Mutational re-modeling of di-aspartyl intramembrane proteases: uncoupling physiologically-relevant activities from those associated with Alzheimer’s disease

The intramembrane proteolytic activities of presenilins (PSEN1/PS1 and PSEN2/PS2) underlie production of β-amyloid, the key process in Alzheimer’s disease (AD). Dysregulation of presenilin-mediated signaling is linked to cancers. Inhibition of the γ-cleavage activities of PSENs that produce Aβ, but not the ε-like cleavage activity that release physiologically essential transcription activators, is a potential approach for the development of rational therapies for AD. In order to identify whether different activities of PSEN1 can be dissociated, we designed multiple mutations in the evolutionary conserved sites of PSEN1. We tested them in vitro and in vivo assays and compared their activities with mutant isoforms of presenilin-related intramembrane di-aspartyl protease (IMPAS1 (IMP1)/signal peptide peptidase (SPP)). PSEN1 auto-cleavage was more resistant to the mutation remodeling than the ε-like proteolysis. PSEN1 with a G382A or a P433A mutation in evolutionary invariant sites retains functionally important APP ε- and Notch S3- cleavage activities, but G382A inhibits APP γ-cleavage and Aβ production and a P433A elevates Aβ. The G382A variant cannot restore the normal cellular ER Ca2+ leak in PSEN1/PSEN2 double knockout cells, but efficiently rescues the loss-of-function (Egl) phenotype of presenilin in C. elegans. We found that, unlike in PSEN1 knockout cells, endoplasmic reticulum (ER) Ca2+ leak is not changed in the absence of IMP1/SPP. IMP1/SPP with the analogous mutations retained efficiency in cleavage of transmembrane substrates and rescued the lethality of Ce-imp-2 knockouts. In summary, our data show that mutations near the active catalytic sites of intramembrane di-aspartyl proteases have different consequences on proteolytic and signaling functions.


INTRODUCTION
Presenilins (termed PS1 and PS2 or PSEN1 and PSEN2 [1,2]) are homologous di-aspartyl proteases, capable of cleaving various type I transmembrane proteins within their intramembrane domains. Mutations in PSEN1 and PSEN2 are major causative genetic factors of familial cases of Alzheimer's disease (AD), characterized by early onset AD manifestation [1,2]. PSEN1 or PSEN2 intramembranous and BACE1 extracellular cleavages of amyloid precursor protein (APP), produce short 40-, 42-amino acid β-amyloid peptides (Aβ). AD autosomal dominant missense mutations in the presenilins have been reported to increase Aβ production and the ratio of Aβ42/40 peptides [3]. PSEN cleavage releases the intracellular domains (ICD) of type I proteins that can act as intracellular signaling molecules, activating gene transcription (e.g., Notch-signaling genes) (reviewed in [3]). Presenilins function as components of the multipleprotein γ-secretase complex and have evolutionarily invariant amino acid signatures around two conserved catalytic aspartates and a PAL-motif (human PSEN1 -D257, D385, PAL433-435) ( Figure 1A, Supplementary Figure 1) [4][5][6][7][8][9][10][11]. There are three major proteolytic activities associated with presenilins: (i) "presenilinase"-PSEN autocleavage, (ii) intramembrane γ-cleavage leading to generation of Aβ peptides and (iii) juxtamembrane ε-cleavages of APP, Notch 1 and other type I protein substrates resulting in release of ICDs -intracellular transcriptional regulators (Supplementary Figure 2) Active γ-secretase complex requires four proteins: Nicastrin, PEN2, APH1 and PSEN [12, reviewed in 13, 14]. Although various missense mutations in PSEN1 lead to autosomal-dominant AD (summarized in AlzForum Mutation Database), heterozygous loss-of-function mutations in PSEN1 as well as in Nicastrin and PEN2 (haploinsufficiency) have been shown to cause specific severe inflammatory skin disease, termed acne inversa in humans [15], reviewed in [16]. Clinical trials of drugs for AD inhibiting γ-secretase activity revealed various effects on skin, including a higher risk of skin cancer [16,17]. In mice, loss of PSEN1 causes skin cancer, and a reduction of PSENs function is responsible for myeloproliferative disease [18,19]. An inverse association between AD and cancer has been proposed with multiple regulatory mechanisms, including Pin1-, p53-, Wnt-related signaling, proposed to underlie the diseases [20, 21, reviewed in 22]. Among the important presenilin functions is regulation of Wnt signaling/β-catenin phosphorylation and turnover, which can contribute to skin cancer [18,[23][24][25]. This regulation can occur indirectly via cadherins as described in [26]. Another reported property of PSEN1 is its activity as a low conductance endoplasmic reticulum (ER) Ca 2+ leak channel with a regulatory role in pathways linked to intracellular Ca 2+ homeostasis [27][28][29][30]. Numerous studies have shown the involvement of PSEN1 in the autophagylysosome degradative pathway, which is also a function independent of γ-secretase proteolysis [31][32][33][34]. Since both the up-and down-regulation of presenilins and presenilinmediated signaling pathways, in particular Notch, may lead to various cancers [18,[35][36][37][38][39][40][41], the balanced physiological level of presenilin/γ-secretase activity is essential for normal biological function. Therefore, the direct approach for down-regulation of γ-secretase by γ-secretase inhibitors for reduction of Aβ generation may not be appropriate for AD treatment. On the other hand, suppression or modification of proteolytic activity producing Aβ with retained physiological activity of presenilin is an attractive strategy in AD therapy.
