPES, and 25 NaHCO3, at 2-3ml/min at 28 C. Field EPSPs were elicited by 10-100 μA 699 constant currents pulse (100 microsecond duration) with Pt bipolar electrode (FHC) onto the 700 Schaffer-collateral pathway and recored with glass electrode filled with aCSF (1-2 M ohm). To set stimulus intensities, maximum responses were determined by increasing stimulus 706 intensity stepwise to saturating responses. For paired-pulse experiment and long-term 707 potentiation experiments, stimulus intensity which elite about 33% of maximum 708 response were employed. Input-output relationship were studied by plotting fiber volley 709 response amplitude (mV) against fEPSP slope (mV/ms). Paired-pulse responses were also 710 recorded with various inter-pulse intervals from 500ms, 100ms, 50ms, 20ms and paired-pulse 711 ration (2nd pulse response to first response) were calculated. LTP was induced by four theta 712 burst stimulation, which consist of 4 burst at 0.05Hz of theta burst stimuli (4 bursts at 713 10Hz of 5 pulse at 100Hz ). On the other hands, LTD was induced by low frequency 714 simulation (1Hz 900 pulse). In case of LTD experiments, stimulus intensity which elicited 715 50 % of maximum response was employed.

717
Blood sampling and corticosterone measurements 718 Corticosterone measurements were performed as described previously (Schmidt et al., 2007). 719 To determine basal corticosterone levels, blood samples were collected between 10 and 12 720 a.m. The mice were moved to an adjacent room and placed individually in a transparent 721 restraint tube. Blood samples (approximately 100 µl) were collected into plastic tubes from 722 the tail vein without anesthesia, as described previously (Fluttert, Dalm, & Oitzl, 2000). In all 723 cases, the time between the first disturbance of the animals and the sampling was less than 1 724 min. Following blood collection, the mice were released and returned to their cages. Blood 725 samples were separated by centrifugation at 4°C for 15 min and the plasma was stored in the 726 freezer until use. To measure the change in blood corticosterone levels in response to acute 727 stress, sampling was performed 60 min before and after restraint stress was initiated.

728
Restraining was performed in plastic cylinders identical to those used for the prenatal stress 729 procedure. The AssayMax Corticosterone ELISA Kit (Assaypro, St. Charles, MO) was used 730 to measure corticosterone levels in the blood samples, according to the manufacturer's 731 instructions.

770
Human plasma collection 771 All human plasma collection protocols were approved by the Ethics Committee at Xiangya 772 Hospital Central South University. 11 Autism children and their 10 normal siblings as control 773 were recruited for blood sampling. Peripheral venous blood from patients and controls was 774 collected in EDTA-coated tubes and centrifuged at 1000g, 10 minutes to obtain the raw 775 plasma. Then, the raw plasma was centrifuged at 1000g, 10 minutes again to eliminate 776 residual blood cells. All plasma samples were stored at -80℃ before the proBDNF and 777 BDNF ELISA experiment. 778 779 BDNF and proBDNF ELISA 780 BDNF and proBDNF protein levels were determined by a two-side ELISA (Genstar, China, 781 C643-02 and C546-02) as described by the manufacturer. Plasma were loaded directly into 96 782 well plates without dilution. Absorbance was recorded and analyzed using a Cytation BioTeK 783 plate reader (Winooski, USA). BDNF and proBDNF concentration (pg/ml) was normalized 784 to the volume of plasma in each sample.

786
Statistics 787 Typically, statistical significance was determined using Student's t-tests after confirming 788 normal distribution of the data. In multi-bar figures, statistical significance was determined 789 by analysis of variance followed by a post hoc test. Data are represented as the mean ± SEM. 790 791 792 Table Legends 793 Table 1 The general information of ASD patients and controls  794 Subject ID, gender, birthday, recruit data, and age of all ASD patients and controls were 795 displayed. 796 797 and motor ability, ASD or autisitc features, and intellectual disability were summarized. In 800 this and all the following tables, '+' represents a positive symptom in an ASD patient, '-' 801 represents a negative one, and the blank means the imformation is not reported. Please note 802 '±' means the patient exhibit autistic features but not severe enough to be disagoised as the 803 ASD. 804 805 Table 3 Neurological problems in the ASD patients 806 Neurological tests for epilepsy, EEG and MRI abnormities, macrocephaly and sleep disorders 807 in the ASD patients. Please note '±' in the column of epilepsy means seizure was found in 808 that patients, but not severe enough for epilepsy diagnosis. 809 810 Table 4 Behavioral problems in the ASD patients 811 Behavior problems, including, the repetitive behavior, ADHD, anxiety, obsessive behavior, 812 self-injury, and aggressive behavior, were displayed in the ASD patients. 813 814 Table 5 No systemic problems observed in the ASD patients 815 Systemic features were documented for these ASD patients. No systemic problems, such as 816 eye abnormalities, recurrent infections, hypotonia, hand deformity, and short stature were 817 observed in all the ASD patients. 818 819 Table 6 The mBDNF and proBDNF level of plasma in the ASD patients 820 The mBDNF and proBDNF level of plasma in each of ASD patients were displayed. 821 822 823    Elucidation of neural network function in the brain-from the Ministry of Education, Culture, 840 Sports, Science and Technology of Japan (40344171)    with PSD-95 in +/+ and BDNF met/leu hippocampi. Hippocampal lysates (2 mg protein) were immunoprecipitated (IP) with an anti-PSD-95 antibody, and equal aliquots of the immunoprecipitants were subjected to immunoblotting (IB) with the indicated antibodies. Note that the interaction between PSD95 and GluN2B was decreased in BDNF met/leu hippocampus, compared with that in the +/+ hippocampus.    The recording of locomotor activity of BDNF +/+ (N=4) and BDNF met/leu (N=4) mice during 4.5 days, with 12 h light and 12 h dark phases. Y axis indicates the horizontal activity for 5 minutes.

Figure 6
Elevated stress response in BDNF met/leu mice.
(A) Tail suspension test of BDNF +/+ and BDNF met/leu mice at the age of 10~12 weeks. Animals were administered saline or fluoxetine 30 min before the tail suspension test (10 mg/kg i.p.). *, p < 0.05; ANOVA with post-hoc test. N=7~8 littermate animals. (B) Corticosterone levels in the serum before and after acute stress exposure. The tail blood samples were collected in group-housed BDNF +/+ and BDNF met/leu mice (10-week-old, N = 8 per genotype) (Left panel, before stress). After an additional 2 weeks of group-housing, the animals were subjected to immobilization stress (60 min) and the tail blood samples were collected immediately (Right panel, after stress). The serum corticosterone concentration was determined immediately after sampling.