IMAGING VULNERABILITY IN NEURODEGENERATION

Transcript

• 00:03Hi. Our next speaker is
• 00:04doctor Carolyn Fredericks.
• 00:06She graduated from Brown University
• 00:08and completed her medical training
• 00:10at Stanford,
• 00:11Johns Hopkins Hospital, and UCSF
• 00:13specializing in neurology
• 00:15and did a fellowship in
• 00:16behavioral neurology at the UCSF
• 00:19memory and aging center. Her
• 00:21lab uses multimodal neuroimaging to
• 00:23better understand the relationship between
• 00:25functional circuitry,
• 00:27neurodegenerative
• 00:28pathology, and symptom trajectory in
• 00:30people with preclinical neurodegenerative symptoms.
• 00:33And, clinically, she sees individuals
• 00:35with a variety of cognitive
• 00:37and behavioral complaints, including those
• 00:39with less common variants of
• 00:40neurodegenerative
• 00:41disease.
• 00:48Thanks so much for having
• 00:49me.
• 00:51So I'm gonna talk to
• 00:52you a little bit about
• 00:52what I consider the dimension
• 00:54of vulnerability in Alzheimer's disease
• 00:55and related dementia. So if
• 00:57all of you picture someone
• 00:58with Alzheimer's, you're probably picturing
• 00:59someone elderly with short term
• 01:01memory deficits,
• 01:02but there's also this period
• 01:03before that onset of clinical
• 01:05illness, right, where someone has
• 01:06a lot of Alzheimer's pathology
• 01:07in the brain, especially amyloid,
• 01:09but doesn't have clinically recognizable
• 01:11symptomatology
• 01:12yet. And even before that,
• 01:13there are groups of people
• 01:14who are much more vulnerable
• 01:15for a variety of reasons.
• 01:17So we're interested in that
• 01:18entire dimension in the lab.
• 01:20Many of you may know
• 01:21that age is the single
• 01:21biggest risk factor for Alzheimer's
• 01:23disease. You may not know
• 01:24that female sex is number
• 01:26two and that even common
• 01:27genotypes that we've all we're
• 01:29all aware of like the
• 01:29APOE4
• 01:30allele, heterozygotes where women have
• 01:32much more risk than men
• 01:33do, so we're interested in
• 01:35vulnerability
• 01:36like that as well. And
• 01:37in the next few minutes
• 01:38I'm going to take you
• 01:39on a bit of a
• 01:40whirlwind tour of several projects
• 01:41in my lab that look
• 01:42at vulnerability,
• 01:43from different different dimensions.
• 01:46The first we'll talk about
• 01:47what the functional connectome can
• 01:48tell us about vulnerability to
• 01:49Alzheimer's disease and how that
• 01:51may actually present some opportunities.
• 01:53Next, we'll talk a little
• 01:53bit more about why women
• 01:54may be at greater risk
• 01:55for aggressive Alzheimer's disease.
• 01:57We'll take a detour away
• 01:58from Alzheimer's and into the
• 01:59world of alpha synucleinopathy,
• 02:01and in particular, the preclinical
• 02:03state of alpha synucleinopathy called
• 02:04REM behavior disorder.
• 02:05And then I also wanna
• 02:06touch on some really exciting
• 02:07work by my colleagues at
• 02:09the Pet Center looking at
• 02:10the relationship between perfusion and
• 02:11tau.
• 02:13So to start off with
• 02:14in terms of the functional
• 02:15connectome and what it can
• 02:16tell us about vulnerability to
• 02:18Alzheimer's disease, The work I'm
• 02:19gonna share with you is
• 02:20spearheaded by a graduate student
• 02:22in my and Todd Constable's
• 02:23lab who just defended defended
• 02:24his dissertation, Hamid Abu Warda.
• 02:26And Hamid took advantage of
• 02:28a technique developed by Todd
• 02:29Constable's group called connectome based
• 02:31predictive modeling, which many of
• 02:32you here may be familiar
• 02:33with, essentially a linear machine
• 02:35based machine learning based model
• 02:37that allows us to predict
• 02:38measures like how the edges
• 02:40of the functional connectome can
• 02:41predict fluid intelligence or scores
• 02:43on cognitive or functional
• 02:45behavioral assays.