The gene family for polytopic proteins termed intramembrane di-aspartyl proteases (IMPAS) or signal peptide proteases (SPP) includes the five known paralogous genes, designated as IMP1, -2, -3, -4, -5; or SPP, SPPL2a, -2b, -2c ,-3; or PSH gene family [42][43][44]. These proteins were described as structural homologs of presenilins, registered in MEROPS database as peptidase subfamily A22B [45]. PSEN and IMP1/SPP proteases share identical evolutionarily conserved motifs for the catalytic sites YD and GxGD and the PAL domain in their C-termini ( Figure 1) [4][5][6][7]42]. Unlike PSEN1 and PSEN2, which cleave type I transmembrane proteins, IMP1/SPP proteins cleave type II transmembrane substrates with the N-terminus oriented to the cytosol [43,46,47]. Some potent γ-secretase inhibitors can efficiently work for IMP1/SPP [46]. In our previous work, we showed that human hIMP1/SPP can cleave polytopic PSEN1 protein within its membrane domain in vitro [47]. In another independent study, a signal peptide peptidase (SPP) that regulates the cleavage of specific short signal peptides in the ER was isolated [43]. The major hIMP1 isoform is identical to human SPP encoded by the same gene and orthologous to C. elegans Ce-imp-2 gene [42,43,47]. We identified a phenotype caused by inactivation of the Caenorhabditis elegans hIMP1/SPP orthologue, Ceimp-2, possibly related to a cholesterol-dependent cellular pathway [48]. In vitro assays demonstrated that hIMP1/ SPP cleaves short signal peptide remnants tethered in ER membranes. This activity may generate short signal sequences that are essential for HLA-E epitopes [43,49]. This protease also participates in processing of the hepatitis C core protein (HCV core protein, [50], ERresident tail-anchored proteins, the unfolded protein response (UPR regulator XBP1u and other proteins (data is summarized in Supplementary Figure 3).
Modulating the different activities of PSEN and IMP1/SPP-related di-aspartyl intramembrane proteases is important for a better understanding of their functions and is essential for studies aimed at dissociating their proteolytic properties. Here we created a series of constructs bearing mutations focused upon (i) the highly conserved signature GLGD; (ii) the PAL-motif that is evolutionarily invariant in both the PSEN and IMP1/ SPP protein families; and (iii) AD mutation positions of PSEN1 (Figure 1). Various mutations in the same www.impactjournals.com/oncotarget  Tables 1 and 2). We tested how the designed mutations affect the processing and functional properties of PSEN1 enzyme in intramembrane proteolysis assays comparing to similar mutations in the hIMP1/SPP protease (preliminary data were reported in [54]). In addition, for selected mutations, we performed rescue experiments in a transgenic C. elegans. To test the effects of the mutations on non-proteolytic functions of PSEN1 and whether IMP1/SPP possess such functions, we employed ER Ca 2+ leak assays using mouse embryonic fibroblasts deficient in the PSEN or IMP1/SPP genes.
We transfected PSEN1 -/-/PSEN2 -/mouse embryonic fibroblasts (MEF) cells with wild type or mutant PSEN1 constructs and examined PSEN1 auto-cleavage products via immunoblotting using PSEN1 N-terminal antibody ( Figure 2). We confirmed previous findings  [4,6] that mutations in specific aspartate residues of PSEN1 (D257A, D257E, D385A, D385E) completely abolish PSEN holoprotein cleavage as well as γ-secretase activity. We used cells over-expressing PSEN1 D385A as a dominant-negative control in all proteolytic assays. Our data on the effects of PSEN1 mutations indicate that mutations that critically change physicochemical properties (hydrophobicity) of the ultra-conserved amino acids G382, P433 and L435, strongly inhibit, but not completely abolish, the presenilinase activity. Similar amino acid substitutions G382A, P433A and L435V in the evolutionary conserved sites had much less or no effect on PSEN1 processing ( Figure 2, Supplementary Table 1). We observed similar weak inhibitory effects of the various mutations in less conserved and AD-related amino acid positions ( Figure 2). As previously described, AD-associated C410Y mutation led to a strong inhibition of PSEN1 processing ( Figure 2) [64,65].

