Marc Paradis led this live discussion on 20 March 1998. Readers are invited to submit additional comments by using our Comments form at the bottom of the page.
Paper under Discussion: Li JH, Xu M, Zhou H. Ma J. and Potter H. Alzheimer presenilins in the nuclear membrane, interphase kinetochores, and centrosomes suggest a role in chromosome segregation. Cell, Vol 90, September 5, 1997. Abstract
Live discussion held 20 March 1998.
Participants: Jinhe Li, Huntington Potter, Marc Paradis, June Kinoshita, Oksana Berezovska, Chris Weihl, MQ Xia.
Note: Transcript has been edited for clarity and accuracy.
Marc Paradis: Hello Jli and June.
June Kinoshita: Hi Marc!
Jli: Hi Marc!
June Kinoshita: Jinhe, where are you these days?
Jli: Hi June! I'm with Pharmacia & Upjohn since Jan.
Marc Paradis: Hello Hunt.
June Kinoshita: Hi Hunt!
Hpotter: Hi folks. Jinhe and I are on the phone together, so I may do the answering for both of us.
June Kinoshita: That's real multimedia! Hello Chris and Oksana.
Cweihl: Hi June. Am I late?
June Kinoshita: No. We haven't begun yet. Glad to see you Oksana. Why don't we begin. Anyone want to lob the first question?
Jli: I'm ready.
Marc Paradis: I'm ready as well
Cweihl: Me too.
Jli: Ready to take the bullets.
Marc Paradis: I would like to start by thanking June for giving me the opportunity to review and also to thank Hunt and Jinhe for writing an excellent paper that was fun to read and think about.
Hpotter: You did a great job.
June Kinoshita: I would like to thank Marc for the wonderful job he did.
Marc Paradis: I would just start with a question from my review if that is alright?
Marc Paradis: One of the most interesting questions, I thought, concerned your mechanisms for non-disjunction in post-mitotic neurons and how this might cause AD. I would like to hear more of your thoughts on this.
Hpotter: We don't actually know how many neurons are totally post-mitotic. As we mentioned in our formal response, there is a fairly large number of new neurons generated in the dentate. However, if we accept the premise that neurons are normally post-mitotic, we have to explain why there are so many mitotic markers in AD neurons.
Jli: So, 1-neurons have the potential to divide, 2-there are non-neuronal cells undergoing cell division, 3-the effect may occur during early development in all cells including those in brain.
Cweihl: Dr. Potter, I was interested if there is any precedent for a multi-transmembrane protein associating with chromosomes? And serve a role in mitosis?
Marc Paradis: Perhaps a very simple question, but I have never heard of this neuronal replacement before. Are there supposedly stem cells that generate the new neurons or do neurons actually make a decision to perform one division at some point in life?
Hpotter: I don't think that is known. Apparently the new neurons just appear in the midst of the old ones. It is not as though there were a region where all the stems cells are and then they migrate as in other regions.
June Kinoshita: I believe there are significant numbers of stem cells in adult brain.
Marc Paradis: If a structure like the dentate can turn over every 6 years, but is only allowed one turnover (because each cell only divides once) what happens after one complete turnover?
Hpotter: I did not mean to imply that each cell would turn over only once. It could turn over many times during the course of a life. But for the sake of the calculation I assumed one turnover per cell. If some cells turned over more than once in a given time and others not at all, then it might take more than 6 years for the entire dentate to turn over.
June Kinoshita: Have you been able to look at the number of trisomic neurons in the dentate across different age groups?
Hpotter: I have a question for discussion. What do you folks think of the idea that all of the effects of the APP and presenilin mutations on Aβ production are mediated through their effects on apoptosis?
Marc Paradis: I would say that it can't simply be apoptosis, because many stimuli and insults result in apoptosis, but only APP and PS mutations seem to cause AD (ignoring sporadic issues for the moment).
Hpotter: What I mean is that since apoptosis induced by normal means causes increases in Aβ, then maybe it is the apoptosis that the APP and PS mutants induce that leads to the increase in Aβ1-42.
June Kinoshita tells Oksana I'm curious what you might be observing with respect to presenilin localization....
Hpotter: Marc has a good point.
Jli: To Cweihl's Q: after looking at so many stainings for a long time, I think PSs are associated with some kind of cytoskeleton protein. And centrosome is the center of the organization of cytoskeleton protein. So they seem to have a chance to get together. But I don't know how, especially for a 7-tm protein.
Oksana: In mouse and human tissue we occasionally see the PS1 in nuclei, but it is also localized in the cytoplasm in cell body and processes.
Marc Paradis: The real test would be other neurologic diseases which cause apoptosis but do not cause AD or AD like symptoms.
Hpotter: June: we have tried to use fluorescence in situ hybridization to look at brain sections for trisomy 21. The problem is that the lipofuscin is so strongly fluorescent in aged brain that it obscures the hybridization spots. We will have to try other techniques.
Marc Paradis: I agree that Cweihl's question is a good one although not impossible given ER and Golgi localization.
Marc Paradis: Sorry, maybe also nuclear membrane localization too!
Cweihl: In your experiments involving interactions with other proteins. (2 hybrid) what regions of PS are you interested in?
Hpotter: To cweihl: Yes we see cytoplasmic staining in sections of brain tissue too. We discuss this a bit in the formal response. Probably different cells use the presenilins for different things. That might make a neuron that hopes not to divide use the PS proteins for some other membrane-cytoskeleton interactive function such as vesicle transport. Good point about other apoptosis diseases. One possibility is that they do cause some Aβ deposition, as does head trauma, but that if the apoptosis is transient, so will be the Aβ deposits.
