Influence of the familial Alzheimer's disease–associated T43I mutation on the transmembrane structure and γ-secretase processing of the C99 peptide

Extracellular deposition of β-amyloid (Aβ) peptides in the brain is a hallmark of Alzheimer's disease (AD). Upon β-secretase–mediated cleavage of the β C-terminal fragment (β-CTF) from the Aβ precursor protein, the γ-secretase complex produces the Aβ peptides associated with AD. The familial T43I mutation within the transmembrane domain of the β-CTF (also referred to as C99) increases the ratio between the Aβ42 and Aβ40 peptides largely due to a decrease in Aβ40 formation. Aβ42 is the principal component of amyloid deposits within the brain parenchyma, and an increase in the Aβ42/Aβ40 ratio is correlated with early-onset AD. Using NMR and FTIR spectroscopy, here we addressed how the T43I substitution influences the structure of C55, the minimal sequence containing the entire extracellular and transmembrane (TM) domains of C99 needed for γ-secretase processing. 13C NMR chemical shifts indicated that the T43I substitution increases helical structure within the TM domain of C55. These structural changes were associated with a shift of the C55 dimer to the monomer and an increase in the tilt of the TM helix relative to the membrane normal in the T43I mutant compared with that of WT C55. The A21G (Flemish) mutation was previously found to increase secreted Aβ40 levels; here, we combined this mutation in the extracellular domain of C99 with T43I and observed that the T43I/A21G double mutant decreases Aβ40 formation. We discuss how the observed structural changes in the T43I mutant may decrease Aβ40 formation and increase the Aβ42/Aβ40 ratio.

The extracellular deposition of ␤-amyloid (A␤) 4 peptides in the brain is a prominent feature of Alzheimer's disease (AD) and is caused by proteolysis of the amyloid precursor protein (APP) via a sequential two-step process. Cleavage of APP by ␤-secretase near the boundary between the ectodomain and extracellular juxtamembrane sequence causes shedding of the ectodomain and is required for a second cleavage within the single transmembrane (TM) helix of APP (1). The ␥-secretase complex catalyzes the second cleavage, which leads to the formation of the A␤ peptides with lengths ranging from 38 to 42 amino acids. A␤40 is the most prevalent A␤ peptide. However, A␤42 has a higher propensity to form aggregates than the shorter isoforms and is the most toxic peptide generated by ␥-cleavage (2). A␤42 is also the principal component of amyloid plaques in AD patients (3).
Familial AD mutations occur in both the ␥-secretase complex and in the APP substrate. The first mutations identified in APP were at Ala 21 , Glu 22 , and Asp 23 within a cluster in the extracellular region of the protein. (In the following, we use the ␤-CTF numbering, which also coincides with the numbering of the A␤ peptides (i.e. Asp 1 is the first residue of the ␤-CTF and the A␤ peptide).) These mutations included the A21G Flemish, E22Q Dutch, E22G Arctic, E22K Italian, and D23N Iowa mutations (4 -7). Within this cluster, mutations have been found to influence different stages of the conversion of APP to amyloid fibrils. For example, the A21G mutation increases the total amount of secreted A␤ by enhancing ␥-secretase processing of the ␤-CTF (8,9). In contrast, mutations at Glu 22 and Asp 23 increase the rate of fibril formation following the formation of the A␤ peptides (10) but do not influence proteolysis by ␥-secretase (9). Over the past few years, the mechanisms by which these mutations influence the proteolysis of APP or the conversion of A␤ monomers to fibrils have begun to emerge (11,12).
A second cluster of amino acids in the ␤-CTF that are mutated in familial AD occurs between the cleavage site (Ala 42 ) that releases the A␤42 peptide and the cytoplasmic end of the TM domain (Leu 52 ). These include the T43A Iranian (13), T43I Austrian (14), V44M French (15), V44A German (16), I45V Florida (17), V46I London (18), V46L Indiana (19), and V46G (20) and L52P Australian (21) mutations. Gorman et al. (22) measured the influence of three FAD mutations (T43I, V46G, and V46F) on the monomer-dimer equilibrium in the context of TM peptides. They found a correlation between the dissociation constant for the TM peptide dimer and the A␤42/A␤40 ratio observed in AD patients, suggesting that processing of the APP monomer results in an increase in A␤42. More recent studies indicated that mutations in the four-residue stretch from Thr 43 to Val 46 decreased homodimerization and increased the A␤42/A␤40 ratio in the context of the ␤-CTF (11). However, it is not clear how four sequential residues in the C99 sequence can decrease homodimerization because only two residues at most might lie in the dimer interface.
Here, we focus on the Austrian T43I mutation, which is located in the TM cluster of mutations to establish the structural changes that occur upon mutation. Recent NMR studies suggest that the region near the ␥-secretase cut site does not form a stable ␣-helix in the WT protein (23). One possibility is that it is unraveled to facilitate cleavage in the dimer or binding to the ␥-secretase complex. It is known that the TM helix must unravel to undergo proteolysis. We propose that in the monomer, backbone hydrogen-bonding interactions favor ␣-helix formation, which impedes enzyme-substrate interactions, substrate unraveling, and proteolysis.
T43I causes the largest increase in the A␤42/A␤40 ratio within the cluster of TM FAD mutations (14). Thr 43 is one amino acid after the ␥-cleavage site that generates the A␤42 peptide. This site is roughly midway between the extracellular and intracellular boundaries. On the extracellular side of Thr 43 , there are several GXXXG motifs that mediate dimerization of the TM domain in membrane bilayers (24) and have been implicated in cholesterol binding (25). On the intracellular side of Thr 43 is a cluster of ␤-branched amino acids (Val 44 , Ile 45 , and Val 46 ) whose mutation results in early onset AD.
To better understand the influence of the T43I mutation on ␥-secretase processing, we also combine it in a single construct with the A21G mutation located in the extracellular cluster of FAD mutants. We have previously shown that the A21G mutation lengthens the TM helix, stabilizes the TM dimer, and destabilizes ␤-sheet structure in the extracellular region of the ␤-CTF (8,12). These structural changes are associated with an increase in the total amount of A␤ peptide released upon ␥-secretase cleavage as well as with an increase in the A␤42/ A␤40 ratio (8,9). By combining an extracellular FAD mutation (A21G) and a TM FAD mutation (T43I), we can probe whether the two clusters of FAD mutations act independently or synergistically in ␥-secretase processing.
We characterize the influence of the T43I and A21G mutations on the structure of C55, a 55-residue peptide that includes the extracellular and TM regions of C99. We have previously shown that the extracellular and TM structures are similar in C55 and C99 and that differences in the amounts of A␤38, A␤40, and A␤42 produced using WT or selected mutant sequences are similar (8). The similarity in processing likely stems from the fact that the intracellular domain of C99 is unstructured and removed in the initial ⑀-cleavage of C99, yielding a product that is the same as C55 after ⑀-cleavage.
Structural measurements using solution and solid-state NMR spectroscopy are carried out in detergent micelles and in membrane bilayers, respectively. Chemical shift measurements upstream and downstream of the point of mutation provide evidence for global changes in structure, whereas measurements of interhelical dipolar couplings assess the dimerization state of WT C55, the T43I mutant, and the A21G/T43I double mutant of C55. FTIR measurements also provide information on the global secondary structure of these mutants in membrane bilayers.