(2) Notch1ΔE cleavage by PSEN1
Among the major PSEN substrates is a Notch family of proteins, which mediate important signaling pathways in the development and function of multicellular organisms [66]. PSEN1 has been shown to cleave NH 2terminally-truncated Notch derivates (NotchΔE) within their intramembrane domains (ε-or S3-cleavage) and to release the Notch intracellular domain (NICD), which functions as a transcriptional activator [67][68][69][70].
We co-transfected Notch1ΔE along with different PSEN1 mutant constructs in PSEN1 -/-/PSEN2 -/-MEF cells and studied the cleavage products in cell-free membrane assays (see Material and Methods). Consistent with the data from the "presenilinase" cleavage assays, mutations in highly conserved amino acids G382, P433 and L435, which affect the physicochemical properties of the protein, completely suppress Notch1ΔE ε-cleavage, whereas mutations G382A, P433A and L435V had a less inhibitory effect ( Figure 3). Among other mutations, including ADlike mutations, only C410Y substitution abolished or strongly inhibited Notch1 processing ( Figure 3).

(3) Production of carboxy-terminal fragments and Aβ peptides by APP processing
Another important type I protein proteolytically processed by presenilins is APP. Extramembrane endoproteolysis of APP by BACE produces the 99-amino acid C-terminus fragment anchored in the membrane. This C-terminus fragment is the substrate for γ-secretase cleavage. We used the transgenic construct APP695 isoform (APP695ΔNL) that bears the Swedish mutation for AD (KM670/671NL), and enhances the overall production of Aβ [71][72][73] (Supplementary Figure 2C) for co-expression with the mutant PSEN1 constructs in PSEN1 -/-/PSEN2 -/-MEF cells. Mutations in PSEN genes are responsible for changes in intramembrane γ-proteolysis events, resulting in accumulation of the amyloidogenic product (Aβ40, Aβ42) that is the major component of amyloid senile plaques in AD [74-76, reviewed in 3]. After shedding of the extracellular N-terminal part of APP, cleavage in the APP ɛ-site by γ-secretase complex generates the carboxy-terminal amyloid intracellular domain (AICD) [77]. AICD, similar to NICD, can regulate transcription of various genes, including the genes encoding EGFR, a protein tyrosine kinase up-regulated in tumors. Reduction of PSEN1 activity can regulate EGFRmediated tumorogenesis [78]. PSEN1 mutants P284A/ L/Q, G382A, L435V, P433A, T440V/D, transfected into PSEN1 -/-/PSEN2 -/-MEF cells along with APP695ΔNL, retain their proteolytic capacity to produce AICD showing various inhibitory or no inhibitory effects ( Figure 4). No APP ɛ-proteolysis products were detected when D257A/E, D385A/E, G382K/L/Q/V, P433L/Q PSEN1 mutants were transfected into the PSEN1/PSEN2 double knockout cells. A strong inhibitory effect on APP ɛ-cleavage was demonstrated for mutation in evolutionary conserved site, L435D, and for AD mutation C410Y ( Figure 4).
For the mutations in PSEN1 highly conservative sites, we also analyzed the Aβ-peptide profiles. A spectrum of Aβ peptides were immunoprecipitated from PSEN1 -/-/PSEN2 -/-MEF cells co-transfected with wild type or mutant PSEN1 and APP695ΔNL constructs. We showed that PSEN1 P433A, that retains Notch1 S3-and APP ɛ-cleavage activities, is also active in APP γ-site cleavage generating elevated level of Aβ peptides. In contrast, for the PSEN1 G382A that shows Notch1 S3-and APP ɛ-cleavage efficacy, we observed a strong inhibition of Aβ40 and other Aβ derivates ( Figure 5). PSEN1 D385A, G382L, P433L mutations lead to complete inhibition of Aβ-peptide production compared to wild type PSEN1. The weak intensity low band suggesting the specific Aβ fragment that differs from 1-42 was also observed in assays for PSEN1 G382A using high-resolution SDS-UREA-PAGE electrophoresis. This band was not observed using any other mutant constructs ( Figure

Proteolytic properties of IMP1/SPP (1) Presenilin substrate cleavage
Our previous experiments, in which we coexpressed PSEN1 holoprotein and human hIMP1/SPP in cultured cells, revealed that hIMP1/SPP is capable of cleaving a multipass transmembrane PSEN1 protein substrate [47]. We provided evidence that hIMP1/SPP may induce intramembrane proteolysis of PSEN1 in its last hydrophobic domain [47]. The significance of such cleavage activity in vivo has yet to be elucidated but a straightforward immunodetection assay has been developed to monitor proteolytic activity of IMP1 isoforms [47].
