Somatostatin slows Aβ plaque deposition in aged APPNL-F/NL-F mice by blocking Aβ aggregation in a neprilysin-independent manner

The molecular underpinnings that govern the endoproteolytic release of the amyloid beta peptide (Aβ) from the amyloid precursor protein (APP) are now quite well understood. The same cannot be said for the events that precipitate the aggregation and amyloid deposition of Aβ in Alzheimer’s disease (AD). The 14-amino-acid cyclic neuroendocrine peptide somatostatin (SST-14) has long been thought of as playing a role, foremost by controlling the expression of the Aβ clearing enzyme neprilysin, and more recently by directly interacting with Aβ oligomers. Missing have been in vivo data in a relevant Aβ amyloidosis model. Here we addressed this shortcoming by crossing AppNL-F/NL-F mice with Sst-deficient mice of identical genetic background to assess if and how the presence of Sst influences key pathological hallmarks of Aβ amyloidosis that develop in AppNL-F/NL-F mice after 10 months of age. Surprisingly, we found that Sst had no influence on whole brain neprilysin transcript, protein or activity levels, an observation that cannot be accounted for by a compensatory upregulation of the Sst paralog, cortistatin (Cort), that we observed in 15-month-old Sst-deficient mice. The absence of Sst did lead to a subtle but significant increase in the density of cortical Aβ amyloid plaques. Follow-on western blot analyses of whole brain extracts indicated that Sst interferes with early steps of Aβ assembly that manifest in Sst null brains through the appearance of SDS-stable smears of 55- 150 kDa. As expected, no effect of Sst on tau steady-state levels or its phosphorylation were observed. Results from this study are easier reconciled with an emerging body of data that point toward Sst affecting Aβ amyloid plaque formation through direct interference with Aβ aggregation rather than through its effects on neprilysin expression.


INTRODUCTION 45
The peptide hormone somatostatin (SST) was initially discovered as an inhibitor of growth 46 hormone secretion (1, 2). It is now understood that SST, along with its paralog cortistatin 47 (CORT) (3), modulates a host of activities in the gastrointestinal, nephritic, immune, and central 48 nervous systems (4). Both hormones are released from larger proproteins exceeding one 49 hundred amino acids in length through endoproteolytic cleavages of C-terminal peptides which 50 cyclize by disulfide bonding. Alternative endoproteolytic processing produces prominent 51 variants named according to their amino acid sequence lengths as SST-14, SST-28, CORT-17, and 52 CORT-29. These peptides have similar yet distinct physicochemical properties and affinities for 53 five human somatostatin receptors (SSTRs), which they bind to following their release into the 54 extracellular space (5-8). The distinct binding profiles of SST and CORT, in turn, elicit distinct 55 physiological responses that point toward overlapping but non-identical functions (9-11). 56 Separate lines of biochemical, histological, and genetic evidence implicate the 57 somatostatinergic system in Alzheimer's disease (AD) (12-14). In fact, among the first 58 biochemical analyses of postmortem AD brains were two studies that drew attention to a 59 profound reduction in somatostatin levels relative to age-matched control brains (15, 16), a 60 finding that has since been validated in human postmortem brains (17) and an AD mouse model 61 (18). 62 Somatostatin-expressing neurons or interneurons may be more vulnerable to certain 63 AD-associated stressors. For instance, cortical SST immunoreactive neurons were reported to 64 be more susceptible to degeneration than neighboring neuropeptide Y expressing cells, with 65 the loss of these neurons proportional to amyloid plaque and neurofibrillary tangle burden (15, 66 monomeric Aβ (31). We then established that nanomolar concentrations of SST or CORT 89 modulate the aggregation of Aβ42 (31). Since then, the selective SST binding to oligomeric Aβ42 90 has been shown to extend to a β-sheet pore-forming Aβ42 tetramer with putative neurotoxicity 91 (32). The possibility that SST may modulate Aβ plaque formation directly is heightened by the 92 fact that this molecule spontaneously forms amyloid fibrils in vitro (33) and is, in fact, stored 93 prior to its synaptic release in dense granules as a natural amyloid (34). Upon their release, it 94 takes in the order of minutes to hours before these SST amyloids are dissolved (35), providing 95 an opportunity to influence nearby Aβ aggregation in AD (12). 