Cweihl: To me the most compelling evidence of PS and apoptosis is the antiapoptotic nature of ALG-3.
June Kinoshita: It's been speculated that apoptosis is involved in all kinds of neurological diseases: ALS, Huntingtons, etc. How convincing is the evidence? I guess I should ask that about Alzheimer's too.
Hpotter: Cweihl: Absolutely. Also PS mutants cause apoptosis as do APP mutants.
Jli: Maybe apoptosis is not a process specific for any disease, but rather one of the final steps in many diseases.
Marc Paradis: I would tend to agree with Jli, simply by Occam's razor type arguments.
Cweihl: As far as PS. It appears to play a real role in apoptosis with caspase cleavage etc. Whereas APP mutants may just add to the overall sickness of the neurons.
Hpotter: June: Apoptosis is hard to be sure about in post-mortem brain because of all the insults that have occurred before you get the tissue. Also some of the techniques for measuring apoptosis are so sensitive that you can easily get false positive results.
Oksana: Hi, this is MQ Xia from Brad Hyman's lab, I just jump in the middle of this discussion. We have observed both nuclear staining and cytoplasmic staining of PS-1 in mouse and in human brains using polyclonal antibodies against the N-terminal of PS-1. In human brain, a lot of nuclear staining was observed, particularly granule cells of dentate gyrus. The nuclear staining appears to be related to postmortem interval. This paper is coming out very soon in Journal of Neurological Sciences.
June Kinoshita: Very interesting, MQ!
Marc Paradis: So perhaps PS1 is moved to the nucleus with apoptosis/cell death/post-mortem interval.
Cweihl: I am confused. What is post mortem interval?
Hpotter: Has anyone else used commercial antibodies (as we have, with an antibody from SANTA CRUZ ) We found the same nuclear membrane staining in the fibroblasts.
Marc Paradis: Such movement would be analogous to localization of steroid receptors to the nucleus on binding to steroids. This was actually a possibility that came to mind while reading Jli and Hpotter's paper, but that I didn't have the space to get into in my review. BTW I would also be interested to hear about commercial antibodies.
Hpotter: PMI is the time from death to the time the autopsy is complete and you can get the brain. You can imagine what your tissue culture cells would look like if you left them with no oxygen or glucose for a few hours--lots of death.
Oksana: Postmortem interval is the time after death and before the brain was fixed. The same phenomenon was also observed in mouse brains of different postmortem interval. In human, after 12 hrs, PS-1 nuclear staining gets really weak -- sometimes cannot be seen at all.
Marc Paradis: Throwing out ideas here, perhaps the nuclear localization is related to the caspase cleavage and thereby to cell death/apoptosis?!
Cweihl: I have used Chemicon's rat monoclonal on western blots and Immunohistochem, and it works well.
June Kinoshita: Oh, so PS1 staining gets WEAKER with increased PS1.
Marc Paradis: Sorry, looks like I may have the sequence backwards here, agree with June?
Jli: I'm really happy to hear that from MQ Xia.
Oksana: Sorry , that I forgot to mention that our PS-1 antibody is a rabbit polyclonal to N-terminal 81 amino acid, similar to Dr. Potter's.
Hpotter: Actually the antibody was.
Oksana: MQ Xia: usually in human brain, after 12 hr PMI, nuclear staining gets weaker or invisible. This was most obviously observed in granule cells of dentate gyrus.
Marc Paradis: So I did have it backwards.
June Kinoshita: Are you suggesting others may have missed seeing the nuclear staining because the PMI was too long?
Oksana: MQXia. Very likely, and the antibody we used was by far the strongest. We seem to prefer polyclonal antibody against longer peptide.
Cweihl: Nuclear staining is far different from kinetechore localization during mitosis.
Hpotter: That is fascinating. We were happy to see the association with notch and PS expression during development since that would fit with a joint function in cell cycle regulation. Indeed several components of the notch signaling pathway have been shown be involved in cell cycle and in chromosome segregation.
Marc Paradis: The rapid disappearance post-mortem could be explained by rapid cleavage and processing of the PS, which is consistent with a lot of previous work.
June Kinoshita: For Hunt and Jinhe: Did Bruce McEwen's paper indicate whether granule cells were undergoing division more than other types?
Hpotter: J: He only could see it there for technical reasons.
June Kinoshita: So is it still an open question whether there is neurogenesis going on in adult animals in other regions of the brain?
Oksana: About PS1/notch colocalization. We have data that they do colocalize both temporally and, anatomically (in the same brain areas) and in the same neurons. It was double fluorescence.
Hpotter: The cleavage of PS proteins does seem to be related to some mistreatment of the cells. When we harvest, we put the cells directly into lysis/sample buffer with SDS and see full length protein. Others may trypsinize or harvest by centrifugation that would stress the cells. Maybe that is why so many see only the cleaved forms.
Jli: to Xia: Very interesting. Is it coming out soon?
Hpotter: Also Dewji and Singer said that full cleaved PS was only present when the cells were harvested carelessly: PNAS a few months ago.
Cweihl: Hunt, is the full length that you see from cell overexpressed or endogenous PS
Oksana: Xia: yes, it is in press, if not this month then next month will be likely.
Hpotter: What journal?
Oksana: xia: J. Neurological Sciences.
Cweihl: So Hunt do you believe in the cleavage of PS1?
Hpotter: Actually. we think it possible that the cleavage is a part of the normal function of the presenilins. If the proteins function as kinetochore receptors, they need to let go every time the cell enters enaphase. What better way than to be cleaved and stop functioning.
