It is important to stress that cysteine scanning mutagenesis is a biochemical technique which permits one to draw conclusions on the water accessibility of amino acid residues and their relative position versus each other in the tertiary conformation. It is, however, an extreme extrapolation to draw conclusions from such experiments on the global structure of a protein domain,...  Read more

It is important to stress that cysteine scanning mutagenesis is a biochemical technique which permits one to draw conclusions on the water accessibility of amino acid residues and their relative position versus each other in the tertiary conformation. It is, however, an extreme extrapolation to draw conclusions from such experiments on the global structure of a protein domain, like the funnel-like structure for the catalytic site of presenilin, as Sato et al. claim. Especially for presenilin this might be a dangerous exercise, since we have no hard structural information or even structures that can be used as models to interpret the data, and therefore caution with this type of conclusion is indicated. Sato et al. also did not take into account that presenilin might be present in the cell in distinct functional and therefore most probably structural conformations, depending on whether it is associated or not with the other γ-secretase complex components. The interpretation of the cysteine scanning data should take this into account as we actually did in our paper.

We feel somewhat surprised by Sato and colleagues’ aggressive criticism of our work. In contrast to Sato et al., we conclude that many of the data in our papers significantly overlap, suggesting that cysteine scanning mutagenesis is indeed a reliable methodology to probe structure function (within the indicated limits). For instance, Sato et al. confirm a differential labeling pattern for TM6 and TM7 which is in accordance with our conclusions and suggesting that TM6 and TM7 contribute in a different way to the activation and/or catalytic mechanism. In addition, Sato et al. show a similar labeling pattern in TM7 as we reported in our paper. Interestingly, this result suggests that the cytosol-oriented part of the TM7 exhibits a solvent accessible (flexible) non-regular secondary structure. A similar unstructured region was observed in the catalytic site-contributing TM4 in the crystal structure of rhomboid GlpG, another intramembrane cleaving protease from an entirely different functional class from presenilin. [3]

Sato et al. stress quite extensively the differences between their and our results. While they allude to methodological differences (for instance, the chemistry of the compounds used) and the absence of controls (which is, in fact, not the case), we believe that a more fundamental and interesting fact could explain at least part of the observed differences. Indeed, the basic construct (cysteine-less presenilin) used in the two studies is quite different, because our cysteine-less mutant is functionally similar to the wild-type presenilin, while Sato et al. used a cysteine-less presenilin that behaves as a clinical mutation, in fact, a loss-of-function form of wild-type presenilin. Indeed, Sato et al. (2006) introduced a Cys92Ser substitution, a known AD mutation, while we used a Cys92Ala form in order to create our cysteine-less template [2]. We provide evidence that the latter mutant behaves like wt PS1 in the in vivo processing of at least three tested substrates. In contrast, the Sato et al. cysteine-less version of PS1 behaves clearly as an FAD mutation in the processing of APP (~threefold decrease in Aβ40 and ~twofold increase in Aβ42, resulting in a more than a fivefold increase of the ratio Aβ42/40). Thus, the few differences we see in both studies could reflect differences in the conformation of mutant versus wild-type presenilin.

Sato et al. also heavily criticized us about including inactive mutants in our accessibility study, such as G384C, D257C, and D385C, whereas they excluded them from their study arguing that they are liable to significant structural changes, again suggesting that we did not perform adequate controls. We would like to point out here briefly 1) that we demonstrated that none of our cysteine substitutions affected the incorporation of PS1 into the γ-secretase complex as assessed by blue native gel electrophoresis and 2) that this criticism implies that Sato et al. also do not accept any of the previous publications that investigated, for instance, the substitution of the catalytic aspartates. In contrast, we believe that one of the major findings in our manuscript was the close apposition of the two cysteines replacing the putative catalytic aspartates in TM6 and TM7, confirming for the first time that these residues, which are quite remote in the primary structure, are in close proximity to each other in the native conformation.

On the other hand, we feel that Sato et al. could have been a bit more critical when interpreting their own data with regard to the γ-secretase inhibitor binding experiments. Briefly, changes in the accessibility of certain amino acid residues in the complex upon inhibitor treatment does not prove that these residues are directly involved in the binding of the inhibitor. The alternative interpretation is, indeed, that the inhibitor induces a conformational change in the catalytic site, like transition state analogues do, and that this actually changes the accessibility of the residue. This possibility must also be taken into account when “modulators” of enzyme activity are studied, such as the non-transition state inhibitor DAPT.