96 Remarkably, despite all these observations implicating somatostatin as a possible factor 97 in AD spanning decades, no prior experimental study has investigated the influence of SST in an 98 in vivo Aβ amyloidosis paradigm. Here, we set out to begin to address this unmet need by 99 studying intercrosses between Sst null mice and a humanized App knock-in mouse line that is 100 known to develop a profound Aβ amyloidosis as the mice age. Breeding the cross as Sst 101 hemizygotes provided progeny of mixed Sst genotype which consistently developed Aβ amyloid 102 plaques from middle to late adulthood. We present data that queried how Sst deficiency 103 affected (i) transcript levels of cortistatin and neprilysin, (ii) Aβ amyloid plaque burden and 104 plaque distribution, (iii) neprilysin protein and activity levels, (iv) SDS-stable Aβ oligomers, as 105 well as (v) steady-state tau levels and tau phosphorylation. To generate these data, we made 106 use of App NL-F/NL-F Sst -/mice, with age-matched Sst-expressing App NL-F/NL-F Sst +/+ mice serving as 107 controls. 108

Western blot analyses 151
Frozen left cerebral hemispheres of mice were weighed and homogenized with 0.7 mm 152 diameter zirconia beads (catalog number 11079107, BioSpec Products, Bartlesville, OK, USA) in 153 a buffer composed of 100 mM Tris-HCl, pH 8.3, 100 mM NaCl, 2x Roche PhosSTOP phosphatase 154 inhibitor cocktail (catalog number 4906837001, Roche, Basel, Switzerland), 2x Roche cOmplete 155 protease inhibitor cocktail (catalog number 11836170001, Roche) with three 1-minute pulses of 156 bead-beating and 1 minute of cooling on ice between each pulse. Note that detergent was 157 omitted during homogenization to minimize foaming but was added for the solubilization of 158 membrane proteins during the subsequent protein extraction. Specifically, homogenates were 159 extracted with 1% NP40, 100 mM Tris-HCl, pH 8.3, 100 mM NaCl, 1x Roche PhosSTOP 160 phosphatase inhibitor cocktail, 1x Roche cOmplete protease inhibitor cocktail, at 4 °C for 30 161 minutes with agitation before centrifugation at 4 °C for 5 minutes at 2000 g and 15 minutes at 162 21130 g. Supernatants of brain protein extracts were collected following centrifugation, then 163 protein concentrations were determined by bicinchoninic acid assay (catalog number 23227, 164 To assess neprilysin activity, brain samples were analyzed using the Sensolyte 520 Neprilysin 177 Activity Assay Fluorometric Kit (catalog number AS-72223, AnaSpec) and following the 178 manufacturer's instructions. Briefly, 10% brain extracts were generated and cleared of insoluble 179 material as described in the western blot analysis section. Throughout these steps, the brain 180 extracts were kept on ice. Next, 50 µL of brain extract supernatants were pipetted onto a 96-181 hippocampal regions were manually defined (as exemplified in Fig 3B), creating two annotation 206 layers per slide. The anterior olfactory nucleus and corpus callosum were included in the 207 cortical annotation. Vasculature and meninges were manually excluded from the annotations. 208 Classifiers for plaques (4G8-positive tissue), non-plaque (4G8-negative tissue) and background 209 (voids in the tissue) were defined manually and iteratively to compensate for differences in 210 background staining such that the minimum number of classifiers were applied to the dataset. 211 Following automated detection of the classifiers, the quantified annotations of each slide were 212 manually examined to verify that the classifier settings discriminated plaque, non-plaque, and 213 background. Raw data (object data) from each analysis were exported in text format. The areas 214 of all objects classified as plaques and non-plaques from a given annotation on a given slide 215 were summed to give the total cortical or hippocampal area, respectively. 216 intercrossing if the genes were linked and recombination disfavored. A closer analysis mostly 285 dispelled this concern because the Sst gene maps to cytoband qB1 (23889573-23890958 bp, 286 Genome Reference Consortium mouse build 38/mm10) and the App gene is coded in cytoband 287 qC3.3 (84952666-85175255 bp), located 32 cM apart (Fig 1A). 288 Sst and App proteins are coded by multi-exon gene sequences (Fig 1B). Several (Fig 1C). 