In the present study we co-transfected different IMP1/SPP mutant forms with PSEN1 substrate in HEK293 cells. Substitution mutations to structurally www.impactjournals.com/oncotarget  bearing the Swedish mutation (ΔNL) were performed in PSEN1 -/-/PSEN2 -/-MEF cells. Cell-free intramembrane cleavage was induced by incubation of the samples at 37°C for 2 hours. APP C99 and γ-secretase cleavage product AICD were detected by APP CTF antibody (Sigma). No cleavage was observed for the control samples incubated on ice (B). mAb 22C11 antibody, which detects the full-length APP (APP FL), was used as a transfection control in the assay (B). (C) Histogram shows the percent ratio of AICD to uncleaved APP C99. P-value significance codes: '*' <0.05. similar amino acids (such as G262A, G264A, P317A and L319V) lead to more efficient rescue capacities than mutations to structurally more distant amino acids in the same positions of hIMP1/SPP ( Figure 6, Supplementary Table 2). Any changes in catalytic aspartate residues and distant mutations in evolutionarily conserved sites (e.g. G262K, P317L/Q, L319D) completely or nearly completely inhibited the proteolysis of PSEN1 substrate. The G262L mutation reduced the cleavage of PSEN1 by approximately 40%. We have also shown that Ce-IMP-2, which is the C. elegans orthologue of human IMP1/SPP, can also cleave the C-terminal domain of human PSEN1 with even higher efficacy than hIMP1/SPP (Figure 7 and below for more details).

(2) HCV substrate cleavage
We next examined how wild type and mutant hIMP1/SPP cleave an HCV (hepatitis C virus) substrate ( Figure 8). hIMP1/SPP is known to cleave the immature form of HCV core protein in the hydrophobic transmembrane/C-terminal domain and this feature has become incorporated into a commonly used assay to study proteolytic properties of IMP1/SPP [50,79]. The resulting cleavage product, p21 protein, along with the envelope proteins E1 and E2, are important structural components of the virus capsid [50,80,81]. Here we cotransfected FLAG-tagged HCV core protein along with different hIMP1/SPPs and detected the cleavage products in HEK293 cells (Figure 8). There is a strong inhibition of HCV core protein cleavage by hIMP1/SPP mutations D265A, P317L, G262V, and to a lesser extent, by G262L and G262A (Figure 8). Interestingly, hIMP1 G264A and hIMP1 A318C demonstrated high proteolytic activity in the HCV cleavage assay.

Low conductance ER Ca 2+ leak assay
In addition to the protease function, presenilins may function as low conductance ion channels [27,28]. Planar lipid bilayer reconstruction techniques showed that wild type presenilin 1 can form a Ca 2+ -permeable ion channel in the ER. Interestingly, mutations in the PSEN1 catalytic aspartate D257A, which is essential for PSEN1 proteolytic function, does not disrupt ion currents; whereas the AD mutation PSEN1 M146V impaired the channel function in a dominant negative manner [27]. The critical pore structural positions were mapped to the PSEN1 hydrophobic domains 7 and 9 [28].
We next tested whether the mutations in glycine at position 382 (G382) that show variable effects on the proteolytic activities of PSEN1 and located in transmembrane domain 7 affect the cellular ER Ca 2+ leak in PSEN1 and PSEN2 double-knockout MEFs. In a series of experiments, we observed that all tested mutations in glycine in position 382, including the G382A, which retains the γ-secretase activity, were unable to reconstitute the normal ER Ca 2+ leak. In contrast, PSEN1 mutants with substitutions in proline 433 (P433A, P433L) restore the Ca 2+ leak function in PSEN1 -/-/PSEN2 -/double-knockout cells similar to wild type PSEN1 function (Figure 9). PSEN1 and IMP1/SPP are structurally related multipass proteins which cleave transmembrane domains  of Type I and Type II proteins with opposite orientations [3,43] (Figure 1). Whether the IMP1/SPP can also function as Ca 2+ channels has never been studied. We addressed this question using cultured mIMP1/SPP -/-MEFs obtained from mIMP1/SPP knockout mice that were generated in our laboratory (unpublished). We tested the Ca 2+ flow in mIMP1/SPP -/and wild type mIMP1/SPP +/+ MEF cells in comparison to PSEN1 -/-/PSEN2 -/and wild type PSEN1 +/+ MEF cells. In contrast to PSEN1 knockout cells, the absence of IMP1/SPP protein does not affect the ER Ca 2+ leak function, indicating that IMP1/SPP is not a Ca 2+ pore ( Figure 10).