Mutations at Thr 43 change the production and ratio of A␤42 to A␤40
There are two mutations at Thr 43 that have been described in the literature, the T43A Iranian (13) and T43I Austrian mutations (14). These introduce different size side chains at position 43, and both result in early onset AD. To better understand the dependence of amino acid type at position 43 on processing, we analyzed the distribution of secreted A␤38, A␤40, and A␤42 produced by the expression and processing of WT C99 and a series of Thr 43 mutants in Chinese hamster ovary (CHO) cell cultures, which are widely used as they generally produce more A␤ peptides through ␥-secretase processing than primary neurons.
The mutants within the TM of C99 selected for study comprise substitutions of Thr to hydrophobic or weakly polar amino acid residues (Fig. 1). The amino acid substitutions chosen are conservative and transfected DNA constructs express at comparable levels in CHO cells. The observed FAD mutations in cluster 2 tend to all be conservative and involve substitutions with different hydrophobic side chains. This observation suggests that nonconservative mutations prevent C99 processing. The overall production of A␤ peptides is generally decreased by substitution of Thr 43 . The familial AD Thr 43 mutations, T43I and T43A, exhibit decreased total A␤ production (i.e. A␤38, A␤40, and A␤42) by ϳ65 and 45%, respectively. Mutation of T43I and T43V dramatically decreases the level of A␤40 and increases A␤42. The A␤42/A␤40 ratio is highest for T43I. These results are dramatically different from those of the A21G mutant and other sites in the adjacent LVFF sequence in which mutation leads to an increase in total A␤ production (8,9).
Serine and valine are perhaps the most similar amino acids to threonine. Both have roughly the same molecular volume as threonine. Serine has a ␤-hydroxyl group, whereas valine has a ␤-methyl group. The T43S mutation results in a similar loss of total A␤ production (ϳ50%), as observed for T43I and T43A,

Structure of the T43I mutant of C99
but has almost no effect on the A␤42/A␤40 ratio. The T43V mutation closely resembles T43I in total production and has the second highest increase in the A␤42/A␤40 ratio. Together, these results suggest that loss of the ␤-hydroxyl group of Thr 43 has a crucial influence on the A␤42/A␤40 ratio.
Of the amino acids tested, glycine and phenylalanine are structurally the most different from threonine. A␤ production is blocked by the T43F mutation, which places a large hydrophobic residue at position 43, whereas T43G exhibits a striking increase in A␤38 relative to A␤40 and A␤42.
Overall, the changes in ␥-secretase processing are not correlated with the size or character of the amino acid at position 43, other than loss of the ␤-hydroxyl group. The T43F mutant exhibited the lowest amount of soluble A␤40 of the mutants tested, whereas T43M generated a level of A␤40 close to WT. It is interesting to note here that phenylalanine substitutions at Val 44 , Ile 45 , Val 46 , Ile 47 , and Thr 48 have very different influences on total A␤ and the A␤42/A␤40 ratio (26). One general correlation of the Thr 43 mutants is a decrease in the total A␤ produced as observed previously for T43A and T43I (27). A large decrease in A␤40 secretion is the primary cause of an increase in the A␤42/A␤40 ratio for the T43I mutation.