293 To our knowledge the sequence of this insert was not originally published, which is why we 294 inserted it here as a supplement (S1 Fig). This Sst null line was selected because it had been 295 reported to feature normal hippocampal App expression. Moreover, the aforementioned data, 296 which suggested Sst-deficiency leads to diminished neprilysin activity and elevated Aβ42 levels, 297 were also generated with this line (26). 298 The selection of a compatible Aβ amyloidosis model fell on a previously reported knock-299 in mouse line sharing with the Sst null mice the same C57BL/6 inbred genomic background (36). 300 This App NL-F/NL-F line carries several mutations that flank or alter the Aβ42 coding sequence. 301 Cumulatively these changes humanize Aβ, increase production of Aβ42 three-fold, and drive 302 the Aβ42-to-Aβ40 brain production ratio more than 10 times higher, relative to wild-type (36). 303 We chose this knock-in model to avoid insertion artefacts and variances that are frequently 304 observed in transgenic lines due to the large and fluctuating number of transgene copies. Prior 305 global proteome analyses of whole brain and hippocampus samples from these mice indicated 306 that Sst expression is comparable in APP NL-F/NL-F and APP wt/wt control mice (Cort was not 307 detected in these analyses) (40, 41). 308 To generate cohorts for this study, three consecutive intercrosses were undertaken. 309 More specifically, we initially crossbred the Sst null line with homozygous App NL-F/NL-F mice (Fig  310   1D). Next, the F1 generation (App NL-F/wt Sst +/-) was backcrossed with APP NL-F/NL-F mice to produce 311 intercrosses that were homozygous for the NL-F/NL-F mutation but heterozygous for the Sst 312 knockout (App NL-F/NL-F Sst +/-). Due to the need for a Chromosome 16 crossover event to achieve 313 this genotype only approximately 15% of the F1 offspring from this second intercrossing, i.e., 13 314 out of a total of 87 mice, were confirmed as App NL-F/NL-F Sst +/-. Finally, App NL-F/NL-F Sst +/were 315 intercrossed and progeny were genotyped from tail clippings using a customized PCR analysis 316 (Fig 1E). This scheme minimized undesired variability by ensuring that Sst null mice (App NL-F/NL-317 F Sst -/-) shared the same parents and housing as their wild-type Sst littermates (App NL-F/NL-F Sst +/+ ). 318 Genotyping validated that App NL-F/NL-F Sst -/-, App NL-F/NL-F Sst +/-, and App NL-F/NL-F Sst +/+ progeny were 319 obtained in approximate Mendelian ratios. App NL-F/NL-F Sst +/hemizygotes and App NL-F/NL-F Sst -/-320 knockouts were physically and behaviorally unremarkable and lived past 16 months of age with 321 a mortality equivalent to their Sst wild-type littermates. 322

Sst, Mme, and Cort transcript expression in App NL-F/NL-F Sst -/mouse brains 323
Prior work indicated that mRNA transcript levels of the Sst paralog Cort may undergo a gender-324 dependent compensatory increase in Sst knockout mice (42). Moreover, transcription of the 325 neprilysin gene (Mme) has been reported to be induced by Sst (26), a finding that could have 326 direct implications for interpreting effects of Sst on Aβ amyloid deposition, due to the enzyme-327 substrate relationship between neprilysin and Aβ. To explore these possibilities, we aged Sst 328 wild-type, hemizygous, or knockout App NL-F/NL-F mice to 12 or 15 months, ages at which the 329 App NL-F/NL-F genotype gives rise to subtle and prominent Aβ amyloid deposition respectively ( Fig  330   2A). We then determined transcript levels of Sst, Mme, Cort, and Ppia, the latter a common 331 proxy for total RNA in RT-qPCR of brain tissue. As biological source materials served mid-sagittal 332 cut half brains, as opposed to specific brain regions, to minimize variances that can be 333 introduced during dissection steps. 334 Distributions of transcript abundance for Sst (Fig 2B), Cort (Fig 2C), and Mme (Fig 2D)  335 were approximately normal, and coefficients of variation ranged from 0.35 to 0.43, 0.12 to 336 0.34, and 0.07 to 0.44, respectively. Moreover, for all transcripts measured, mRNA levels in 337 males and females were equivalent, as assessed by t-test. As expected, these analyses revealed 338 that Ppia transcript levels were highly consistent between all samples (Fig 2E), indicating that 339 both the neurophysiological backgrounds sampled, and the sample preparations were 340 consistent. 341 The brain Sst mRNA concentration of App NL-F/NL-F Sst +/hemizygotes averaged around half 342 that of Sst wild-type App NL-F/NL-F Sst +/+ animals, suggesting that in most hemizygote animals a 343 compensatory transcriptional upregulation of the remaining Sst allele did not occur (Fig 2B). 344

