Developing Autism Spectrum Disorder Therapeutics

Part 1: Developing Autism Spectrum Disorder Therapeutics: Treatment Targets and Tools from 30,000 Feet

00:00:07;18 Hi.
00:00:08;18 My name is Bryan King.
00:00:09;18 I'm the Robert Porter Distinguished Professor of Child Psychiatry
00:00:13;28 and the Vice-Chair for Child and Adolescent Psychiatry
00:00:16;23 at the University of California, San Francisco.
00:00:18;25 And I'm delighted to be here to walk you through developing autism therapeutics.
00:00:26;03 What we're aiming to do today is to take a high-level view of the process of the development
00:00:32;02 of therapeutics in autism spectrum disorder, and we'll highlight the objectives.
00:00:39;04 In order to develop therapeutics, you need to know what it is you're aiming at.
00:00:42;25 And one of the things that we spend a considerable amount of time thinking about, clinically,
00:00:47;17 is what the common therapeutic targets are in autism spectrum disorder.
00:00:52;05 Also, I want to highlight some of the rate-limiting steps that we've experienced over time
00:00:57;11 in terms of progress to date and challenges that have limited that progress, and use this opportunity
00:01:04;04 to highlight where the future is going to take us and where the emerging opportunities are
00:01:08;08 in this field.
00:01:10;24 So, it's good to start out with what it is that we're talking about in terms of autism.
00:01:16;01 And the definition of the disorder has evolved over time.
00:01:20;08 And at the same time, it has maintained its fidelity to the original descriptions of the disorder.
00:01:28;11 Leo Kanner described the criteria that distinguished this rare condition back in the 1940s.
00:01:36;08 And among the criteria that he put forward were an obsessive desire for the
00:01:45;11 maintenance of sameness, an anxious desire to retain sameness.
00:01:50;23 And you'll see that in the current definition.
00:01:55;28 There are two criteria that the DSM-5 calls out for the diagnosis of autism.
00:02:02;03 One is deficits in social communication and social interaction.
00:02:07;15 And the three areas that we need to see in order to satisfy this major criterion are
00:02:13;28 deficits in social-emotional reciprocity, in nonverbal communicative behaviors,
00:02:19;13 and in developing, maintaining, and understanding relationships.
00:02:22;09 You can see how these stack on to one another, that particularly the first two criteria
00:02:27;20 would almost always lead to problems developing, maintaining, and understanding relationships.
00:02:33;27 These need to occur across multiple contexts and multiple settings.
00:02:37;21 And this... this difficulty needs to significantly impact an individual's function.
00:02:45;26 And then the other major criterion is restricted, repetitive patterns of behavior or interests.
00:02:52;20 And here, of the four criteria that can be present, one needs at least two in order to
00:02:58;06 meet diagnostic criteria.
00:02:59;16 So, we're looking for typically stereotyped and repetitive motor movements; or repetitive speech;
00:03:06;19 idiosyncratic speech; insistence on sameness -- as Kanner called out in the very first definitions;
00:03:14;28 highly restricted interests -- and these can be on areas that might
00:03:20;22 not necessarily be unusual in and of themselves, so one might have an interest in things relating
00:03:26;26 to the Titanic, for example, but the degree and focus on that interest is characteristically
00:03:35;17 unique in autism in that it can become all-consuming... the entire world is focused around
00:03:41;11 one of these particular interests;
00:03:43;19 and sensory hyper- or hypo-sensitivities were added in the DSM-5.
00:03:48;22 They had been present in the earliest descriptions but had fallen off the radar in terms of criteria,
00:03:53;28 and now have been added back.
00:03:57;04 One comment about the diagnosis of autism over time is that it is not unlike this image
00:04:03;09 that one sees on the approach into the city of San Francisco to the airport from the north.
00:04:10;10 You circle around these wetlands.
00:04:13;10 And you have a very characteristic picture of the wetlands.
00:04:16;23 And frankly it's...
00:04:18;11 I think everyone would agree that these are wetlands.
00:04:21;10 You... you can recognize them from 30,000 feet.
00:04:25;20 But when you drill down into the weeds, as it were, of the wetlands, and think
00:04:33;00 particularly about where the boundaries are around the wetlands, it becomes more challenging.