Comparison of WT C55 and Thr 43 mutants reveals global changes in structure
C55 is the minimal sequence containing the entire extracellular and TM domains of C99 needed for ␥-secretase processing (8). To assess the global structure of C55, we undertook solution NMR measurements in dodecylphosphocholine (DPC) detergent micelles. These studies parallel those in DPC micelles on the full ␤-CTF molecule (residues 1-99) (25, 28), a truncated ␤-CTF (residues 15-53) (29), and TM domain alone (residues 28 -55) (30). Fig. 2 presents the 2D 15 N-1 H HSQC spectrum of detergent-solubilized C55. This construct (designated C55*) contains the full-length A␤ extracellular sequence, the APP transmembrane domain, and a few residues in the juxtamembrane sequence (KKK) of APP. These were cloned into a pET21a vector with an added N-terminal methionine and an attached linker/His 6 tag (KLAAALEHHHHHH) at the end of the sequence for purification. We have previously shown that WT C55 and the C55* exhibit similar FTIR spectra, suggesting that the His tag is unstructured (8). In C99, the amino acids following the KKK sequence at the C terminus are unstructured and extend into the cytosol (28). Consequently, in both C55 and C99, the hydrophilic C terminus is not expected to influence the structure of the TM and extracellular domains.
The HSQC spectrum exhibits roughly one well-resolved resonance of equal intensity for each backbone and side-chain NH group. The assignments were made on the basis of a series of 3D NMR experiments and are in agreement with those determined for the full-length ␤-CTF (25, 28) and a truncated ␤-CTF (29).
The chemical shifts of the glycine residues provide a potential probe of changes in C55 structure or dynamics. The GXXXG motifs mediate TM dimerization, whereas the Gly 37 -Gly 38 sequence is thought to form a hinge facilitating substrate These constructs express at comparable levels. The secretions of A␤38 (A), A␤40 (B), and A␤42 (C) were measured by multiplex 4G8 or 6E10 A␤ electrochemiluminescent assays (MesoScale Discovery, Gaithersburg, MD). The A␤42/A␤40 ratio (C) increased 10.5-and 7.2-fold in T43I and T43V mutants, respectively, relative to the WT peptide. The A␤42/A␤40 ratio also increased 4.3-and 5.2-fold in the T43G and T43L mutants, respectively, compared with WT. Overall, the mutations on Thr 43 increase the A␤42/A␤40 ratio by decreasing the A␤40 production. The statistics are based on a one-way analysis of variance using a Dunnett post hoc test (all columns are compared with WT). The number of samples was as follows: n ϭ 4 for WT; n ϭ 3 for mutants. The different samples (n) correspond to biological replicates in different cultures. In each culture, we pooled n ϭ 3 (different wells) for each condition (WT, mutants). *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. a.u., arbitrary units.

Structure of the T43I mutant of C99
flexibility (31). Fig. 3 presents the region of the 2D 15 N-1 H HSQC spectra containing the glycine NH resonances of WT C55* and a series of Thr 43 mutants in DPC micelles. The mutations include T43A, T43V, T43I, T43M, T43F, and T43G. The striking observation is that all of the mutations influence the Gly 37 and Gly 38 resonances within the TM sequence. The T43V and T43I mutants, which involve a change to another ␤-branched amino acid, result in only changes to Gly 37 and Gly 38 . The T43M and T43F mutants, which involve changes to large hydrophobic residues that are either flexible or relatively rigid, respectively, lead to additional shifts in Gly 33 . The T43G mutation is the most influential, leading to changes in all of the glycine residues except the N-terminal Gly 9 resonance.
We previously observed differences between the structure of the extracellular domain of C55 in detergent and membrane bilayer environments (8). The hydrophobic 17 LVFF 20 sequence adopts helical secondary structure in detergent micelles, whereas it contributes to a small ␤-sheet domain that binds to the bilayer surface in a bilayer environment. The structure of the TM domain may be less sensitive to the change between detergent micelles and membrane bilayers than the extracellular sequence because the TM domain in both detergent micelles and bilayers resides within a hydrophobic environment. To probe if similar chemical shift changes occur at Gly 37 and Gly 38 in the T43I structure in bilayers, solid-state NMR measure-ments were undertaken by incorporating specific 13 C labels into C55 with solid-phase peptide synthesis. The C55 peptide was reconstituted into dimyristoyl-phosphocholine (DMPC), dimyristoyl-phosphoglycerol (DMPG) membrane bilayers to characterize the chemical shift changes between WT and T43I at residues Gly 33 , Gly 37 , and Gly 38 , as well as residues 39 -42 and 44 -45 that bracket position 43 in the sequence. The NMR chemical shifts of the 13 C-labeled resonances were made using one-dimensional magic angle spinning (MAS) NMR. Table 1 summarizes the 13 C chemical shifts of the specific labeled amino acids within the TM region of C55. The glycine 13 C␣ chemical shifts are in the range of 40 -50 ppm, whereas 13 CϭO chemical shifts are in the range of 174 -180 ppm. Chemical shifts at the high ends of these ranges generally correlate with helical secondary structure, whereas chemical shifts at the lower ends of these ranges generally correlate with ␤-sheet secondary structure. The higher chemical shifts for both the 13 CϭO and the 13 C␣ resonances in the TM glycine residues in the T43I mutant suggest that the structure of the TM domain is more helical than in WT C55. The increase in helical structure is illustrated by the plotting the chemical shift differences relative to random coil chemical shifts (Fig. 4), where the more positive values in the T43I mutant relative to WT C55 reflect an increase in helical structure. There are larger changes in Gly 37 and Gly 38 compared with Gly 33 . Both 13 C resonances for Gly 37 and Gly 38 shift upfield by ϳ2 ppm. The large changes in Gly 37 and Gly 38 upon mutation agree with the solution NMR results showing that these glycines are the most sensitive to mutation.
The 13 CϭO and 13 C␣ chemical shifts of Val 40 , Ile 41 , Ala 42 , Val 44 , and Ile 45 on the N-terminal side of Thr 43 also exhibit changes characteristic of an increase in helical structure upon mutation (Fig. 4). The only exceptions are the backbone 13 CϭO of Val 40 and the 13 C␣ of Ile 45 . The Val 40 13 CϭO has an unusually high chemical shift in WT C55 and is in a position to backbone hydrogen-bond to the Thr 43 C␤-OH. Loss of this interaction upon the T43I mutation would be expected to result in a lower frequency.
Together, the solution and solid-state NMR measurements suggest that global changes occur in the structure of C55 upon introduction of the T43I mutation. The chemical shift changes of the backbone 13 C resonances within the middle of the TM helix in the solid-state NMR experiments suggest that the helix is distorted in WT C55 and adopts a more canonical helical structure in the T43I mutant. We show below that the T43I mutation results in a shift in the monomer-dimer equilibrium toward monomer. Fig. S1 presents overlays of the full HSQC spectra of the T43I and T43A mutants that illustrate large changes occur throughout the sequence, consistent with a dimer-to-monomer transition.