Jli: PS1/2 may be cleaved, but not necessarily completely for their function.
Marc Paradis: That's great stuff Hunt! What kind of cells are you looking at?
Hpotter: Full length in fibroblasts and transfected cos cells, and also fragments.
Cweihl: Do you see the cleavage products at all or just full length?
Jli: Using both anti-flag and anti-ps abs.
June Kinoshita: If you speculate that AD pathology is seen early in hippocampus because of cell division there, how do you account for the spread to other brain. regions? Do you think there is cell division also going on at a slower rate, or is there some other mechanism by which dysfunction and death are spreading?
Hpotter: Actually, we are much more inclined to though think that the regional specificity reflects an astrocyte-microglial cell difference coupled to a difference in neuronal cell vulnerability. Both astrocytes and microglia divide throughout adulthood, and we know they are weird in Down syndrome. They could be the most affected by PS mutations causing problems with the cell cycle.
Jli: To June: different brain regions or cell types may have different extent of vulnerability to the aβ toxicity. That’s why only some of the regions are affected in AD.
June Kinoshita: In what ways are astrocytes and microglia weird in Down's?
Hpotter: One thing is clear. The cerebellum has amorphous amyloid but not mature amyloid. It also has no expression of the amyloid promoting factor ACT and almost no expression of the ACT and inflammation-inducing cytokine IL-1.
Marc Paradis: Hello again, Mqxia.
June Kinoshita: Welcome back MQ. You're not in drag anymore!
Hpotter: Astrocytes and microglia are spontaneously activated even in fetal brain There are 30 times as many IL-1-positive microglia in DS brain of all ages (Griffin et al , 1989).
Mqxia: Hi, nice to be back here.
June Kinoshita: Didn't know that. So inflammation may be important in Down's too?
Marc Paradis: Inflammation, I think, plays a role analogous to apoptosis, a final common pathway, not necessarily specific to AD. Of course, this doesn't negate the importance of studying or designing treatments based on inflammation or apoptosis. It just means that you probably won't cure AD that way, just slow it down.
Hpotter: Yes, inflammation (which we define in the brain as activated microglia and astrocytes and induced expression of inflammation proteins such as ACT and IL-1) is apparently constitutively activated in DS brain. I would expect that anti-inflammatory drugs would help them [Down syndrome individuals] develop better.
June Kinoshita: Interesting. I wonder if anyone is studying that possibility.
Jli: Very good point.
Marc Paradis: Hunt, you should start epidemiological studies of DS mothers with arthritis or some other condition for which they chronically take anti-inflammatories. Their DS children would be expected to have better outcomes... By DS mothers, I mean mothers of children with DS.
Hpotter: The only thing is that the inflammation is clearly responsible for expressing ACT and apoE, which appear to be necessary to get Aβ to polymerize efficiently. We have shown this in vitro, but Steve Paul's group at Lilly has shown it for ApoE in vivo.
Hpotter: I agree. That would be a good experiment. No one is trying I don't think.
Marc Paradis: Well, I'd be happy to help get that experiment going in any way that I can.
Hpotter: Lets talk. Maybe a drug company would help.
Jli: Let's try it.
June Kinoshita: It's a few minutes past 3 PM, so I would like to release Hunt and Jinhe of their obligation to stay on line. Of course, everyone is free to continue chatting here. I want to thank Hunt and Jinhe for being here, and Marc for the wonderful discussion text. Our next live journal club is next Thursday, 4-5 PM EST, with Rachael Neve, Nick Robakis and Dennis Selkoe. Should be another lively event!
Jli: Thanks to June and Marc.
Marc Paradis: I would like to say my thanks as well to Hunt, Jinhe and June. As well as to everyone who participated today. I learned quite a bit.
June Kinoshita: So did I.
Hpotter: Great. See you all next week at the next roast.
June Kinoshita: I'll say farewell for now. Off to lunch (I'm in California). Thanks again!
By Marc Paradis
The early-onset and aggressive form of Alzheimer's Disease (AD) evident in individuals trisomic for Chromosome 21 (Down's Syndrome) was one of the earliest clues to the genetic etiology of AD, and today it remains one of the cornerstones of the amyloid hypothesis of AD causation. Amyloid plaques, one of the two diagnostic lesions of AD, are composed of a peptide (Aβ) which is derived from the protein product of the Amyloid Precursor Protein (APP) gene located on Chromosome 21. This fact, in conjunction with the AD symptomatology and the overexpression of all genes on Chromosome 21 that is present in Down Syndrome (DS) patients, has often been taken as strong evidence for the central role of too much APP (and hence of too much Aβ) in the etiology of AD. Thus, in this more traditional view, AD is a separate disease entity of which the early onset form of AD seen in DS is a subset, albeit a subset due to the particulars of DS. That is to say, if APP were located on any other chromosome, then AD, in this view, would not be part of the constellation of symptoms that make Trisomy 21 a syndrome.
In 1991, Hunt Potter proposed an alternative hypothesis which turned this dogma on its head. He postulated that AD was actually a subset of Trisomy 21, and more generally, that AD was a subset of the disorder of chromosomal non-disjunction. Non-disjunction is the failure of two members of a chromosome pair to disjoin (separate) during meiosis I, or of two chromatids to disjoin either during meiosis II or mitosis, such that both members of the pair pass to one daughter cell while the other daughter cell receives neither member. A quick review of mitosis and meiosis will help to clarify the act and consequences of non-disjunction.