However, overlapping distributions of Sst transcript levels in Sst wild-type App NL-F/NL-F Sst +/+ and 345
Sst hemizygote App NL-F/NL-F Sst +/mice indicated a degree of flexibility of Sst expression and 346 raised the possibility that some hemizygous individuals may produce sufficient levels of Sst to 347 be phenotypically wild-type. The Sst transcript was undetectable in all App NL-F/NL-F Sst -/mice, 348 confirming that the mutated allele was not expressed. 349 At both ages examined, Sst wild-type App NL-F/NL-F Sst +/+ mice had consistent levels of Cort, 350 Mme, and Sst transcripts, indicating that each of these genes was stably expressed over the 3 351 month age range (Fig 2B-D). Cort mRNA transcript levels increased in Sst-deficient App NL-F/NL-352 F Sst -/mice by approximately one third (35%), relative to age-matched Sst wild-type App NL-F/NL-353 F Sst +/+ mice. This increase was significant (p<0.05) between Sst wild-type and knockout App NL-354 F/NL-F cohorts at 15 months of age (Fig 2C). Remarkably, there was no significant difference in 355 neprilysin transcript levels between Sst wild-type App NL-F/NL-F Sst +/+ and Sst knockout App NL-F/NL-356 F Sst -/animals at 12 or 15 months of age (Fig 2D). 357 In summary, these RT-qPCR data validated a compensatory increase in Cort mRNA levels 358 in response to Sst knockout but did not corroborate the previously reported male sex-specificity 359 of this increase (42). These data also did not validate the concept of Sst acting as an inducer of 360 whole brain neprilysin expression. 361

Amyloid plaque deposition in App NL-F/NL-F Sst -/mice 362
The RT-qPCR data may lead to an anticipation that the compensatory increase in Cort transcript 363 levels could diminish effects of Sst knockout on Aβ deposition. To address this question 364 experimentally, we next assessed cohorts of App NL-F/NL-F Sst +/+ , App NL-F/NL-F Sst +/and App NL-F/NL-F Sst -365 /mice for Aβ deposition at 12 and 15 months of age. To this end, half brains were formalin 366 fixed, paraffin-embedded and parasagittally cut. Next, 4G8 anti-Aβ immunoreactivity was 367 visualized with the horseradish peroxidase substrate 3, 3-diaminobenzidine, and Aβ plaque 368 densities were quantified (Fig 3A). Immunoreactivity was confined mainly to the cortex, 369 hippocampus, and olfactory bulb (Fig 3B-D). A small number of amyloid plaques were 370 occasionally observed within the corpus callosum and the anterior olfactory nucleus, 371 particularly in brain sections of 15-month-old mice. The brainstem, midbrain, and cerebellum 372 were devoid of 4G8 immunoreactivity in all samples. Of all cortical areas, the prelimbic and 373 orbital areas generally had the lowest plaque density. Cortical plaque density increased from 12 374 to 15 months of age in each Sst genotype (App NL-F/NL-F Sst +/+ : 3.7 times; App NL-F/NL-F Sst +/-: 4.6 375 times; App NL-F/NL-F Sst -/-: 4.5 times). At both ages studied, the sex of the animals did not appear to 376 influence the propensity or speed of plaque formation that we captured in these analyses. 377 When comparing the cortical Aβ amyloid plaque density for all plaques larger than 0.25 378 µM in diameter (the smallest plaque size that could be reliably detected in these analyses), we 379 observed that Sst-deficient App NL-F/NL-F Sst -/mice exhibited a statistically significant increase in 380 plaque density, relative to Sst wild-type App NL-F/NL-F Sst +/+ at 15 months (Fig 3E). 381 To dig deeper into how Sst deficiency might affect Aβ amyloid deposition, we 382 segregated Aβ plaques with diameters up to 300 µM into three equally proportioned size 383 ranges. Perhaps not surprisingly, the smallest plaque size range contained the largest 384 proportion of total plaques in all three mouse populations at both ages. We were particularly 385 interested in the relative densities of amyloid plaques of the smallest size in 12-month-old 386 mice, hypothesizing that Sst influences the earliest steps of Aβ aggregation. The Sst-associated 387 difference in plaque densities was indeed most pronounced at this relatively early timepoint in 388 the size range that included the smallest detectable plaques (0.25-100 µm) but did not reach 389 statistical significance (Fig 3F). In the second smallest size range examined (100-200 µm), Sst-390 deficient mice had 17% more plaques than Sst