The T43I mutation alters the monomer-dimer equilibrium of C55
The dimer structure of the TM domain of APP has been reported in both detergent micelles (29,30) and in membrane bilayers (24). In DMPC:DMPG and POPC:POPS bilayers, the TM domain dimerizes via the G 29 XXXG 33 XXXG 37 interface (24). In contrast, the 38 GxxxA 42 sequence was found to mediate dimerization in DPC detergent micelles (29). The GXXXA

Structure of the T43I mutant of C99
interface is the same as that predicted computationally in an earlier study by Gorman et al. (22), who proposed a comprehensive model for the influence of FAD mutations in cluster 2. In their studies using TM peptides, they showed that mutation of Thr 43 shifts the structure of the TM dimer toward monomer. To test the dimer-to-monomer equilibrium in WT C55 and in the T43I mutant, 2D solid-state NMR experiments were undertaken. The experiments were designed to establish whether either the GXXXG or GXXXA dimer interface exists in the T43I mutant. These experiments involve co-mixing of two C55 peptides containing different 13 C-labels and probing interhelical magnetization transfer. Specifically, 13 C labels were selectively incorporated at Gly 33 and Ala 42 . Peptide 1 of the co-mixture contains 13 C-carbonyl-labeled Gly 33 and 13 C␤-labeled Ala 42 , whereas peptide 2 contains 13 C␣-labeled Gly 33 and 13 C-carbonyl-labeled Ala 42 . In the C55 dimer, the Gly 33 residues are expected to be closely packed if the 33 GXXXG 37 motif mediates dimerization, whereas the Ala 42 residues are in close proximity if the 38 GXXXA 42 motif mediates dimerization. Fig. 5 presents 2D NMR measurements of potential GXXXG and GXXXA contacts using dipolar assisted rotational resonance (DARR), a solid-state NMR approach for determining through-space 13 C . . . 13 C distances. Close proximity of 13 C labels (Ͻ6 Å) is manifested as a cross-peak in the 2D NMR spectrum. Fig. 5A shows the full 2D spectrum, where the region containing cross-peaks is boxed in red. For the WT C55 peptide, the cross-peak between the 13 C␣ carbon of Gly 33 at 46.9 ppm and the 13 CϭO carbon at 174.2 ppm is observed (red box). A row through this region of the spectrum clearly reveals this cross-peak (Fig. 5B) and is consistent with a 33 GXXXG 37 motif mediating dimerization. There is no cross-peak between the Ala 42 positions (8). In contrast to the WT sequence, incorporation of the same 13 C labeling scheme into peptides containing the T43I mutation does not result in interhelical cross-peaks for either the GXXXG interface or the GXXXA interface (Fig. 5C). We interpret the absence of cross-peaks for either interface as a shift of the GXXXG dimer to monomer, consistent with the previous FRET measurements of Gorman et al. (22).

Global structural changes in helix orientation and the extracellular ␤-sheet
FTIR measurement of the frequency of the amide I vibration of TM peptides provides information on secondary structure, whereas measurement of the dichroic ratio provides information on the orientation of TM helices. Fig. 6 presents FTIR spectra of WT C55* and a series of Thr 43 mutations obtained

Structure of the T43I mutant of C99
by expression. The peptides were reconstituted into DMPC: DMPG membrane bilayers. The amide region of the spectrum exhibits an intense band at 1730 cm Ϫ1 corresponding to the acyl chain carbonyls of the membrane lipids. The band at 1657 cm Ϫ1 is characteristic of ␣-helical global secondary structure, whereas the band at 1626 cm Ϫ1 is characteristic of ␤-sheet secondary structure. The small 1695 cm Ϫ1 band is often associated with anti-parallel ␤-sheet. We have recently assigned the ␤-structure in C55 to the extracellular sequence between Tyr 10 and Ala 21 (12). Upon mutation of Thr 43 , there is a general increase in resolution and an increase in intensity of the 1626 cm Ϫ1 band. These changes indicate that mutations in the TM domain have an influence on the extracellular region and suggest that the TM mutations induce a global change in structure.
Measurement of the dichroic ratio of the amide I normal mode using FTIR spectroscopy provides a probe of helix orientation relative to the membrane bilayer normal ( Table 2). The amide I vibration is dominated by the stretching vibration of the backbone CϭO group, which has an orientation roughly parallel to the helix axis. Polarized IR light is preferentially absorbed when the polarization is aligned parallel to the CϭO bond. The dichroic ratio measures the intensity of absorbed IR light with orientations of 90 and 0°relative to the surface of the IR plate (see "Experimental procedures"). Polarized FTIR measurements were made on the TM sequence alone (residues 28 -55) or on C55 (residues 1-55). The TM peptide has a high dichroic ratio of 3.5, consistent with an orientation that is nearly parallel to the bilayer normal. The TM domain of C55 exhibits a slightly  . Dimer contacts between TM helices of C55 probed using 2D DARR NMR spectroscopy. DARR NMR spectra were obtained of C55 peptides corresponding to the WT sequence (A and B) T43I (C), A21G (D), and A21G/T43I (E). A, full 2D spectrum. B-E, rows through the boxed region in A. For these measurements, 13 C-carbonyl and 13 C-C␣ labels were separately incorporated at Gly 33 in two C55 peptides. The peptides were co-mixed in equimolar ratios and reconstituted into membrane bilayers (A-E). Cross-peaks are present as a result of interhelical dipolar couplings. To rule out the possibility that the cross-peaks observed originate from 13 C labels within a single peptide (i.e. intrahelical dipolar couplings between the Gly 33 C␣ and Ala 42 CϭO), we also reconstituted the peptides individually into membrane bilayers and ran parallel experiments. Neither peptide that we used in our co-mixture exhibited intramolecular crosspeaks when measured separately.