After a cell has duplicated its DNA (in S phase) and discerned no copy errors (in G2 phase), the cell commits (by entering M phase) to the processes of division that will either yield two identical diploid daughter cells (mitosis) or four allelicly shuffled haploid daughter cells (meiosis). In mitosis, M phase begins with prophase where the diffuse chromatin condenses into compact chromosomes (which are actually pairs of identical chromatids attached by a centromere), the nuclear membrane breaks down and the mitotic spindles, anchored to the centrosomes at either pole of the dividing cell, begin to form. Prophase is followed by prometaphase where the microtubules of the mitotic spindle capture individual chromosomes by binding to the specialized protein complexes at their centromeres called kinetochores. During metaphase, the chromosomes align along the metaphase plate and then suddenly separate at their centromeres to initiate anaphase. Anaphase is completed when the chromosomes (now single chromatids) reach their respective spindle poles. In telophase, nuclear membranes begin to reform around each spindle pole, the mitotic spindles disassemble and the mother cell pinches into two daughter cells. These daughter cells then enter G1 and the cell cycle is complete (as S phase follows G1). If a pair of chromatids fails to separate at anaphase, then non-disjunction occurs and the resultant daughter cells are no longer identical: one daughter is trisomic for the chromosome that failed to separate, while the other daughter is monosomic. The condition of having a number of chromosomes that is not a whole multiple of the haploid chromosomal number (as in monsomy or trisomy of a specific chromosome) is known more generally as aneuploidy.
Meiosis is identical to mitosis with the following two major differences. First there are two division events, termed meiosis I and meiosis II, and each meiotic event proceeds through prophase, prometaphase, metaphase, anaphase and telophase, just like mitosis. Second, unlike mitosis, pairs of duplicated homologous chromosomes align during prometaphase I and metaphase I such that at the onset of anaphase I homologous chromosomes (rather than chromatids) are separated. The resulting daughter cells each have one full set of duplicated chromosomes and are therefore diploid, but each chromosome derives from only one of the two homologous chromosomes in the original cell and each chromosome is actually composed of two sister chromatids. Meiosis II proceeds just as mitosis, although the final products are now four haploid gametocytes. A non-disjunction at anaphase I would result in two gametocytes with two copies of the non-disjoined chromosome and two gametocytes with no copies of the non-disjoined chromosome, while non-disjunction at anaphase II would yield one gametocyte with two copies of the non-disjoined chromosome, one gametocyte with no copies of the non-disjoined chromosome and two normal gametocytes. When a gametocyte with two copies of a non-disjoined chromosome fuses with a normal gametocyte, the resultant zygote is trisomic for the non-disjoined chromosome. When a gametocyte with no copies of a non-disjoined chromosome fuses with a normal gametocyte, the resultant zygote is monosomic.
Some final facts about non-disjunction. Non-disjunction during meiosis I is more common than non-disjunction during either meiosis II or mitosis and the rates of non-disjunction increase with both cellular and organismal age (although non-disjunction can, of course, only occur in actively dividing cells). Furthermore, as reported by Potter and his colleagues, Trisomy 21 is more than twice as common in cultured fibroblasts from AD patients than in fibroblasts cultured from controls.
Two years ago in rapid fashion, two more proteins, now named Presenilin I (PS1) and Presenilin II (PS2), were discovered to also be involved in the genetic etiology of early-onset AD. These two proteins have been intensely studied and over 230 papers regarding their genetic and molecular structure, their localization, their function, their processing, their interactions with other proteins and their role in AD have been generated in this short period of time. The current consensus among most researchers is that the Presenilins (PSs) have 8 transmembrane spanning regions (TSRs), with a large cytoplasmic loop between TSR 6 and TSR 7. Furthermore, it is believed that the PSs localize to the endoplasmic reticulum (ER) where, it is hypothesized, they interact either directly or indirectly with the enzymes involved in the processing of APP. Once again Potter and his colleagues seek to turn these hypotheses on their head and they make a start with the data presented in this paper.
With this paper, Li et al. have further expanded the body of evidence that supports Hunt Potter's non-disjunction hypothesis. To begin, the authors generated four novel polyclonal PS antibodies, two to the least homologous regions (the intracytoplasmic N-terminal and loop portions) of both PS1 and PS2. These antibodies were tested for sensitivity and specificity by comparison to overexpressed FLAG-tagged PSs. Furthermore, for their PS1 antibodies, all PS1 immunoreactivity was abolished in cells cultured from the PS1 null mouse (generated by Shen and Tonegawa), while the immunoreactivity of their PS2 antibodies was unchanged in these cultures.
Prior work on the localization of PSs had indicated that they were highly abundant in the ER, however Li et al. rightly criticize the validity of this conclusion based as it is upon cell lines vastly overexpressing either PS1 or PS2. They argue that the overexpression of any transmembrane or secreted protein will lead to ER enrichment and that the appropriate experiment to determine localization is one that assays endogenous PS. When they used their PS1 and PS2 antibodies on untransfected fibroblasts and lymphoblastoid cells, they observed staining primarily of the nucleus and of the nuclear membrane, accompanied by suggestive staining of structures reminiscent of the centrosome and interphase kinetochores. Double label experiments appeared to demonstrate co-localization of PS1 and PS2 with markers of the centrosome, the nuclear membrane, and the kinetochores. To further confirm and localize these findings, a series of ultrastructural analyses were undertaken. Li et al. coupled their PS2 antibodies to immunogold particles and found that 3.0 (N-terminal antibody) and 2.4 (loop antibody) times as many gold particles resided on the nucleoplasmic side of the inner nuclear membrane as compared to the cytoplasmic side of the outer nuclear membrane, with 10% of the gold particles located in the intermembrane space. Gold particles were also found associated with centrosomal structures.