In vivo rescue experiments in C. elegans
C. elegans is a popular and useful model organism for understanding the conserved mechanisms of ADrelated presenilin/γ-secretase function and in identifying the components of the γ-secretase and signaling pathways regulated by presenilins [82]. Presenilin regulation of Ca 2+ channel activities, similar to mammals, has been recently demonstrated in C. elegans [30].
In the present study, we examined the in vivo effect of the mutation in PSEN1 (G382A) that does not suppress the ɛ-proteolytic activity, but inhibits the γ-secretase proteolytic and Ca 2+ channel activities of PSEN1. We generated a C. elegans presenilin sel-12 gene construct carrying a G361A mutation that corresponds to human PSEN1 position G382. We tested whether this mutant gene construct rescues the egg-lying defect (Egl) phenotype of sel-12(ar171)unc-1 C. elegans mutant, which has a premature stop codon and inactive Sel-12/presenilin protein [83]. We found that sel-12 G361A, as well as sel-12 wild type transgene expression can efficiently rescue the egg-lying defect (Figure 11). Li and Greenwald have described that reduction of the activity of another C. elegans presenilin homolog, hop-1, can cause lethality in sel-12 mutant background strains [84]. Dead embryos/ arrested larvae phenotypes are associated in general with a reduction of lin-12/glp-1 Notch gene function, but does Histogram shows the percent ratio of cleaved HCV to uncleaved HCV core protein, detected by anti-FLAG antibody. P-value significance codes: '***' <0.001, '**' <0.01. not exclude the influence of other signaling pathways [84]. In our experiments, sel-12 G361A and sel-12 wild type strains, but not sur-5 GFP controls, can successfully survive for multiple generations being fed by hop-1 RNAi bacteria, proving the rescue effect of sel-12 G361A in C. elegans when both presenilin homologues are suppressed (data not shown). All together, the data demonstrate that ɛ-proteolytic activity, but not the Ca 2+ channel leak function, is essential for biological functions of presenilin, at least in early development and Notch-signaling.
In cultured cells, both human IMP1/SPP and Ce-IMP-2 are capable of cleaving human PSEN1 holoprotein. Ce-IMP-2 had the high proteolytic efficiency, which is consistent with a functional relationship between hIMP1 and Ce-IMP-2 orthologous proteins ( Figure 7A). To determine whether Ce-IMP-2 is capable of cleaving a C-terminal PSEN1 fragment, we expressed a C-terminal derivative (corresponding to the PSEN1 form processed by "presenilinase") with hIMP1 and Ce-IMP-2 in double knockout mouse PSEN1 -/-/PSEN2 -/cells. The efficient cleavage of the C-terminus fragment was in the same domain and likely in an identical site that we described in the PSEN1 holoprotein ( Figure 7B).
We then tested the effect of mutations in the Ce-imp-2 gene in the rescue assays performed in the tm1397 Ce-imp-2 knockout strain (Figures 12, 13; Supplementary Figure 5). C. elegans G350A and P417A mutations, corresponding to human G262A and P317A, successfully rescued the embryonic lethality on a Ceimp-2 -/background (Figure 13), although the progeny number was overall lower compared to the Ce-imp-2 wild type construct ( Figure 13). Mutations with distant physicochemical properties G350K and P417L showed no ability to rescue the mutant phenotypes compared to Ceimp-2 single and Ce-imp-1 x Ce-imp-2 double knockout strains ( Figure 14).

DISCUSSION
Uncoupling the AD-related and physiologically important functions of presenilins represents a promising approach in the search for novel strategies for AD therapy. Genetically inherited forms of PSENs bearing the AD mutations most likely do not effect the early development of the organism, but rather, accumulate deleterious effects throughout the lifetime of the individual. Less dramatic early effects of the mutations on CNS function and development, however, cannot be excluded [85,86]. Evolutionary analysis of two distantly related families of PSEN and IMP/SPP intramembrane di-aspartyl proteases shows a highly conserved amino acid signature around invariant aspartates in two transmembrane domains and the PAL motif at the C-terminal domain (Supplementary Figure 1). Here, we studied whether PSEN1 amino acid substitutions in the conserved regions of the protein have differential effects on PSEN1 proteolytic and non-proteolytic functions (Supplementary Figure 2). A similar mutational remodeling analysis was performed for the IMP1/SPP protein, which belongs to the family of evolutionarily ancient proteins distantly homologous to presenilins.