Structure of the T43I mutant of C99
smaller dichroic ratio (3.3). For the T43I, the dichroic ratios of the TM peptide and C55 mutants are less than the corresponding dichroic ratios of the WT peptides, indicating a tilt away from the bilayer normal upon mutation.
As a result, the FTIR and NMR measurements suggest that mutation of Thr 43 in the middle of the TM helix causes a change in the structure of C55 upstream of the site of mutation. A shift of the TM dimer to monomer would result in changes in the Gly 33 , Gly 37 , and possibly Gly 38 chemical shifts and allow the helix to tilt more in the bilayer.

The T43I mutation has a stronger influence on processing than A21G in the A21G/T43I double mutant
We have recently characterized the structure and influence on proteolysis of the A21G Flemish mutation in the context of the C55 peptide (8). Compared with WT C55 (or C99), the A21G mutation increases A␤ production by ␥-secretase. This mutation reduces ␤-sheet structure of the extracellular sequence of C55 and increases ␣-helical structure from Gly 25 to Gly 29 . The latter corresponds to a region near the membrane surface and thought to interact with cholesterol. In addition, the A21G mutation appears to slightly favor C55 dimerization (8). These effects are opposite to those seen for the T43I mutant.
To compare the relative influence of the A21G and T43I mutations, we examined processing of the ␤-CTF in the single mutants and together as a double mutant (Fig. 7). We individually compared the production of A␤38, A␤40, and A␤42. The most significant changes are decreases of the amount of A␤40 and increases in the amount of A␤42 in both the single T43I mutant and the double A21G/T43I mutant. The relative concentrations of both A␤38 and A␤42 increase in the double mutant, A21G/T43I, relative to the single mutants.
The net result is that the T43I mutation has a stronger influence on processing than A21G in the A21G/T43I double mutant. Solid-state NMR of the double mutant of C55 indicates that the protein is monomeric (Fig. 5E), suggesting that the conversion of dimer to monomer is driving the low production of A␤40 and high production of A␤38 in the double mutant. The individual mutations combine in an additive or synergistic fashion to increase the amount of A␤42 in the double mutant.

Coupling of the T43I mutation to changes in the extracellular region of C55
The results above suggest that the influence of the T43I mutation on the extracellular region of C55 occurs via a shift toward a monomer. It has previously been shown that the extracellular sequence of C55 folds into ␤-sheet secondary structure and that this structure inhibits the production of soluble A␤. The FTIR results in Fig. 6 show that the ␤-sheet component of C55 increases in T43I and the other Thr 43 mutants, consistent with the overall decrease in soluble A␤. We have recently shown using fluorescence spectroscopy that the hydrophobic 17 LVFF 20 motif within extracellular ␤-sheet structure is inserted in bilayer and Phe 19 penetrates into the headgroup region of the bilayer (12).
In this section, we compare how the mutation of A21G and T43I influence membrane interactions using fluorescence spectroscopy of WT C55 and the A21G, T43I, and T43I/A21G mutants. In each of these four peptides, Phe 19 has been changed to tryptophan, whose fluorescence is highly sensitive to environment. Membrane-embedded tryptophan generally exhibits a more intensive, blue-shifted fluorescence emission band than tryptophan in solution. This substitution does not change the ␤-sheet component of C55 in the context of the WT peptide. In the current studies, we find that the fluorescence intensity increases and there is a slight blue shift in the fluorescence maximum in the T43I and T43I/A21G mutants relative to WT C55 (Fig. 8). In contrast, the fluorescence intensity decreases, and there is a slight red shift in the fluorescence maximum the A21G C55. These changes are consistent with a more membrane-buried tryptophan at position 19 in the context of the T43I mutant. We discuss below the potential effects of an increase in the ␤-sheet component in the T43I mutant.

Discussion
Familial mutations in the APP gene fall into several clusters. One cluster is within the TM domain of the protein between Ala 42 and Leu 49 , the ␥and ⑀-cleavage sites, respectively. This cluster includes amino acids Thr 43 , Val 44 , Ile 45 , and Val 46 . We have targeted the T43I mutation and have shown that it induces global structural changes in the extracellular region of C55 as well as local changes in the region of Gly 37 and Gly 38 . These changes are associated with a shift in the monomer-dimer equilibrium toward more monomer as first observed by Chakrabartty and co-workers (22) and an increase in helical secondary structure near the ␥-cleavage site. An analysis of the cleavage products resulting from ␥-secretase processing of C99 shows that the T43I mutation reduces the total amount of secreted A␤ peptides mainly due to a decrease in A␤40, while increasing the A␤42/A␤40 ratio. We hypothesize that the increase in helical secondary structure in monomeric A␤ peptide impedes enzyme-substrate interactions, substrate unraveling, and proteolysis.