• 02:47For me, it had the
• 02:47idea that perhaps we could
• 02:48use this method to predict
• 02:50pathology instead.
• 02:51So in people in this
• 02:52preclinical cohort, this is from
• 02:54the A four study, which
• 02:55is a study of anti
• 02:56amyloid antibody in people with
• 02:57preclinical Alzheimer's disease, The study
• 02:59unfortunately failed, but the data
• 03:00were released to the research
• 03:02community.
• 03:02So we have this huge
• 03:03cohort of people in this
• 03:04preclinical stage of disease, and
• 03:06Hameed developed models based on
• 03:07the connections between the brain
• 03:09from fMRI to see if
• 03:10we could predict focal tau
• 03:11in different regions of the
• 03:12brain.
• 03:13And now you might expect
• 03:14that we'd be best at
• 03:15predicting tau in the very
• 03:16early Brock stage regions of
• 03:17the brain, such as entorhinal
• 03:19cortex,
• 03:19perhepocampal
• 03:20gyrus, because these people are
• 03:21so early in illness. That
• 03:23is not what we found,
• 03:23and those models were actually
• 03:25some of the worst performing
• 03:26that Hamid built, and those
• 03:27are in the blue box.
• 03:28Instead, we were best able
• 03:29to model tau in regions
• 03:30like the posterior cingulate and
• 03:32precuneus,
• 03:33which are important hubs in
• 03:34the default mode network, a
• 03:35network that seems to be
• 03:36preferentially targeted in Alzheimer's and
• 03:37that's observed short term memory
• 03:39function, but also are two
• 03:40of the biggest brain hubs
• 03:41just in general,
• 03:43in terms of the sheer
• 03:44number and importance of connections
• 03:45they have with other brain
• 03:46regions. So in retrospect, it
• 03:48may make sense that we're
• 03:49able to use the functional
• 03:50connectome best to predict tau
• 03:51in areas with the most
• 03:52functional connections.
• 03:54So building on that work,
• 03:55Mead wanted to move forward
• 03:57and see if we could
• 03:58identify,
• 03:59based on these functional connections,
• 04:01patients who might be more
• 04:02vulnerable to aggressive disease or
• 04:03patients who might benefit from
• 04:05specific treatments.
• 04:06And to do this, I
• 04:07should also say he he
• 04:08took this work and externally
• 04:09validated it in a clinical
• 04:11cohort, the ADNI cohort. So
• 04:12these models still apply for
• 04:14people who have clinical disease,
• 04:15not just preclinical.
• 04:18So moving forward, he used
• 04:19a clustering approach that was,
• 04:21actually pioneered by Leanne Williams
• 04:23right here at Yale and
• 04:23published in Nature Medicine last
• 04:25summer, and that essentially uses
• 04:27the functional connectivity scores of
• 04:28individuals that were derived from
• 04:30connectome based predictive modeling and
• 04:32puts them into clusters based
• 04:33on how much their edge
• 04:34weights resemble each other. So
• 04:36for our particular study, the
• 04:37best model was, was a
• 04:39two cluster model.
• 04:41And no need to read
• 04:42through this in-depth, but just
• 04:43to say that the patients
• 04:44who were in cluster one
• 04:45versus cluster two didn't differ
• 04:46in terms of their APOE
• 04:47status or how much amyloid
• 04:48they had in the brain.
• 04:49None of these sort of
• 04:50obvious factors that you might
• 04:52measure.
• 04:53However, when he looked at
• 04:54where their tau was, and
• 04:55again these clusters were derived
• 04:56using the functional edges only,
• 04:59they look quite different. So
• 05:00cluster one is individuals who
• 05:01have a very typical distribution
• 05:03of tau for someone with
• 05:04early Alzheimer's disease. They have
• 05:05a lot of it in
• 05:06the limbic regions, mesial and
• 05:07lateral temporal lobes. Cluster two
• 05:09looks really different, right?