In conclusion, we welcome the work of Sato et al. as an interesting study on the structure function of the catalytic site of presenilin, and we are very pleased with the partial confirmation and the extensions of our previous work. We would like to stress, however, that it is important to take the limitations and considerations discussed above into account when interpreting the data in these two first studies. We also anticipate that further studies around the lines of ours [2] and others [1] will shed further light into the complex structure of γ-secretase. The real value of this work, as with other site-directed mutagenesis work, will, however, be fully realized when it is combined with other approaches.

References:
1. Sato, C., et al., Structure of the Catalytic Pore of {gamma}-Secretase Probed by the Accessibility of Substituted Cysteines. J. Neurosci., 2006. 26(46): p. 12081-12088. Abstract

2. Tolia, A., L. Chavez-Gutierrez, and B. De Strooper, Contribution of presenilin transmembrane domains 6 and 7 to a water-containing cavity in the gamma-secretase complex. J Biol Chem, 2006. 281(37): p. 27633-42. Abstract

3. Wang, Y., Y. Zhang, and Y. Ha, Crystal structure of a rhomboid family intramembrane protease. Nature, 2006. 444(7116): p. 179-80. Abstract

View all comments by Lucia Chavez-Gutierrez
View all comments by Bart De Strooper
View all comments by Alexandra Tolia

SCAM analysis and the pore within γ-secretase
Over 20 percent of the genes sequenced so far are known or predicted to encode polytopic transmembrane proteins. Structural analysis of these hydrophobic, sticky proteins remains difficult in general. Nonetheless, several molecular engineering techniques have provided detailed information about the structure of polytopic membrane proteins in relation to their functions. We [1] and Bart De Strooper’s group [2] utilized the substituted cysteine accessibility method (SCAM) to analyze the structure of presenilin 1 (PS1), an unusual membrane-embedded catalytic subunit for γ-secretase, which is linked to the pathogenesis of Alzheimer disease.

First, to solve the detailed atomic structure of membrane proteins, we should await the realization of X-ray crystallographic analysis. So, the “funnel-like structure of the catalytic pore” that we suggested in our paper remains a hypothetical model. However, recent X-ray structural analyses of Rhomboids [3,4] provided us with a structural clue to the mechanism of intramembrane...  Read more

SCAM analysis and the pore within γ-secretase
Over 20 percent of the genes sequenced so far are known or predicted to encode polytopic transmembrane proteins. Structural analysis of these hydrophobic, sticky proteins remains difficult in general. Nonetheless, several molecular engineering techniques have provided detailed information about the structure of polytopic membrane proteins in relation to their functions. We [1] and Bart De Strooper’s group [2] utilized the substituted cysteine accessibility method (SCAM) to analyze the structure of presenilin 1 (PS1), an unusual membrane-embedded catalytic subunit for γ-secretase, which is linked to the pathogenesis of Alzheimer disease.

First, to solve the detailed atomic structure of membrane proteins, we should await the realization of X-ray crystallographic analysis. So, the “funnel-like structure of the catalytic pore” that we suggested in our paper remains a hypothetical model. However, recent X-ray structural analyses of Rhomboids [3,4] provided us with a structural clue to the mechanism of intramembrane proteolysis: the existence of a hydrophilic environment around the catalytic motif that is encoded within hydrophobic amino acid sequences, which fits well with our model. This finding suggests that this type of structure is common to a set of intramembrane proteases.

Second, as pointed out in De Strooper’s comment, our cysteine-less PS1 indeed showed an increase in the generation in Aβ42, although this mutant cleaved Notch as well (data not shown in the paper). However, we would like to stress that our study is not based on a structure of PS1 that preferentially generates Aβ42, because some mutants did show normal Aβ generation similarly to wild-type PS1, after introduction of a single Cys residue to the Cys-less PS1, whereas others showed variable levels of Aβ42 overproduction (see supplementary figure in [1]). This type of problem has been quite common in previous SCAM analyses on transporters or channels [5]. Thus, we envisage our structural model as representing a “common denominator” structure of PS1 in an active state. To specify the structural changes in PS1 related to Aβ42 production, we are currently trying to analyze the effect of γ-secretase modulators (e.g., NSAIDs or Aβ42-augmenting compounds) on the Cys labeling.

Third, we do not know the exact mechanism as to how some Cys mutations of PS1 caused the loss of γ-secretase activity. However, we know that the acquisition of γ-secretase activity requires several additional processes that are likely to take place after the formation of a ~450 kDa high-molecular-weight complex: for example, we have recently shown that some PS1 TMD swap mutants lose γ-secretase activity, although forming a high-molecular-weight complex on BN-PAGE complexed with other γ-components (Watanabe et al., SfN meeting, 2006). We surmise that the proper conformation that provides the complex with stability, oligomerization of components, and trafficking may also be required for the proteolytic activity. Even the D385A mutant was reported to exhibit a slight difference in the migration pattern in BN-PAGE, which is likely to reflect the molecular or structural differences from wild-type PS1 [6]. We cannot even exclude the possibility that γ-secretase activity per se is required for the acquisition of the complete active structure, in a similar manner to other classical proteases with prodomains. These are the reasons why we carefully avoided the detailed SCAM analysis of the loss-of-function mutants, exclusively focusing on the structural analysis of “proteolytically active γ-secretase,” like previous studies [5].