wild-types. Larger plaques (>200 µm) made up a 391 smaller proportion of the total plaque area, namely 16%, 12%, and 11% of total plaques in Sst knockout App NL-F/NL-F Sst -/mice (Fig 4A). Next, we measured the in vitro activity of Sst wild-type 412 and deficient App NL-F/NL-F brain extracts toward a commercial fluorescent neprilysin substrate (5-413 FAM/QXL-520). In these analyses, background (non-neprilysin) activities within brain extracts 414 toward the fluorescent substrate were elucidated by incubating brain extracts in the presence 415 of the neprilysin-specific inhibitor thiorpan (Fig 4B). To estimate total neprilysin activity levels in 416 our extract fractions we compared their activity to the activity of a known amount of 417 recombinant human neprilysin. These analyses established that the presence or absence of Sst 418 did not affect neprilysin expression or activity in a significant manner in total brain extracts. 419

Sst ablation promotes the formation of high molecular mass Aβ oligomers 420
We considered that Sst may influence the earliest Aβ aggregation steps, namely 421 oligomerization. This idea was based on our earlier observation that the addition of Sst to a 422 synthetic Aβ1-42 preparation delays Aβ in vitro aggregation in a concentration-dependent 423 manner (31) and on data from molecular dynamics simulations by our collaborators showing 424 that the presence of Sst interferes with early Aβ oligomer assembly (43). We made use of the 425 monoclonal 82E1 antibody, which exclusively detects Aβ assemblies with exposed Aβ N-termini 426 and, consequently, will not bind APP or its derivatives encompassing the noncleaved Aβ 427 sequence (44). The same 15-month-old brain extracts analyzed by 82E1 western blot had 428 striking differences between Sst wild-type App NL-F/NL-F Sst +/+ and knockout App NL-F/NL-F Sst -/mice: 429 Whereas Sst knockout App NL-F/NL-F Sst -/mice gave a strong 82E1-reactive smear covering 430 apparent molecular weights of 55-150 kDa, wild-type App NL-F/NL-F Sst +/+ gave rise to considerably 431 weaker signals in this molecular weight range (Fig 5A). Densitometry showed that the 432 difference in signal intensities in this molecular weight range were highly significant (Fig 5B). 433 Next, we generated western blots with the 6E10 anti-Aβ epitope, which predominantly 434 detects full-length APP in extracts of 15-month-old App NL-F/NL-F upon low exposure of blots, to 435 determine if Sst influences the expression of this precursor. We also produced a long exposure 436 of the low molecular weight region of the 82E1 western blot for the detection of monomeric Aβ 437 (Fig 5C). Finally, we probed western blots with antibodies directed against total tau and a 438 prominent tau phosphorylation epitope encountered in AD brain samples (Fig 5D). No changes 439 in APP, monomeric Aβ, total tau or phospho-tau were observed. 440 Taken together, experiments in this section revealed that Sst slows Aβ deposition 441 primarily through a neprilysin-independent effect on Aβ oligomerization that appears to affect 442 early steps in Aβ oligomer assembly. Accordingly, Sst ablation increased the formation of SDS-443 resistant Aβ oligomers but did not influence levels of the APP precursor or the production of 444 monomeric Aβ. 445

DISCUSSION 446
This study was designed to reveal whether Sst gene ablation affects Aβ amyloidosis in vivo. To 447 this end, we crossed the extensively studied C57BL/6-derived Sst null mouse line (37) and 448 App NL-F/NL-F mice (36). The Sst gene knockout was validated by genotyping and RT-qPCR and was 449 observed to cause a slight compensatory upregulation of the mRNA levels of the Sst paralog 450 Cort. Remarkably, neither whole brain neprilysin transcript nor protein or activity levels were 451 impacted by Sst genotype, yet Sst ablation still led to significantly stronger cortical Aβ 452 deposition in 15-month-old App NL-F/NL-F mice. When Aβ plaques were binned by size in 12-453 month-old App NL-F/NL-F mice, which are at early stages of plaque deposition, it became apparent 454 that the Sst gene ablation promoted the formation of the smallest Aβ plaques. A follow-on 455 investigation into the mechanism revealed that in the absence of Sst the signal intensities of 456 SDS-stable Aβ oligomers were strongly increased relative to wild-type App NL-F/NL-F mouse brain 457 levels, in the absence of an effect on APP precursor or monomeric Aβ levels. The present work revealed whole brain