Discussion of Figures
FIGURE 1. Immunolocalization of FLAG-Tagged Presenilin Proteins Expressed in Transfected Cells: COS cells were transiently transfected with 5'-tagged PS1 and PS2 and expression was driven by a weak promoter. Panel A demonstrates strong perinuclear and punctate nucleoplasmic staining for PS1, although strong ER staining is also clearly visible. Panel B demonstrates strong perinuclear staining accompanied by diffuse and punctate nucleoplasmic staining for PS2, again, strong ER staining is also clearly visible.
FIGURE 2. Western Blot Analysis of PS1 and -2 in Transfected COS Cells and Primary Fibroblasts: Panel A represents a cartoon of PS1 and PS2, emphasizing the non-homologous N-terminal and loop regions of these two proteins. Panel B compares transfected COS cell extracts of FLAG-tagged PS2, and PS2 fragments, as detected by either the anti-FLAG antibody or the PS2-N and PS2-L antibodies generated by the authors. Non-specific immunoreactivity can be observed in both the PS2-N and PS2-L blots, especially as compared to the anti-FLAG blot. Panel C demonstrates that the PS2-N and PS2-L antibodies can detect full-length PS2 in primary human fibroblasts and that pre-absorption of the antibodies with their respective immunogen abolished staining. Using PS1-L and PS1-N, Panels D and E paralleled Panels B and C. Again, non-specific immunoreactivity of PS1-L and PS1-N was visible in transfected COS cell extracts.
FIGURE 3. Immunohistochemical Localization of Endogenous PS2: Primary cultures of human fibroblasts were labeled using the authors' PS2 antibodies. Panel A shows a low power view of theses labeled primary cultures, dark nuclear staining is evident as is diffuse cytoplasmic staining. Panel B demonstrates that the secondary antibody alone exhibits no staining of these cultures. Panel C shows a higher power view of a single fibroblast, diffuse and punctate nuclear staining can be observed as well as strong centrosomal staining, however, diffuse and punctate staining of the cytosol and ER is also clearly evident. Panel D captures a fibroblast in the act of mitosis and staining of both centrosomes is visible as well as two diffuse and weakly staining ring-like structures connecting the centrosomes.
FIGURE 4. Punctate Immunofluorescent Labeling of PS2 in the Nuclear Membrane: This figure is a key part of the authors' argument that the PS2 staining observed in COS cells and in fibroblasts is really a punctate staining of the nuclear membrane. Round lymphoblastoid cells in solution were labeled with either PS2-L or PS2-N and nuclei were counterstained with DAPI. The authors argue that by changing the plane of focus they can "catch" the surface of the nuclear membrane and compare it to a "slice" through the center of the nucleus to demonstrate localization of signal on the nuclear membrane. Panel A represents a slice through the nucleus with PS2-N; perinuclear punctate staining can be observed. Panel B represents the capture of the surface of the nuclear membrane with PS2-N; punctate staining of what would appear to be the nucleoplasm is now observable. Panels C and D parallel Panels A and B but make use of the PS2-L antibody.
FIGURE 5. PS2 Colocalization with Centrosome, Nuclear Membrane, and Kinetochore Antigens: This figure is a rather extensive series of fluorescent photomicrographs that nicely demonstrates the co-localization of PS2-L and PS2-N immunoreactivity with centrosomal, kinetochore and lamin antigens in fibroblast cultures. This figure also includes photomicrographs of the controls that rule-out cross-fluorescence of antibodies or cross-reactivity of PS2-L with GST as confounds; furthermore, immunofluorescence of the PS2-L was abolished by pre-absorption with the immunogen. Panels A and D show clear co-localization of PS2 with a centrosomal antigen. Panels B and E demonstrate a more questionable co-localization of PS2 with the lamin antigen (a marker of the nuclear membrane). The PS2 nuclear staining in this series is primarily punctate while the lamin nuclear staining is primarily diffuse, with such patterns, co-localization of fluorescence is inevitable. Panels C and F demonstrate incomplete co-localization of PS2 and a kinetochore antigen, both signals are principally punctate, with some immunofluorescence co-localizing, and some independent kinetochore and PS2 immunofluorescence as well. In the text of the paper, the authors admit to a wide range of variation in staining intensities and staining patterns which they principally ascribe to differences in the stages of the cell cycle.
FIGURE 6. Colocalization of PS1 with Centrosome, Nuclear Membrane and Kinetochore Antigens: This figure parallels Figure 5 using the PS1-L and PS1-N antibodies with similar results and similar criticisms. Iin general, PS1 nuclear staining is less punctate and more diffuse than PS2 nuclear staining.
FIGURE 7. PS1 Knockout Mice Lose PS1 but Not PS2 Centrosome and Nuclear Membrane Labeling: Embryonic spleen cells (Panels A-H) and adult fibroblasts (Panels I-P) from PS1 knockout mice and their wildtype littermates were cultured and labeled with fluorescent tagged PS1-L, PS1-N and PS2-L antibodies. PS2-L staining was retained in all cells, while both the PS1-L and PS1-N staining present in the wildtype littermates was lost in their knockout brethren. Although close examination of Panel F and H suggests that PS1-N centrosomal staining in embryonic spleen cells may be preserved (see lone fluorescent spot in Panel F and compare it's location to the DAPI stain of the same cell in Panel H).