• 05:11Even if you're not a
• 05:12brain imager you can appreciate
• 05:13that there's there's a bunch
• 05:14more tau in these these
• 05:16cortical regions, especially the parietal
• 05:18nodes that we talked about
• 05:19before,
• 05:19that are such important brain
• 05:21hubs.
• 05:22So Hamid next asked so
• 05:23the a four study, again,
• 05:24is the dataset we're using.
• 05:25They had a bunch of
• 05:26people who were just kept
• 05:27on placebo through the course
• 05:28of the study, so we
• 05:29can sort of see what
• 05:29the natural history is for
• 05:31people who are in cluster
• 05:32one or cluster two based
• 05:33on their functional connectomes at
• 05:34baseline.
• 05:36And what he found is
• 05:37that both groups do more
• 05:38poorly over time as you
• 05:39would expect for anyone with
• 05:40Alzheimer's disease,
• 05:42but that cluster one has
• 05:43a much slower progression, the
• 05:44ones with the typical tau
• 05:45distribution, whereas cluster two has
• 05:47a much more dramatic decline
• 05:49on this primary out endpoint,
• 05:51composite cognitive measure that the
• 05:53a four study reports, the
• 05:54PAC.
• 05:56This is the data from
• 05:57the entire phase three,
• 05:59study that that report of
• 06:00the negative study. So solanizumab
• 06:02in general failed to delay
• 06:03cognitive decline. You can see
• 06:03there's no separation between the
• 06:03placebo and decline. You can
• 06:04see there's no separation between
• 06:06the placebo and solenozumab groups
• 06:08at any point.
• 06:09And this was an expensive
• 06:10study too, so a big
• 06:12disappointment to all of us
• 06:13in the field. What if
• 06:15you look at how people
• 06:16do in cluster one versus
• 06:17cluster two on the drug?
• 06:19And this fascinated us. So
• 06:20if you look at people
• 06:21with that typical tau pattern,
• 06:23they show a similar response
• 06:24or lack of response to
• 06:25the drug as the overall
• 06:26study did. But when you
• 06:27look at those people with
• 06:28the more aggressive clinical course
• 06:30and that cortical pattern of
• 06:31tau who ended up sorted
• 06:33into cluster two, they actually
• 06:34do really well on this
• 06:36drug.
• 06:36So they have a forty
• 06:37eight percent reduction in cognitive
• 06:39decline on solenozumab
• 06:40versus placebo, and this is
• 06:41about a third of those
• 06:42who enrolled in the study.
• 06:44Hamid then went back and
• 06:45looked at how much tau
• 06:46was in that that region
• 06:48that includes the precuneus and
• 06:49posterior cingulate. And it turns
• 06:51out that individuals who had
• 06:52a high baseline level of
• 06:54tau in those regions are
• 06:55the ones who benefit most
• 06:56from the drug, as you
• 06:57can see in this heat
• 06:58map here.
• 07:00So we're really excited about
• 07:01this finding, and I think
• 07:02that it'll be,
• 07:03it has the potential to
• 07:04open up a new era
• 07:05where we can use fMRI
• 07:06to identify individuals who may
• 07:08more be maybe more vulnerable
• 07:09to aggressive disease or may
• 07:11be more responsive to specific
• 07:12treatments,
• 07:13at a fraction of the
• 07:14cost of Taupep.
• 07:18Moving on and and turning
• 07:19completely to a different topic,
• 07:21we are also very interested
• 07:22in why women are at
• 07:23more risk for aggressive AD,
• 07:24and this is work that
• 07:25I did soon after my
• 07:26arrival at Yale with Bronte
• 07:28Fisiktani, who is now an
• 07:29MD and about to start
• 07:30her psychiatry residency after graduating
• 07:32from the University of Washington.