Previous in vivo and in vitro studies have shown that aspartates at positions 257 and 385 of PSEN1 play an essential role in γ-secretase proteolysis [4,6,83,[87][88][89]. These results were confirmed in our study.
The C-terminal proline 433 of PSEN1 has been described as being critical for γ-secretase function, participating in stable complex formation and catalytic pore structure [90][91][92][93][94]28]. Our data demonstrate that substitutions of the conserved proline to structurally distant, but not to similar amino acids, suppress the proteolytic activity of PSEN1. However, PSEN1 with any mutation in this position can restore the normal Ca 2+ leak function in PSEN1 knockout cells [28].
Mutation of the conserved leucine 435 of PSEN1 was found in familial cases of early onset AD with the unusual phenotype of cotton wool plaques [95,96]. In our study, substitutions to structurally distant amino acids in this position have a prominent inhibitory effect on proteolytic function of PSEN1 with complete inhibition of Notch1 proteolysis and a strong inhibition of APP ɛ-proteolysis. Presenilinase function was also inhibited by L435D mutation, but to a lesser extent. Another unusual phenotype for AD, "Lewy bodies," is associated with deletion of threonine 440 [97,98]. We have shown that both PSEN1 T440D and T440V support cleavage of various substrates.
PSEN1 with a G382A mutation in the highly conserved motif of the second aspartate does not change physiologically essential proteolytic properties of the protein. The PSEN1 G382A mutation was functionally active in presenilinase, Notch and APP ɛ-proteolytic assays. Transgenic C. elegans experiments showed a successful rescue of an egg-laying defect linked to Notch proteolysis by the G361A Sel-12 construct. Distant mutations G382 K/L/Q/V completely abolished Notch1 and APP ɛ-site proteolysis and have a strong inhibitory effect on presenilinase cleavage. The G382 position is critical for supporting normal Ca 2+ signaling, as none of the mutants at this position were able to restore normal Ca 2+ leak function (Figure 9 and [28]). The most unexpected results were obtained in the Aβ peptide assay. While PSEN1 G382L lacks both ɛ-and γ-proteolytic activities and does not produce any Aβ peptides, the PSEN1 G382A mutant retains the ɛ-cleavage activity, but significantly reduces the total production of Aβ peptides. In the reported data the occurrence of Aβ40 and Aβ42/43 fragments was detected for the G382A PSEN1 mutant in HEK293 cells expressing APP with the Swedish mutation [99]. HEK293 cells have a substantial endogenous PSEN1 activity [48]. In our study, we used a PSEN1, PSEN2 double knockout MEF model for expression of the mutant PSEN isoforms, thereby excluding any effect of endogenous PSEN activity. We also used a single-amino acid high resolution PAAG assay and noticed the minor production of a novel Aβ-peptide form that does not correspond to Aβ40 or Aβ42. We demonstrated that the PSEN1 G382A mutation retains Notch1-proteolytic activity, which has also been described in [99]. However, PSEN1 G382A was reported to block the processing of the cell membrane glycoprotein CD44 [99]. Nonetheless, our data revealed that Sel-12 G361A transgenic worms with no endogenous sel-12 presenilin activity are viable for multiple generations, providing further evidence that proteolytic activity of PSEN in processing of Notch is an essential biological function of PSEN.
Here we showed that both C. elegans Ce-IMP-2 and human hIMP1/SPP proteins are capable of cleaving the same multipass transmembrane protein (PSEN1holoprotein and -CTF) in co-transfection experiments using mammalian cells (Figure 7). Interestingly, Ce-IMP-2 induced cleavages with efficiency higher than those observed for hIMP1. Despite strong structural similarities, the Ce-IMP-2 and hIMP1 proteins have differences in some domains, particularly at the N-terminus. We also observed that human IMP1/SPP likely forms dimeric complexes, while the Ce-imp2 protein homolog does not (data not shown). Altogether, this may explain the difference in efficiency of cleavage of the same substrate between C. elegans and human IMP/SPP homologs. The Figure 13: Rescue of Ce-imp-2 knockout phenotype in C. elegans. The constructs producing Ce-IMP-2 wild type and several mutant isoforms were generated. Several independent viable transgenic strains on Ce-imp-2 -/-(tm1397) background were obtained by picking 20-30 roller phenotype animals (pRF4 was used as a co-marker) from injected Ce-imp-2 +/worm progeny. Rescue of Ce-imp-2 knockout phenotypes was analyzed in 40-61 animals for each transgenic Ce-imp-2 -/strain. Worms that can reach adult stage and have viable progeny were considered as rescued. The loss-of function Ce-imp-2 phenotype was rescued by G350A, P417A Ce-imp-2 constructs.
data also indicate that Ce-imp-2 is a true orthologue of human IMP1/SPP.