FIGURE 8. Electron Microscopic Localization and Conformational Analysis of PS2 on the Inner Nuclear Membrane: All panels are electron micrographs of immunogold coupled PS2-L labeling of primary human fibroblasts. Panels A and B clearly demonstrate labeling, but the area encompassed by the micrograph is too small to give any indication of the localization of this labeling to the nuclear membrane or of the relation of this labeling to the topology of the nuclear membrane. Furthermore there is no indication of the frequency of these regions of strong immunogold labeling per unit length of nuclear membrane. Panels C and D clearly show immunogold labeling in and around centrosomes, but again, there is no indication of the relation of this level of labeling to the level of labeling of other sub-cellular structures, or even to the level of labeling of the cytoplasm in general.
Li et al. state that in this paper they "... have obtained immunocytochemical evidence that, in normal dividing cells, the Alzheimer presenilin proteins are primarily located in the nuclear membrane, in the centrosomes, and in spots on the inner nuclear membrane associated with interphase kinetochores. (Furthermore, they) suggest that the point mutations in the presenilin genes that cause FAD may affect the ability of the presenilin proteins to link the chromosomes to the nuclear membrane and to release them at the appropriate time during mitosis, thus leading to chromosome missegregation (i.e. non-disjunction) and consequent abnormalities such as inappropriate apoptosis." Intriguing though it is, Hunt Potter's non-disjunction hypothesis as it is supported and demonstrated in this paper, still clearly lies outside of the mainstream of thought about, and the facts on, the role of the PSs in AD, and, as the saying goes, extraordinary hypotheses require extraordinary proofs. For this reason I have been rather critical of most of the data and interpretations presented by Li et al. in this paper, I was however struck by several of their observations.
The authors' criticism of overexpression experiments as an accurate method of determining sub-cellular localization was rather insightful. Their decision to investigate the endogenous expression patterns of the PSs under conditions devoid of genetic manipulation or overexpression should be taken as a paradigm for all such experiments. If their PS antibodies are actually specific to the PSs, then it is also unquestionably clear that the centrosomal structures of these cells, and their nuclear membranes, are intensely immunoreactive for both PS1 and PS2. Furthermore, the paper is to be commended on its very methodical approach to experimentation. With only two exceptions (see below), the paper very clearly and adequately addressed and implemented issues of appropriate controls, checks of cross-reactivity and cross-fluorescence, confirmation of findings by multiple experiments in different cell types utilizing independent, complementary investigational techniques, and full presentation of relevant data.
Questions and Discussion Topics
Two Methodological Issues
1) It would be nice to see some PS1 ultrastructural or lymphoblastoid localization work. More generally, it would have been desirable to see parallel experimental series for both the PS1 and PS2 antibodies, as most of the data in the paper are from PS2 immunoreactivity, but the authors frequently generalize these results to PS1 as well.
Response from Li and Potter: For completeness' sake, it would of course be ideal to carry out ultrastructural analysis on all of our antibodies to PS1 and 2. However, given the near identity of the localization identified at the light microscopic level with the four antibodies, the ultrastructural localization of PS-1 might not be expected to add a great deal to that obtained for PS-2. One argument in favor of the experiment is the fact that PS-1 and PS-2 are not functionally able to replace each other inasmuch as knockout mice for either gene are inviable. This suggests that, at least at some level, the two proteins have complementary but not identical functions and therefore might have slightly different localizations.
2) One wishes there had been a table of descriptive statistics and significance values for the work relating to the preferential localization of PS2-linked immunogold particles to the nucleoplasmic surface of the inner nuclear membrane.
Response from Li and Potter: Given that several hundred particles were counted and the percentage difference between the inner nuclear membrane association and the outer nuclear membrane association was strong, we did not calculate the statistical significance since it was clearly very high.
Four Experimental Directions Questions
1) Will the same staining patterns be observed with the use of the PS1 and PS2 antibodies generated by other labs?
Response from Li and Potter: We have not yet used antibodies from other laboratories, but we have used an antibody that was commercially available for PS-1 and obtained the same staining. In addition, at least four laboratories (those of Tom Wisniewski and Blass, Frangione, Bruce Yankner, Gerry Schellenberg, and Greg Cole) have reported to us obtaining similar nuclear membrane staining with their presenilin antibodies. (In the case of Greg Cole, a photograph showing punctate nuclear membrane localization for PS1 was presented at the 1996 Neuroscience Symposium, but not commented upon.) We suspect that when other labs have a chance to reexamine their previous presenilin immuno- photomicrographs, especially of dividing cells, they will often find staining similar to that which we reported. It is not hard to obtain, but in the face of the conventional wisdom that presenilins were in the endoplasmic reticulum, and in the absence of the the kind of extensive analysis that we carried out (encouraged by the reviewers), it was probaly easier to dismiss such puntate nuclear membrane staining from artifact.
2) Can either endogenous PS1 and/or endogenous PS2 immunoreactivity be detected in biochemically confirmed nuclear membrane fractions, as separated by high-speed centrifugation techniques, from either cultured cells or whole tissues?
Response from Li and Potter: Fractionation of cell components has not yet been carried out by us, although it is at least as tricky with respect to artifact, cross-reacting epitopes, and contamination as is immunocytochemistry. Nonetheless, such an approach is worth carrying out, especially in our future searches for presenilin interacting proteins.
3) What happens to PS1 and PS2 localization and co-localization when you lock cells in various phases of the cell cycle, especially in the specific phases of mitosis and meiosis? Of special interest would be changes in PS distribution during G2, when the kinetochore number doubles. Individual kinetochores can also be visualized by electron microscopy in anaphase cells, PS-linked immunogold labeling of anaphase cells therefore should determine whether the PSs really do co-localize with kinetochores and/or kinetochore associated microtubules.