The T43I mutation causes structural changes upstream of Thr 43
Intramembrane cleavage of substrates by the ␥-secretase complex is unusually slow (32,33). Moreover, it has not been established what distinguishes substrates from nonsubstrates other than the apparent requirement of extracellular domain shedding. For APP, the major ␥-secretase cleavage sites within the ␤-CTF are at residues 38, 40, and 42. The TM cluster of FAD mutants occurs in the helical turn of amino acids that are just downstream of Ala 42 . De Jonghe et al. (34) found that T43I, V44M, V44A, I45V, V46I, and V46L all increase the A␤42/ A␤40 ratio. Importantly, these mutations are all relatively conservative, involving substitutions with different hydrophobic side chains. This observation suggests that nonconservative mutations prevent C99 processing. All mutations except I45V decreased A␤40 secretion, and all mutations except V44M increased the amount of secreted A␤42. These results suggested that there are local changes within the ␤-CTF substrate in the vicinity of the Val 40 and Ala 42 cleavage sites.
Moreover, there are several observations that imply that the FAD mutations within the TM domain can influence the structure of the ␥-secretase substrate upstream of Thr 43 . For example, De Jonghe et al. (34) and Ancolio et al. (15) found that mutations within the TM cluster of FAD mutants increased x-A␤42, where x-A␤42 are peptide isoforms that start with cleavage at amino acid Glu 11 (alternative ␤-secretase cleavage) (1) or Leu 17 (␣-secretase cleavage) (15,35).

Structure of the T43I mutant of C99
Our results provide support for structural changes both upstream (this section) and downstream (discussed below) of Val 40 or Ala 42 , the ␥-cleavage sites. We further show that these changes are associated with a shift of the WT C55 dimer to monomer.
Both solution and solid-state NMR studies show that mutations at Thr 43 influence the structure of the ␤-CTF in the region of the TM glycine residues (Gly 33 , Gly 37 , and Gly 38 ), upstream of the ␥-cleavage site. These glycines have been implicated in both dimerization of the TM domain and flexibility of the TM helix. For example, Munter et al. (36) found that mutations of Gly 29 and Gly 33 of the GXXXG motif gradually attenuate the strength of TM dimerization, reduce the formation of A␤42, and increase A␤38 and shorter A␤ species. The correlation between a reduction in dimerization and an increase in secreted A␤38 is similar to the results reported above.
One possibility is that dynamics of the monomer hinders the ␥-secretase cleavage of C99 particularly to A␤40, and leads to a more equal distribution of A␤ species. Recent studies (37) indicate that the cluster 2 FAD mutations, including T43I, decrease the stability of the ␥-secretase-substrate complex, consistent with a change in the structure or dynamics of the substrate.
The double glycine sequence (Gly 37 -Gly 38 ) has also been implicated in forming a kink in the TM domain (31). Computationally, residues Val 40 -Ala 42 were found to be less ␣-helical than at the ⑀-cut site, and the side-chain hydroxyl of Thr 43 was found to hydrogen bond to the backbone CϭO of the i Ϫ 4 residue in the TM domain (i.e. Val 39 ) (31). Structurally, Bechinger and co-workers (23) found that the TM helix is disrupted around Ala 42 and that this break provides flexibility for increased topological heterogeneity of the C-terminal domain. Their data are consistent with Gly 37 mediating dimerization, with a kink or break in the TM helix occurring below this point in the sequence. We find that Gly 37 mediates TM helix dimerization in the WT peptide and that the region at and below Ala 42 (i.e. Ala 42 -Val 44 -Ile 45 ) becomes more helical in the T43I mutant, consistent with their observations. Moreover, the disruption of the TM helix in the region of Ala 42 may lead to hydrogen bonding of the Thr 43 ␤-OH to the backbone CϭO of Val 40 . This interaction may explain the unusual chemical shift of the 13 CϭO group of Val 40 and the strong influence of the Thr 43 ␤-OH in modulating ␥-secretase processing.
The FTIR data presented in Fig. 2 also provide evidence for a change in the structure of the extracellular region of C55 upon mutation of Thr 43 . We observe reproducible changes in the intensity of the 1626 cm Ϫ1 band assigned to ␤-sheet structure between Tyr 10 and Ala 21 . This structured region influences the level of A␤40 secretion by influencing ␥-secretase cleavage (8).
In particular, mutations of the 17 LVFF 20 sequence disrupt this extracellular ␤-sheet and are correlated with an increase in the secretion of A␤ peptides. Here, we observe the opposite effect; namely the T43I, T43A, T43G, and T43S mutants exhibit enhanced ␤-sheet in the extracellular domain of C55 and a Figure 8. A, fluorescence spectroscopy of the F19W mutation reveals changes in membrane association of the extracellular ␤-sheet domain of C55. Tryptophan fluorescence spectra are shown of WT C55, along with C55 containing the A21G, T43I, and A21G/T43I mutations. The WT C55, T43I C55, and A21G/T43I C55 have shifted fluorescence spectra. The spectrum of A21G is more similar to that of the F20W mutant (12). These spectra were obtained with a 200:1 lipid/protein molar ratio. Experiments with 300:1 and 100:1 gave similar results. B, schematic showing the proposed structural changes that result from the A21G and T43I mutations. The WT C55 peptide formed a dimer mediated by its TM domain with the extracellular region adopting ␤-sheet structure that lies on the membrane surface. The T43I mutant disrupts the TM dimer. The TM helix tilts ϳ10°in the membrane, and the extracellular ␤-sheet becomes more membrane-embedded. Both the extracellular and TM regions have binding sites in the ␥-secretase complex (dashed line). A recent cryo-EM structure of the substrate in complex with ␥-secretase exhibits an unraveled C terminus and extended extracellular domain, in agreement with previous NMR and FTIR studies of membrane-embedded C55 (8,12).