We also searched available SNP databases to identify if any mutation in PSEN1 and hIMP1/SPP that we studied can be found in common human populations. We have not identified such SNPs in more than 2,500 human genomes in the 1000 Genomes Project catalog [100]. The Ensemble variants database contains reports for PSEN1 AD-mutations M146V, C410Y and also AD-linked mutations in protein positions 284 and 435, which differ from the mutations in this position described in our paper. Interestingly, the PSEN1 D257N somatic mutation, leading to PSEN1 haploinsufficiency, has been found in lung carcinoma" (Supplementary Table 1). Amino acid substitutions in positions 245 and 319 of hIMP1 were found in malignant melanoma and endometrium carcinoma correspondingly (Supplementary Table 2). Thus, it would be of interest to elucidate further the potential role of these proteases in cancer.
In summary, we confirmed that any mutations at functionally essential aspartates of members of two distantly related families of intramembrane proteases, PSEN1 and IMP1, completely abolish all their proteolytic activities. Mutations in other highly conserved sites reduce, but do not completely suppress, at least some of the proteolytic activities of PSEN1. Substitutions to structurally distant amino-acids in the most conserved sites most dramatically change efficacy of proteolytic cleavages. Amino acid substitutions in less conserved sites (e.g. P284 and T440 in PSEN1) can reduce substrates proteolysis, but to a much lesser extent. For PSEN1, among different proteolytic assays, "presenilinase" activity is the most resistant to all mutation modifications in highly conserved sites. The PSEN1 G382A mutant retains the functionally important Notch1 and APP ɛ-cleavage activities but has reduced APP γ-cleavage activity and does not support the Ca 2+ leak function, but rescues the lethality of C. elegans Ce-imp-2 knockout. We also provided evidence that IMP1/ SPP is not a Ca 2+ channel as described for the structurally related PSEN1.
The important finding is that a single amino acid alteration in transmembrane domain 7 of PSEN1 can reduce the AD-related intramembrane γ-secretase activity but retain the biologically important juxtamembrane proteolytic function of PSEN1. These data raise the idea that uncoupling AD-and biologically-essential functions of presenilins may represent a promising novel therapeutic approach. rescue phenotype on Ce-imp-2 -/background, then ~200 single worms, synchronized mixed viable progeny of several transgenic Ce-imp-2 +/animals, were analyzed. In addition, for G350K strain #1, we checked phenotypes of Ce-imp-2 +/+ animals progeny. The rate of the losing transgenic status was estimated as percentage of non-roller wild type phenotype animals/total number of wild type phenotype animals. P-value significance codes: '**' <0.01, '*' <0.05. www.impactjournals.com/oncotarget

Transfection of mammalian cell lines and immunoblotting
Transient transfections were performed using Lipofectamine Plus Reagent (Invitrogen, Thermo Fisher Scientific) according to the manufacturer's instructions. After 24-48 hours of transfection, cells were briefly washed 2 times in cold PBS, lysed in modified RIPA-buffer (50 mM Tris-HCl, pH 7.4, 1% NP-40, 0.25% deoxcholate Na, 150 mM NaCl, 1 mM EDTA) supplemented with Complete Mini Protease Inhibitor Cocktail (Roche Diagnostics, Indianapolis, IN) for 15 min at 4°C and subsequently centrifuged at 14,000 rpm for 10 min at 4°C. A NOVEX mini-cell gel electrophoresis system (Invitrogen, Thermo Fisher Scientific) was used for protein separation and electrotransfer procedures. Cell lysates (10-20 μg of total protein) or 10 μl of conditioned culture medium were mixed with 5x Laemmli sample buffer with reducing agent (1x sample buffer composition: 60 mM Tris-Cl pH 6.8, 2% SDS, 10% glycerol, 5 % β-mercaptoethanol, 0.01% bromophenol blue) centrifuged at 12,000 rpm for 5 min with or without prior boiling for 5 min, and loaded onto SDS PAAG mini-gels (Invitrogen, Thermo Fisher Scientific). Prestained molecular weight markers (Invitrogen, Thermo Fisher Scientific) were loaded into a separate well. Electrophoresis was run either in 10% (PSEN1, HCV detection) or 8% (Notch detection) Tris-Glycine PAAG or in Novex ® 10-20% Tricine Gel (for APP-CT detection) in 1X SDS running buffer at 125 V. After electrotransfer, Immobilon-P PVDF membranes (EMD Millipore, Billerica, MA USA) were washed in TBS-T buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.05% Tween 20) 1 time for 10 min and 3 times for 5 min, incubated in blocking buffer (5% milk in TBS-T) at room temperature for 1 hour and placed in 10 ml of hybridization buffer (1% milk in TBS-T), containing primary antibodies in 1:500 -1:5000 dilution, at 4°C overnight. After incubation with appropriate secondary antibodies, signal visualization was performed using an ECL Western blotting detection reagent kit and exposure to an X-ray film or via image capturing using a VersaDoc 5000 imaging system (Bio-Rad Laboratories, Hercules, CA). Experiments were replicated multiple times ( ≥3).