Response from Li and Potter: This is an excellent experiment that we are in the process of carrying out. Preliminary observations suggest that cells in G2 do have increased numbers of presenilin spots on the nuclear membrane, but need to be confirmed. Cells in metaphase seem to lose their presenilin localization to kinetochores, which might be expected inasmuch as kinetochores at this stage are no longer attached to the nuclear membrane and the presenilins are apparently still membrane-associated proteins. Thus the presenilins appear to have a dynamic localization during the cell cycle, and may even have different functions at different times in the cell cycle.
4) Do rates of aneuploidy (in this case specifically the rates of Trisomy 21) differ systematically between primary cultures of fibroblasts from patients with early-onset AD caused by APP mutations, PS1 mutations, and PS2 mutations? Furthermore, do these rates of aneuploidy differ from rates in sporadic (non-genetic) AD, in AD associated with the ApoE e4 genotype, and rates in age-matched controls with or without other neurological disorders? Finally, do rates of PS1 and PS2 induced aneuploidy differ within an individual, or specific class of mutation, between different, dividing cell types i.e. is the rate of aneuploidy for fibroblasts different than the rate for astrocytes different from the rate for thymocytes, etc.?
Response from Li and Potter: A paper which we have submitted several times and will hopefully be submitted for the last time presently has used fluorescence in situ hybridization (FISH) to determine the number of trisomy 21 cells in familial AD fibroblasts compared to fibroblasts from control individuals. The results are highly statistically significant that the FAD fibroblasts have an increase in trisomy 21. Most of these cell lines have been derived from presenilin 1 families, but they include several families that have not yet been genotyped. All of the cells were derived from the National Institute on Aging Cell Repository at the Coriell Institute. In this study, there was no indication of an effect of ApoE4. However, the risk factor effect of ApoE4 seems to be rather small in presenilin 1 mutant families according to Christine Van Broeckhoven and her colleagues, as well as other laboratories. A few sporadic AD individuals also had increased levels of trisomy 21. Whether this indicates that sporadic AD is caused by environmental agents that also cause problems with mitosis, or whether they reflect a difficult-to-identify autosomal recessive genetic defect is unknown. Naturally, it would also be ideal to obtain other cells for analysis, such as astrocytes, but these will be difficult to obtain in a healthy state from postmortem tissue. Such autopsy cells might have suffered mitotic or chromosome segregation defects due to anoxia prior to harvest. Nonetheless, we have attempted to carry out fluorescence in situ hybridization on sections from Alzheimer brain, but found that the high levels of autofluorescence at all wavelengths generated by lipofuscin effectively obscured the subtle FISH hybridization spots and made the analysis impossible. Perhaps FACS sorting of disaggregated cells from Alzheimer and control brain followed by FISH may be helpful.
5) Can any binding between cloned kinetochore proteins and either PS1 or PS2 be observed in vitro?
Response from Li and Potter: We are carrying out some studies to look at the interaction between presenilins and other proteins as are others. Thus far we have detected several PS-interacting proteins by thier binding to GST-PS fusion protein on a column and by the yeast two-hybrid system. We have a series of antibodies to kinetochore and centrosome proteins that can be used to screen the biochemically-identified interactors. Sequencing of the c-DNAs that we have identified by the yeast two hybrid system has identified eight novel proteins, and further analysis will need to be carried out to determine whether these yeast two hybrid identified interactors indeed do interact with presenilins, and what their intracellular localization is.
Four Theoretical Questions
1) If the functional location of the PSs really is in the nucleus as the authors propose, then what is the meaning/interpretation of the strong ER and cytoplasmic PS staining that was also evident in nearly all of their figures?
Response from Li and Potter: When we use our presenilin antibodies to stain sections of brain tissue, we found (as have other laboratories) that the majority of the presenilin localization is in the cytoplasm of neurons, although the nuclear membrane localization is also visible. It is our belief that the presenilin proteins may serve several functions in different cellular locations, at different times in development, in different cells, and at different times in the cell cycle, all of which functions would be related to their membrane association. Specifically, in neurons, the presenilins may be involved in vesicle transport or other intracellular membrane function because in these non- dividing cells they are no longer needed for the process of mitosis. There is no reason to exclude the possibility that other cells, including dividing cells, may also employ the presenilins in a cytoplasmic membrane/vesicle transport role. In that case, some cells may show solely nuclear membrane localization, while others may show a combination of endoplasmic reticulum, vesicular, and nuclear membrane localization depending on the various functions that the presenilin proteins play.
The question then becomes which of the localizations is most important for the pathology of Alzheimer's disease. It appears likely that the FAD mutant presenilins affect APP processing, as evidenced by increases in the Ab 1-42/Ab 1-40 ratios in PS-expressing cells and tissue, but this effect may be indirect. For example, apoptosis has been shown by LeBlanc and recently by Galli and co- workers to increase Aβ secretion from several cell types. It is possible that apoptosis itself is sufficient to cause a change in APP processing. Inasmuch as there is direct evidence linking at least one presenilin (PS2) to apoptosis, and in our hands transfection of PS-expressing plasmids into cells routinely results in substantial cell death by apoptosis, it seems possible that the effect of the presenilins on APP processing is indirect through the apoptotic pathway. Inasmuch as alteration in cell cycle parameters, including chromosome mis-segregation, would be expected to cause reactive apoptosis, then the two major effects of the presenilins—chromosome mis-segregation and changes in APP processing—can be considered manifestations of the same pathogenic effect on the cell cycle.