Structure of the T43I mutant of C99
reduction in A␤40 secretion. Monomer formation is associated with a change in the tilt of the TM helix away from the bilayer normal. This overall change in orientation of the substrate may reduce its ability to bind to the enzyme.

The T43I C55 peptide is monomeric
There has been considerable discussion about whether the C99 forms a monomer or dimer and whether the oligomeric state influences ␥-secretase processing (8, 12, 25, 28, 31, 38 -40). Our structural data indicating that the T43I mutation disrupts the WT C55 dimer agree with fluorescence studies on TM peptides (22) and full-length C99 (11) indicating that the cluster 2 mutations all lead to a shift in the monomer-dimer equilibrium toward monomer.
This conclusion disagrees with solution NMR studies arguing that C99 is monomeric (28) or that TM peptides containing the V44M and V44A FAD mutations are dimeric (30). For C55, both the extracellular and TM domains influence the overall structure of the ␤-CTF (8,12). Deletion or mutation of the extracellular domain as in the WT TM peptides shifts the monomer-dimer equilibrium toward dimer (8,12). For detergentsolubilized C55, the level of detergent used can influence the monomer-dimer equilibrium, with less detergent driving peptide aggregation.
Importantly, the observation that monomer formation of the WT substrate decreases processing implies that the dimers are likely the native form of the substrate prior to complex formation. In fact, we previously found that the A21G mutation increases dimerization and increases the amount of secreted A␤40 (8,12).
Recent observations suggest that monomeric FAD mutant substrates having altered structures influence substrate binding and proteolysis even at the initial ⑀-cleavage step (37). The C99 monomer with an increase in helical secondary structure will likely behave more as a rigid rod, and this may be more difficult to accommodate in the initial enzyme-binding site or in the enzyme catalytic active site. The general change from dimer to monomer may then explain the decrease of total A␤40 production in T43I as well as the other cluster 2 mutants.

Summary
␥-Secretase catalysis of C99 involves at least four steps: 1) substrate binding, 2) translocation to the active site, 3) catalysis, and 4) product release. Changes in substrate structure, dynamics, and/or interactions at one or more of these steps may influence the amount and length of A␤ secreted by the enzyme.
First, the C99 substrate is in a monomer-dimer equilibrium in membrane bilayers, where the structure, dynamics, and membrane interactions of the substrate influence its initial binding to the enzyme complex away from the catalytic site (Fig. 8B). As mentioned above, T43I C99 has been shown to have a weaker interaction with ␥-secretase than WT C99 (37). The monomeric structure of the T43I mutant with a higher tilt angle of the TM helix and a more membrane-embedded extracellular domain may impede initial binding to the enzyme and be the origin of the lower levels of total A␤ peptide produced.
Second, the substrate is translocated to a position within the enzyme complex capable of catalysis (41). Recent cryo-EM structures of the ␥-secretase complex provide evidence for significant rearrangements of the TM helices during the catalytic process (42). The structure, flexibility, and interactions of the substrate likely influence substrate translocation and the final orientation of the TM helix prior to the initial ⑀-cleavage.
Finally, progressive cleavage of the substrate is influenced by the ability of the substrate to unravel and form a productive catalytic complex with the enzyme (37). The WT dimer having distortions both in the region of Gly 37 -Gly 38 and in the region below the ␥-cleavage site likely facilitates unraveling. A rigid ␣-helix in the T43I monomer (as shown here) appears to decrease stability of the complex (37). Recent cryo-EM structures of enzyme-substrate complexes with ␥-secretase reveal a helical substrate that requires chemical cross-linking to observe (43,44) possibly as a result of weak enzymesubstrate interactions.

Experimental procedures
Materials 13 C-Labeled amino acids were purchased from Cambridge Isotope Laboratories (Andover, MA). Other amino acids and octyl-␤-glucoside were obtained from Sigma. DMPC and DMPG were obtained from Avanti Polar Lipids (Alabaster, AL) as lyophilized powders and used without further purification.

Peptide synthesis and purification
C55 peptides corresponding to the TM and juxtamembrane regions of APP were synthesized by solid-phase methods (Keck Facility, Yale University). The purity was confirmed with MALDI MS and analytical reverse phase HPLC.

Protein overexpression and purification
The 55 residues containing the full-length A␤, transmembrane domain, and a few residues in the juxtamembrane sequence (KKK) of amyloid precursor protein were cloned into a pET21a vector with the added N-terminal methionine and an attached linker/His 6 tag (KLAAALEHHHHHH) at the end of the sequence. The plasmid harboring target DNA sequences then was transformed into BL21 Escherichia coli strain and plated on Luria-Bertani broth (LB)/ampicillin plates. A single colony on the plate was selected and inoculated into 5 ml of LB/ampicillin medium overnight at 37°C at 180 rpm (A 600 ϳ1.8). The cultured medium was spun down at 4°C, and cells were resuspended and cultured in 25 ml of sterilized LB medium or M9 minimal medium (40 mM Na 2 HPO 4 , 20 mM KH 2 PO 4 , 10 mM NaCl, 20 mM NH 4 Cl, 0.2% glucose, pH 7.0, supplemented with 0.1 mM CaCl 2 , 1 mM MgSO 4 , trace elements, and vitamin B) containing ampicillin (0.1 mg/ml), until the A 600 reached 0.6. The cultured medium was inoculated to the sterilized 1 liter of LB medium (or M9 minimal medium) at 37°C at 180 rpm, until the A reached ϳ0.8. The cells were induced with 1 mM isopropyl ␤-D-thiogalactopyranoside for 16 h at 23°C at 180 rpm. The induced cells (ϳ1.5 gram) are then spun down at 6000 ϫ g, resuspended in the 35 ml of ice-cold lysis buffer (75 mM Tris, 300 mM NaCl, 0.2 mM EDTA, and pH 7.8), and passed through a French pressure cell press (SLM AMINCO) twice at 1000 p.s.i. The harvested cells were spun