neprilysin levels to be unaffected by Sst ablation 480 at the mRNA or protein levels. It could be said therefore that a caveat in the interpretation of 481 these divergent results is the difference in brain structures analyzed, i.e., hippocampus versus 482 whole brain, including neprilysin-rich structures (basal ganglia, pituitary), which could dilute 483 hippocampus-specific effects. Consistent with this interpretation, the difference in plaque 484 numbers that we observed at 12 months were more pronounced in the hippocampus than 485 cortical areas. More specifically, only in the hippocampi of 12-month-old mice did any plaque 486 size range (plaques with areas 100.1-200 square micrometers) satisfy the t-test 0.05 confidence 487 threshold. As we did not survey RNA expression in the hippocampus specifically, we cannot 488 reject the hypothesis that this Sst-dependent local inhibition of plaque formation is due to a 489 functional interaction between Sst and neprilysin. That said, according to recent Human Protein 490 Atlas data (https://www.proteinatlas.org/ENSG00000196549-MME/brain) the hippocampus 491 expresses a small fraction of total brain neprilysin mRNA and protein, such that if hippocampal 492 neprilysin levels were to increase several-fold while remaining constant elsewhere, total 493 expression would still be subtly affected (49). Therefore, one conclusion of our study is that Sst 494 is likely to play a lesser overall role for the expression of neprilysin throughout the brain than 495 was previously appreciated. 496 We are not the first to suggest that the relationship between Sst and neprilysin 497 expression may be limited to specific brain structures. For instance, the initial report 498 establishing this relationship noted that in the cerebellum neprilysin is insensitive to Sst gene 499 ablation (26). More recently, a reduction in the levels of human SST and CORT in the temporal 500 lobe of AD postmortem brains could not be correlated with neprilysin levels (17). Finally, the 501 aforementioned administration of the Sst-fusion construct only changed neprilysin levels in the 502 hippocampus (48). 503 How else, if not on the basis of SST controlling whole brain neprilysin expression, can we 504 account for our observation that SST ablation increases the proportion of small plaques? We 505 previously showed that Sst binds to oligomeric Aβ and modulates Aβ aggregation in vitro, 506 preventing the formation of amyloid fibrils that can incorporate thioflavin T (31). A closer look 507 at fractions generated by co-incubation of Aβ and Sst showed that Sst delayed Aβ aggregation 508 by stabilizing smaller oligomeric Aβ structures. More recently, a separate group reported that 509 SST can bind certain Aβ oligomers (32). Consistent with these earlier data, the in vivo 510 immunohistochemical observations reported here support the notion that Sst inhibits early 511 events in amyloid plaque formation such that plaques form more readily in Sst knockouts, then 512 grow at a rate unaffected by Sst genotype (Fig 6). 513 Arguably, our most instructive data for mechanistic modelling come from the western 514 blotting of brain extracts derived from 15-month-old App NL-F/NL-F Sst +/+ and App NL-F/NL-F Sst -/mice. 515 These western blot data established that the presence of Sst reduced the levels of 55-150 kDa 516 bands that were reactive to an Aβ-specific antibody. That these signals originated from Aβ 517 oligomers, as opposed to another APP isoform or cleavage product, can be deduced from the 518 specificity of the antibody used, which is known to selectively target the N-terminal cleavage 519 site of Aβ. Less obvious is whether the Sst-dependent slowing of Aβ aggregation would be 520 therapeutic or detrimental for individuals with AD. Although it may be more intuitive to 521 interpret the role of Sst as beneficial, whether this is true will depend on where exactly in the 522 Aβ cascade Sst interferes, i.e., whether the presence of Sst increases or decreases the levels of 523 the most toxic Aβ aggregation intermediates. previous report, we also found Cort transcript levels to increase in Sst null mice by around 35%. 530 In contrast to the prior work, however, we observed no difference in this regard between male 531 and female App NL-F/NL-F Sst -/mice. This apparent distinction may merely reflect our inclusion of 532 the cerebrum, cerebellum, and caudoputamen, areas of the brain that express Cort and 533 perhaps obscured sex-specific differences that may still exist in the hypothalamus. 