• 07:34And what Bronte and I
• 07:35did was we looked at
• 07:35the Human Connectome Project Aging
• 07:37study, which looks at healthy
• 07:38men and women over the
• 07:39course of aging and obtains
• 07:40very high quality fMRI, we
• 07:42found this interesting pattern where
• 07:43these posterior nodes in the
• 07:44default mode network, again the
• 07:46network that serves short term
• 07:47memory performance and that seems
• 07:49to be especially targeted in
• 07:50amnestic disease,
• 07:51they have higher connectivity than
• 07:53men do in the posterior
• 07:54nodes of that network. We
• 07:55broke it down by, by
• 07:57decade and found that this
• 07:58was especially for women in
• 07:59their 50s, which of course
• 08:00is an interesting time in
• 08:01the reproductive life cycle for
• 08:03most women. This pattern still
• 08:05holds in the preclinical a
• 08:06four data set, So women
• 08:08who actually have amyloid pathology
• 08:09in their brains, if anything,
• 08:10have more so, this posterior
• 08:12hyperconnectivity.
• 08:14And now a very talented
• 08:15grad student in the lab,
• 08:16Jordan Galbraith, is taking this
• 08:18idea
• 08:19and looking into hormone levels
• 08:21in perimenopause
• 08:22to see whether levels of
• 08:23estrogen or FSH might correspond
• 08:25to some of these connectivity
• 08:26changes in women who are
• 08:27about to go through the
• 08:28menopausal transition and finding that,
• 08:30in fact, connectivity in the
• 08:31default mode network seems to
• 08:33correlate tightly with higher levels
• 08:34of FSH in perimenopause.
• 08:37I think that I am
• 08:38going too slowly, so I'm
• 08:39going to skip our detour,
• 08:42into alpha synuclein,
• 08:44but I would be delighted
• 08:45to talk about this work,
• 08:46which is in collaboration with
• 08:47Al Powers with our graduate
• 08:48student, Rena Vine, if we
• 08:50have time in the questions.
• 08:52And instead, I will share
• 08:53with you a little bit
• 08:54about,
• 08:55the use of MK sixty
• 08:56two forty by two colleagues
• 08:58in the pet center, Nicola
• 08:59Aigel and Meva Daneout.
• 09:02MK sixty two forty is
• 09:03a second generation tau tracer.
• 09:05It has a lot of
• 09:06selectivity and high affinity for
• 09:07tau, and it also shows
• 09:08relatively fast brain penetration,
• 09:10which allows us to use
• 09:11the early phase of dynamic
• 09:13scans to derive this r
• 09:15one parameter that essentially is
• 09:16an index of relative brain
• 09:17perfusion.
• 09:18And Nicola and Meva and
• 09:19their group have shown that
• 09:20these R1 values reveal regional
• 09:22perfusion differences in people with
• 09:24Alzheimer's disease, including earlier stages,
• 09:26and cognitively unimpaired individuals, and
• 09:28that those correlate strongly with
• 09:30fifteen o water PET. So
• 09:31this is is supporting its
• 09:33uses to surrogate marker marker
• 09:34of perfusion.
• 09:36And that means that essentially
• 09:37if we get an MK
• 09:38scan we can measure both
• 09:39perfusion and tau single scan
• 09:40with no added cost or
• 09:42dose.
• 09:43What's very cool is the
• 09:44work that they've done in
• 09:45dizzy across disease stages of
• 09:46AD. So on the left
• 09:48you see controls who don't
• 09:49have amyloid in their brains
• 09:50there's no consistent correlation between
• 09:52perfusion that r1 parameter and
• 09:53tau, but then if you
• 09:55look at individuals
• 09:56with preclinical AD like the
• 09:57a four cohort we were
• 09:58talking about before you can
• 10:00see that higher perfusion is
• 10:01actually linked with early tau
• 10:02deposition, and this is thought
• 10:04to be maybe a compensatory
• 10:05response. Once you get into
• 10:06clinical illness in the patients
• 10:08with mild cognitive impairment and
• 10:09AD, high tau is associated
• 10:11with low perfusion, which might
• 10:12reflect a breakdown of that
• 10:14compensatory
• 10:15process and a shift towards
• 10:16neurodegeneration
• 10:17and decline.
• 10:18So taken together, r one
• 10:19seems to serve as a
• 10:20dual marker, compensation in early
• 10:21disease and dysfunction in later
• 10:23stages.
• 10:25With that, I want to
• 10:26thank all of you for
• 10:27for your attention and our
• 10:28sponsors and collaborators and the
• 10:30members of the lab. Thanks
• 10:31so much.