To identify novel molecular targets relevant to regulation of Aβ levels in the brain, we screened viable mouse KO strains for decreased brain Aβ40. A total of 1930 different gene ablations giving rise to viable homozygous mice were evaluated. Two of the strains, PIGZ KO and QPRT KO, showed an unequivocal increase of Aβ40 and Aβ42 in the screening samples, whereas changes in Aβ were less obvious for the other KO strains. The two KO strains showing the lowest values of Aβ40 and Aβ42 in the screen, UBE2R2 and ADRM1, were of uncertain significance given the small number of samples tested.
Given the wide variety of mechanisms and proteins that have been reported to affect Aβ, and the relatively large number of gene KO strains tested, it was surprising that we did not identify a single new gene KO strain that caused a robust (≥50%) decrease in Aβ. A combination of several reasons may account for this. First, the choice of gene KO strains entering the screen was based on the 'druggable genome', which mostly limited the KO strains to proteins in gene families known to interact with small molecules. Second, the gene KO strains entering the screen were limited to those strains that resulted in viable homozygous adult mice. Approximately one third of the gene KO constructs made were homozygous embryonic lethal, and were consequently not included in the brain Aβ screen. Therefore, effective Aβ-lowering targets such as presenilin 1, for which genetic ablation is deleterious, would not have been tested in this screen. Third, functional redundancy, e.g. the genes encoding the Aph1B and Aph1C subunits of γ-secretase , could have obscured any effects of single gene ablations. There is also the possibility of developmental compensation, in which alternative pathways functionally substitute for the missing gene, thus restoring Aβ levels in the adult. Fourth, inbred mice have been shown to express significantly different levels of brain Aβ , raising the possibility that genetic changes in these mice may have obscured the function of some genes (epistasis). Fifth, there is the possibility that some genes may affect Aβ levels only in older mice, and therefore the role of these putative genes would not have been apparent at the age of three months when our mice were harvested. Sixth the ability of the screen to detect changes in Aβ was limited by the intrinsic variability in Aβ combined with the small group size, which in most cases was equal to four homozygous KO animals. This limitation to four animals per KO strain was necessary because of the resource and time constraints of producing and maintaining multiple KO mouse colonies, and the use of most of the available KO mice for other phenotypic and biochemical tests not reported here. Nevertheless the statistical power of the screen was favorable. Based on the good fit of the data to the normal distribution (Shapiro-Wilk's test p > 0.05), the false negative error rate (Type II error) was found to be 1.7% for detection of a KO strain with 50% lowering of Aβ, and 5.99E-9 for a KO strain with 85% lowering of Aβ, as observed in the BACE1 KO samples. Thus, there was a high probability for detection of any KO strains robustly lowering Aβ. In addition, the most practical Aβ-lowering targets should have the potential to lower Aβ by a robust and substantial amount, and thus, for the purpose of identifying the most practical targets, a low group size could be tolerated. The false positive error rate (Type I error) for a KO strain more than 3 sd below the mean was 0.13%, and for 10 sd above the mean was negligible at less than 6.7E-16. Thus, the results for the two KO strains, UBE2R2 and ADRM1, which exhibited the lowest brain Aβ levels around three standard deviations below the mean, could have been due to chance, whereas the Aβ increases in the PIGZ and QPRT KO strains were very unlikely due to chance. Finally, despite the possible limitations discussed above, it is hard to escape the conclusion that molecular targets capable of robust Aβ lowering in the relevant context of brain are rare.
Brain Aβ40, Aβ42 and Aβ1-x levels in GPR3 KO mice were evaluated more thoroughly by using a larger number of mice. No changes in brain Aβ42, Aβ40 or Aβ1-x were detected in sagittal brain halves from these mice. This contrasts with the results reported by Thathiah et al.  in which ca. 50% lowering of Aβ40 and Aβ42 was observed in APP-transgenic GPR3 KO mice. There are two noteworthy experimental differences between the two studies, first, we assayed endogenous mouse Aβ, not transgenic human Aβ, and second, we used sagittal brain halves not hippocampal sections. Unfortunately, our analysis did not extend to hippocampal sections, which constitutes only a small fraction of total brain. GPR3 is expressed at high levels in the cortex, which, like hippocampus, is relevant to AD. Thus, evaluation of Aβ in the hippocampus of non-APP GPR3 KO and in the cortex and/or whole brain of APP transgenic GPR3 KO would be interesting.
Two gene ablations, corresponding to the PIGZ and QPRT enzymes, exhibited significantly increased Aβ40 and Aβ42 in the screening samples. While further substantiation of the results for PIGZ and QPRT using larger groups of homozygous KO mice is clearly desirable, plausible molecular mechanisms for increased Aβ can be proposed. PIGZ, also known as SMP3, is a mannosyl transferase that catalyses addition of a fourth side chain mannose to the glycosylphosphatidylinositol (GPI) protein anchor precursor [58, 59]. In cell cultures, GPI anchored proteins are necessary for Aβ synthesis , and targeting of an artificial BACE1-GPI chimera to lipid rafts greatly increases Aβ production , although targeting of BACE1 to lipid rafts is not necessary for Aβ synthesis . Thus, in cell culture, a connection between GPI anchor metabolism and Aβ levels is well established. An effect of PIGZ on brain Aβ would extend these conclusions to a relevant organ in vivo, and further raises the possibility that the fourth mannose residue plays a specific role in Aβ metabolism. QPRT is the enzyme responsible for quinolinic acid turnover in the kynurenine pathway of tryptophan degradation , and therefore ablation of this gene would be expected to increase quinolinic acid levels. Increased quinolinic acid has been reported in AD brain [62, 63], and treatment of primary neuronal cultures with quinolinic acid has been reported to increase cell death and oxidative stress . The association of oxidative stress with increased BACE1 activity and Aβ production has been widely substantiated in AD [10, 15, 65–69], raising the possibility of a mechanistic connection between quinolinic acid and Aβ through activation of BACE1 by oxidative stress. In addition, quinolinic acid has been reported to increase APP expression in rat brain, which could contribute to increased Aβ production .
The two KO strains for which the lowest values of Aβ40 and Aβ42 (ca. 30% lowering) were observed in our screen corresponded to the UPS proteins Adrm1 and Ube2R2. Adrm1, also known as hRpn13, associates with the proteasome 19S regulatory particle, and is required for recruitment of the Uch37 deubiquitinating enzyme to the proteasome [71, 72]. Ube2R2 (sequence NM-017811) is a ubiquitin conjugating enzyme. Decreased expression of several other ubiquitin conjugating enzymes has been reported to decrease Aβ production in cell culture . The UPS has multiple potential roles in AD in addition to possible regulation of Aβ levels, as recently reviewed in detail by Upadhya and Hegde . Possible mechanisms of proteasomal regulation of Aβ include resveratrol-activated clearance of Aβ , and competition with γ-secretase for APP processing . Thus, an intriguing possibility is that selective inhibition of specific sub-pathways of the UPS might decrease brain Aβ levels by both biosynthetic and clearance mechanisms. However, from a drug discovery perspective, this would carry the risk of further exacerbating the already defective proteasome activity prevalent in AD thought to result from the accumulation of toxic Aβ and tau aggregates. Furthermore, assuming that maximal inhibition of Adrm1 or Ube2R2 would elicit only a 30% decrease in brain Aβ, even the effect of an inhibitor with ideal drug properties would be limited, and the expected small changes in Aβ difficult to quantify.