Polyclonal rabbit antibodies against hPSEN1 N-terminal polypeptide and CTF were described previously [47]. Antibody against c-myc epitope for IMP1 and Notch detection was obtained from Invitrogen (Thermo Fisher Scientific). The released APP intracellular fragments were detected in supernatants using polyclonal APP CTF antibody (Sigma-Aldrich). For detection of Aβ peptides, mAb 1E8 (Bayer-Schering Pharma AG, Berlin, Germany or provided by T. Dyrks, Schering, Berline, Germany) was used while mAb 22C11 was used for detection of N-terminal APP fragments (Chemicon, EMD Millipore). For HCV detection, monoclonal mouse anti-FLAG (Sigma-Aldrich) was used.

Cell-free intramembrane cleavage assay
For intramembrane cleavage assays, PSEN1 -/-/PSEN2 -/-MEF cells were transfected with either APP695ΔNL or NotchΔE and one of the PSEN1 isoforms or mock plasmid. After 48 hours of transfection, cells were washed twice with ice-cold PBS, scraped off the dishes, resuspended in 0.5 ml of hypotonic homogenization buffer (10 mM HEPES, pH 7.2; protease inhibitors cocktail), incubated on ice for 1 hour and frozen in liquid nitrogen. Frozen cells were thawed on ice for 1 hour and homogenized by passing through a 22-gauge hypodermic needle 5 times. Homogenates were centrifuged at 1000 at 4°C for 10 min and post-nuclear supernatants were saved. To isolate crude membranes, postnuclear supernatants were subjected to centrifugation at 20000 g, 4°C for 45 min and the pellets were resuspended in assay buffer (150 mM Sodium Citrate, pH 6.4, 5 mM EDTA, protease inhibitors cocktail). Intramembrane cleavage was induced by incubation of the samples at 37°C for 2 hours. Reactions were stopped by placing samples on ice, and membranes precipitated by centrifugation at 16000 g, 4°C for 30 min. The released cleavage fragments were detected in supernatants and analyzed by immunoblot.

C. elegans experiments
Standard methods of C. elegans handling and culture were applied [104]. N2 Bristol strain was used as the wild type. Sel-12(ar171)unc-1 (e538) was provided by I. Greenwald (Columbia University, New York, USA). Tm827 (Ce-imp-1 knockout), tm1397 (Ce-imp-2 knockout) and tm1654 (Ce-imp-3 knockout) were provided by the Mitani Laboratory, Tokyo Women's Medical University School of Medicine, Japan and outcrossed 3-5 times to the N2 strain. A C. elegans ~6 kb sel-12 genomic fragment containing ~3 kb of the 5'-and ~0.8 kb of the 3'-regulatory regions were cloned into the L4440 vector. Ce-imp-2 constructs are described in Supplementary Figure 5. Mutations in the sel-12 and Ce-imp-2 genes were introduced using a QuikChange ® Site-Directed Mutagenesis Kit (Stratagene, Agilent Technologies). RNA interference (RNAi) by dsRNA feeding and germ-line injections were performed as previously described [48]. Sur-5 GFP or pRF4 plasmids were used as germ-line co-injection markers. Ce-imp knockout and transgenic strains were genotyped accordingly (Supplementary Figure 5, Supplementary Table  3). The progeny of the heterozygous deletion mutant Ceimp-2 were analyzed by single worm PCR.

Statistical analysis
Plotted values represent mean ± standard error of the mean. Comparisons of more than two groups were carried out using one-way ANOVA and unpaired T-test against WT values as the control group (Figure 2-4, 6,  8). For the analysis of total Aβ in PSEN1 mutants ( Figure  5) and C. elegans experiments (Figure 13-14) WT values were converted to Z-score using mutant data and two-