We appreciate the caveat that the γ-secretase inhibitors may bind to regions distant from the amino acid residues that were masked upon treatment. In fact, we have reported that a dipeptidic-type inhibitor DAPT bound to TMDs within PS1 CTF, but no data suggesting the binding to NTF have been obtained [7]. To unequivocally address these issues, further rigorous identification of the inhibitor binding sites using fine chemical biological approaches is mandatory.

Finally, we are very pleased that we and De Strooper’s group independently reached the same and the most important conclusion in this field: γ-secretase cleaves its substrate within a hydrophilic environment in the membrane. We also appreciate the comments and discussions from De Strooper and colleagues on our analysis. Some researcher once called the SCAM analysis on lac permease of E. coli as a “heroic KAMIKAZE approach” [8,9], because each new postdoc was forced to make ten single-Cys mutants until the whole protein was covered. Now, combining the results from X-ray crystallographic analysis, these researchers can re-evaluate the results of SCAM and apply the well-characterized mutants to further dynamic analyses (e.g., structural change upon ligand binding) [10]. We are convinced that γ-secretase is one of the fascinating proteases that deserve this heroic, and fruitful, approach. Nonetheless, we have to keep on making every effort to unveil the structure of γ-secretase using different approaches, in order to completely understand the shape of the catalytic pore as well as the whole complex.

References:
1. Sato C, Morohashi Y, Tomita T, Iwatsubo T. Structure of the Catalytic Pore of {gamma}-Secretase Probed by the Accessibility of Substituted Cysteines. J Neurosci. 2006 Nov 15;26(46):12081-12088. Abstract

2. Tolia A, Chavez-Gutierrez L, De Strooper B. Contribution of presenilin transmembrane domains 6 and 7 to a water-containing cavity in the gamma-secretase complex. J Biol Chem. 2006 Sep 15;281(37):27633-42. Epub 2006 Jul 14. Abstract

3. Wang Y, Zhang Y, Ha Y. Crystal structure of a rhomboid family intramembrane protease. Nature. 2006 Nov 9;444(7116):179-80. Epub 2006 Oct 11. Abstract

4. Wu Z, Yan N, Feng L, Oberstein A, Yan H, Baker RP, Gu L, Jeffrey PD, Urban S, Shi Y. Structural analysis of a rhomboid family intramembrane protease reveals a gating mechanism for substrate entry. Nat Struct Mol Biol. 2006 Nov 10; [Epub ahead of print] Abstract

5. Seal RP, Leighton BH, Amara SG. A model for the topology of excitatory amino acid transporters determined by the extracellular accessibility of substituted cysteines. Neuron. 2000 Mar;25(3):695-706. Abstract

6. Gu Y, Sanjo N, Chen F, Hasegawa H, Petit A, Ruan X, Li W, Shier C, Kawarai T, Schmitt-Ulms G, Westaway D, St George-Hyslop P, Fraser PE. The presenilin proteins are components of multiple membrane-bound complexes that have different biological activities. J Biol Chem. 2004 Jul 23;279(30):31329-36. Epub 2004 Apr 30. Abstract

7. Morohashi Y, Kan T, Tominari Y, Fuwa H, Okamura Y, Watanabe N, Sato C, Natsugari H, Fukuyama T, Iwatsubo T, Tomita T. C-terminal fragment of presenilin is the molecular target of a dipeptidic gamma-secretase-specific inhibitor DAPT (N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester). J Biol Chem. 2006 May 26;281(21):14670-6. Epub 2006 Mar 28. Abstract

8. von Heijne G. A Day in the Life of Dr K. or How I Learned to Stop Worrying and Love Lysozyme: a tragedy in six acts. J Mol Biol. 1999 Oct 22;293(2):367-79. Abstract

9. Kaback HR, Sahin-Toth M, Weinglass AB. The kamikaze approach to membrane transport. Nat Rev Mol Cell Biol. 2001 Aug;2(8):610-20. Review. Abstract

10. Abramson J, Smirnova I, Kasho V, Verner G, Kaback HR, Iwata S. Structure and mechanism of the lactose permease of Escherichia coli. Science. 2003 Aug 1;301(5633):610-5. Abstract

View all comments by Takeshi Iwatsubo
View all comments by Chihiro Sato
View all comments by Taisuke Tomita