534 When comparing female and male APP NL-F/NL-F mice of each of the three Sst genotypes, 535 we observed no sex-specific differences in cortical and hippocampal plaque density and plaque 536 size distributions. In the absence of a sex-specific phenotype, which could have clarified the 537 possible influence of a compensatory biology, our data cannot resolve the extent to which Cort-538 based compensation may have precluded the formation of more striking differences in AD-like 539 Aβ amyloidosis. A Cort null mouse (50) and an Sst/Cort double knockout mouse (51) have been 540 reported, raising the possibility that a triple transgenic line could be produced. 541 Another factor that could have masked the physiological influence of Sst on Aβ 542 amyloidosis and AD molecular signatures in this study is the artificially high level of Aβ1-42 that 543 the App NL-F/NL-F knock-in mice are known to produce. This caveat represents a conceptual catch-544 22 because the very ability to measure the influence of Sst on Aβ amyloidosis depends on a 545 paradigm that can produce sufficient Aβ levels to study its aggregation. Thus, any anticipation 546 that the ablation of endogenous Sst would by itself profoundly affect amyloid plaque formation 547 in APP NL-F/NL-F mice needs to be tempered by the recognition that Aβ deposition in this model 548 depends on Aβ1-42 production that far exceeds physiological Aβ production in human AD brains. 549 When considered in this context, it is remarkable that Sst gene ablation alone was sufficient to 550 increase the levels of SDS-stable Aβ assemblies to the extent seen. 551

CONCLUSIONS 552
The results of this study document that Sst does not alter steady-state neprilysin transcript, 553 protein or activity levels in whole brain extracts yet represent robust in vivo evidence of an 554 inverse relationship between Sst levels and Aβ deposition. As such, the results are easier 555 reconciled with a model of Sst affecting Aβ aggregation directly, consistent with recent 556 biochemical and molecular dynamics simulation studies. This scenario also fits with results from 557 new work documenting a relatively close spatial correlation of Sst and Aβ release sites in 558 various brain regions (52). Future work will need to establish if the combined ablation of the Sst 559 and Cort genes will further enhance the impact on Aβ amyloid deposition. Once we understand 560 if the Sst-mediated slowing of Aβ aggregation reduces or exacerbates toxicity in vivo, refined 561 strategies should come to the fore that can harness the still untapped therapeutic potential of 562 this cyclic neuropeptide for the treatment of AD. 563

ACKNOWLEDGEMENTS 564
The authors thank Dr. Paul McKeever, University of Toronto, for highly informative discussions 565 on immunohistochemistry and neuroanatomy, Zhilan Wang, University of Toronto, for tissue 566 embedding, sectioning and slide mounting, and Erica Stuart, University of Toronto, for help with 567 RT-qPCRs. 568  In 15-month-old App NL-F/NL-F mice, the presence or absence of Sst has no effect on whole-brain 644 steady-state neprilysin transcript, protein or activity levels (1). Instead, Sst slows the formation 645 of Aβ oligomers that can be visualized as 55-150 kDa SDS-resistant 82E1-reactive bands (2) and 646 shifts Aβ amyloid formation and deposition (3). No effect of Sst was observed on steady-state 647 tau protein levels or the tau phosphorylation at key AD phospho-acceptor sites. This cartoon 648 was adapted and modified from a previously published figure (see Figure 2 in (12) Hippocampal Aβ amyloid plaque densities increased between 12-and 15-month-old App NL-F/NL-F 808 mice. Differences in hippocampal Aβ amyloid plaque densities were observed when comparing 809 App NL-F/NL-F mice that expressed wild-type Sst versus Sst-deficient mice. More specifically, a 810 trend toward higher Aβ plaque densities of small sizes (0.25-200 µm) was observed in 12-811 month-old Sst gene-deficient mice, echoing the increase in Aβ amyloid plaque densities 812 observed in the cortex of Sst ablated App NL-F/NL-F mice (Fig 3E). 813

Figure 1
App Sst Step

App Sst
Step 2:    Sst genotype +/+ -/- Sst slows formation of larger Aβ oligomers X No effect of Sst on whole brain steady-state neprilysin transcript, protein or activity levels Sst shifts Aβ plaque size distribution toward smaller plaques No influence of Sst on tau levels or phosphorylation X