2) How could an age- or genotypically-induced increase in the rate of non-disjunction make any difference to post-mitotic cells such as neurons? It would seem that Potter's non-disjunction hypothesis would require that the entirety of AD pathology would have to be due to the deleterious effects of non-disjunction in non-neuronal, actively dividing cell types (To be fair, the authors do mention this in their discussion. However, their best explanation is that AD neurons may be inappropriately re-entering mitosis. Even should this be so, the question is simply changed to: what could be causing AD neurons to re-enter mitosis, and would not this cause have to preceed any effects of the PSs on chromosomal segregation?).
Response from Li and Potter: We may have answered much of this already in the paper, but one more point is worth making. That is, that neurons are actually replaced at a fairly high level in the adult brain. Specifically, it has been known for some years that neurons in the dentate gyrus in the hippocampus of the adult rat are replaced at the rate of about 1 in 2000 cells per day. Recently, work from Bruce McEwen's laboratory at the Rockefeller (presented at the 1997 Society for Human Genetics Meeting and soon to be published in PNAS) have extended this finding to the dentate gyrus of monkeys. Although cell counts have not yet been completed for the monkey, the general level of cell division appears similar to that seen in the rat. A simple calculation shows that if one in 2000 cells are replaced per day (and none are replaced more than once) that in less than six years, the entire dentate gyrus neuronal population could potentially be replaced. Any defect in cell cycle control, chromosome segregation, or other mitotic defect would have a potentially devastating effect on this normal neuronal regeneration. Other parts of the brain are apparently more difficult to analyse for neuronal regeneration, but might be worth examining further. Finally, it should be pointed out that a number of laboratories have identified epitopes and enzymes in Alzheimer brain neurons that are normally never seen except in mitotic cells. Such mitotic epitopes and enzymes in Alzheimer neurons may reflect increases in intracellular calcium or other insults that are part of the reactive response to, for instance, Aβ neurotoxicity. However, if the presenilins are involved in the control of mitosis, it is not impossible that they may, when mutant, cause a cell such as a post-mitotic neuron to attempt to reenter the cell cycle.
3) If PS1 and PS2 mutations induce non-disjunction in many cell types (as would seem to be argued by Li et al.'s findings in COS cells, human fibroblasts, lymphoblastoid and spleen cells) then why are there no systemic or syndromic consequences of PS mutations, analogous to the systemic consequences of DS?
Response from Li and Potter: Most of the systemic consequences in Down syndrome are due to the severe problem of development in an organism with an incorrect number of chromosomes in every cell. However, the trisomy 21 mosaicism model of AD predicts that adult Down's syndrome and Alzheimer patients should share some clinical features. Indeed, when we looked for such features in the published literature, we found that Down syndrome individuals were hypersensitive to both cholinergic agonists and antagonists as measured by a number of assays. The most straight-forward one is to place one of these drug types in the eye and measure the dilation or constriction of the pupil. Down syndrome individuals are hypersensitive to both the cholinergic agonists and the antagonists. In collaboration with Leonard Scinto at the Brigham and Women's Hospital, we have found that Alzheimer patients are also hypersensitive to the pupil-dilating effect of the cholinergic antagonist tropicamide. A study published at the same time by Idiaquez and colleagues has shown a similar marked hypersensitivity of Alzheimer patients to the cholinergic agonist pilocarpine. Both of these results have been repeated in other laboratories with the vast majority of the results indicating a clear difference between Alzheimer and control individuals. Furthermore, the hypersensitivity is apparently restricted to Alzheimer patients and does not extend to other forms of dementia. Whether this assay will ever be refined sufficiently to serve as a diagnostic test for Alzheimer's disease or, more likely, as one of a battery of diagnostic tests, remains to be seen. The biggest problem so far is that some studies have shown a substantial number of overlap patients (either Alzheimer patients with normal sensitivity to tropicamide or control patients with hypersensitivity). Longitudinal studies and autopsy confirmation will be necessary to determine whether these patients represent true false-positives and false-negatives, or instead represent incorrect diagnosis in the cases of the false-negative Alzheimer patients, or future Alzheimer's disease in the cases of the false-positive controls. Indeed, preliminary longitudinal studies suggest that tropicamide hypersensitivity is able to identify future Alzheimer patients prior to clinical manifestations of dementia. Finally, it should be pointed out that the frequency of trisomy 21 that we have found in Alzheimer patients is usually less than 10%. Studies of known mosaic trisomy 21 Down syndrome individuals indicate that such a low number of trisomy 21 frequency of trisomy 21 cells is not sufficient to generate the usual manifestations of Down syndrome, such as low intelligence, cardiac problems, physical features, etc., but is sufficient to lead to early Alzheimer- like dementia.
4) If there is no specific mechanism for the non-disjunction of Chromosome 21 over other chromosomes, then shouldn't all aneuploids of all chromosomes be observable? Especially for those chromosomes which are non-lethal in human fetus' such as Trisomy 13 and Trisomy 18. Furthermore, Trisomy 16 is the most common trisomy in abortuses, and is therefore not immediately lethal to cells; this aneuploidy should also be represented at increased incidence if PS mutations generally predispose cells to non-disjunction.
Response from Li and Potter: Yes, we would expect increases in aneuploidy for other chromosomes and indeed these have been reported in Alzheimer fibroblasts by others. Further analysis using FISH might be useful along these lines.
- Alzheimer Presenilins in the Nuclear Membrane, Interphase Kinetochores, and Centrosomes Suggest a Role in Chromosome Segregation
- Li J, Xu M, Zhou H, Ma J, Potter H. Alzheimer presenilins in the nuclear membrane, interphase kinetochores, and centrosomes suggest a role in chromosome segregation. Cell. 1997 Sep 5;90(5):917-27. PubMed.
No Available Further Reading