Structure of the T43I mutant of C99
down at 25,000 ϫ g, and the pellets containing C55 were subjected to subsequent purification.
The pellets were washed with the 35 ml of lysis buffer and spun down at 25,000 ϫ g for 20 min until the washed solution became clear (usually with three iterations). The spun-down pellets were dissolved in 35 ml of urea/SDS buffer (20 mM Tris, 150 mM NaCl, 8 M urea, 0.2% SDS, pH 7.8). The pellet was first homogenized with an 18 G 11/2 needle to increase the solventexposed surface and mixed well overnight at room temperature until the pellet was dissolved. The dissolved pellets were centrifuged at 25,000 g for another 20 min. The supernatants were pooled with the pre-equilibrium nickel beads for 2 h (2.5 ml of resin for 1 g of cell pellet). The beads were then washed with 4 bed volumes of urea/SDS buffer, 4 bed volumes of SDS rinse buffer (20 mM Tris, 150 mM NaCl, 0.2% SDS, pH 7.8), and 8 bed volumes of detergent-contained TBS buffer (20 mM Tris HCl, 200 mM NaCl, pH 7.8) in a pulsed manner. The proteins were refolded and were eluted with 250 mM imidazole in TBS solution (pH 7.8) containing detergent. The eluted fractions are collected, and checked by SDS-PAGE. The protein concentration was estimated with 280 nm absorbance using an extinction coefficient of 1490 M Ϫ1 cm Ϫ1 , after removal of imidazole with an Amicon ultra spin concentrator (3 kDa molecular mass cut-off). The protein purity was further confirmed with MS (MALDI TOF).

Reconstitution of C55 into membrane bilayers
The C55 peptides were co-solubilized in DMPC, DMPG, and octyl-␤-glucoside in hexafluoroisopropanol. The peptide/lipid molar ratio was 1:50; the molar ratio between DMPC and DMPG was 10:3. The solution was incubated for 90 min at 37°C, after which the solvent was removed under a stream of argon gas and then under vacuum overnight. MES buffer (5 mM MES, 50 mM NaCl, pH 6.2) was added to the solid from the previous step and gently mixed at 37°C for 6 h. The octyl-␤glucoside was removed by dialysis (45). The reconstituted membranes were subjected to FTIR analysis and then pelleted and loaded into NMR rotors.

Attenuated total reflection (ATR) FTIR spectroscopy
Polarized ATR FTIR spectra were obtained on a Bruker IFS 66V/S spectrometer. ATR FTIR spectroscopy was used to characterize the global secondary structure and the orientation of the TM domain of C55 in bilayers. The reconstituted peptides in bilayers were layered-down on a germanium plate. The amide I vibrational frequency (1600 -1700 cm Ϫ1 ) is sensitive to the secondary structure. An amide I frequency between 1650 and 1660 cm Ϫ1 is characteristic of the ␣-helix. The absorbance difference between 90 and 0°of polarized light provides information on the orientation of transmembrane helices relative to bilayers. The dichroic ratio (I 90 /I 0 ) on TM helices was used to calculate the tilt angle, with the value of ␣ ϭ 41.8°based on parallel measurements on bacteriorhodopsin (45).

Solution NMR spectroscopy
NMR experiments were performed on a 700-MHz Bruker AVANCE spectrometer. Several detergent systems (DPC, SDS, dihexoylphosphocholine, and octyl-␤-glucoside) were screened to optimize the resolution in multidimensional experiments. TROSY (transverse relaxation-optimized spectroscopy) type of HNCO, CBCA(CO)NH, and CBCANH spectra were collected and analyzed for the sequential assignment. The sample temperature was maintained at 45°C for the full 3D experiments and 40°C for the HSQC spectra in Fig. 3.

Solid-state NMR spectroscopy
NMR experiments were performed at a 13 C frequency of 600 MHz on a Bruker AVANCE spectrometer. The MAS spinning rate was set to 9 -11 KHz. The ramped amplitude cross-polarization contact time was 2 ms. Two-pulse phase-modulated decoupling was used during the evolution and acquisition periods with a field strength of 80 kHz. Internuclear 13 C . . . 13 C distance constraints were obtained from 2D DARR NMR experiments (46) using a mixing time of 600 ms. The sample temperature was maintained at 198 K. The solid-state MAS NMR spectra were externally referenced to the 13 C resonance of undiluted TMS at 0 ppm at room temperature. With TMS as the external reference, we calibrated the carbonyl resonance of solid glycine at 176.46 ppm. The chemical-shift difference between 13 C of DSS in D 2 O relative to neat TMS was 2.01 ppm.

Processing of C99 and Thr 43 mutants
A␤ production from CHO cells was monitored in the cell culture media 48 h after transfection. Briefly, samples were cleared by centrifugation (12,000 ϫ g, 3 min, 4°C). A␤38, A␤40, and A␤42 were quantified in 25 l of cellular medium by multiplex 4G8 or 6E10 A␤ electrochemiluminescent assays according to the manufacturer's instructions (MesoScale Discovery, Gaithersburg, MD).