00:04:39;17 So, we can see that there... you know, there... there are pretty discrete boundaries
00:04:46;02 around what appears to separate the wetlands from... from the ocean.
00:04:53;08 But as the tide comes in and out, that boundary is going to change.
00:04:56;14 And if you... if you look up here, one could argue, are these wetlands as well?
00:05:04;11 Where you don't have quite as much exposure as down below?
00:05:08;02 And one could spend hours trying to figure out where the boundary is around the wetlands.
00:05:14;15 And that problem exists in autism as well.
00:05:16;25 So, while we have really discrete criteria, and it's easy to come to consensus about
00:05:24;12 whether someone has autism or not in most cases, the way the diagnosis is made, as you saw,
00:05:31;01 is based on the behaviors that we see and not what the underlying pathophysiology is.
00:05:38;00 And it could very well be that the behaviors that rise to the level of meeting the criteria
00:05:43;10 for autism are, again, not unlike the wetlands here, where in some instances it's obvious
00:05:50;03 and in some instances the pathophysiology may be the same but may not carry forward
00:05:54;18 to the same degree in terms of behavioral expression.
00:05:57;22 And you're left with this problem that will be resolved in the future, when we have diagnoses...
00:06:05;09 diagnostic criteria that incorporate some of the biology, and we'll hear about that
00:06:10;02 in Dr. State's talk.
00:06:12;08 The diagnostic boundaries around autism are very likely to be redrawn.
00:06:17;17 We'll come back to this image, but in summary the diagnosis is reliable but the boundaries
00:06:22;21 are not discrete.
00:06:24;10 And this is important as we think about the development of therapeutics.
00:06:28;00 Again, if our focus is on the manifested behaviors, it may ultimately be different, or it's
00:06:35;12 likely to be different, than if we're focusing on a treatment that's based on a knowledge
00:06:40;19 of what the underlying pathophysiology is.
00:06:43;13 So, the other fact that has emerged over time is that autism does not prevent the likelihood
00:06:53;21 of experiencing other psychiatric disorders that are common in the general population.
00:06:59;22 In fact, in this study that was done in Kaiser Permanente's Northern California database,
00:07:07;11 where Lisa Croen and colleagues identified 1,500 adults with autism and identified
00:07:14;24 a matched control group of 15,000 typically developing adults without autism, and simply
00:07:22;12 asked the question, if we look in our database at medical conditions, particularly diagnoses,
00:07:29;21 that have been identified in the group with autism and also in this control group,
00:07:36;03 how do they compare?
00:07:37;22 And so, you see in the typical, in the control group, that we have something on the order of
00:07:45;11 10% of the population with depression diagnoses, another other equal number with anxiety diagnoses,
00:07:53;07 suicide attempts, schizophrenia, and the like.
00:07:56;12 We essentially have other chapters in that DSM book that I showed you earlier.
00:08:01;26 When we overlay the population with autism, here in blue, what you see is that virtually...
00:08:09;17 for every one of these psychiatric illnesses, the prevalence in the setting of autism
00:08:16;04 is greater than it is in the general population.
00:08:19;10 And so, in the next slide, just calling this out in terms of odds ratios, what you see is that
00:08:26;03 if you're an adult with autism the likelihood... the likelihood that you are
00:08:31;05 going to be diagnosed with a schizophrenia-like disorder, obsessive-compulsive disorder,
00:08:37;00 psychotic illness, and so on, and particularly important things like suicide attempts, are all elevated
00:08:44;03 in the setting of autism.
00:08:45;17 These of course are also potential therapeutic targets.
00:08:49;25 So, as we're moving through the possibilities here, we have autism as the core disorder,
00:08:56;12 but we have these co-occurring conditions.
00:08:59;18 Now, this is a slide taken from Matt Siegel's group that incorporates reasons for
00:09:07;04 psychiatric hospitalization in children and adolescents with autism at centers that have been specifically
00:09:12;27 set up to support children with autism and their psychiatric illness.
00:09:17;15 And what you see here are, in fact, not psychiatric diagnoses but a list of maladaptive behaviors
00:09:25;20 that have resulted in the need for psychiatric hospitalization.
00:09:29;11 So, we have problems like aggression and self-injurious behavior and the like that are so interfering
00:09:37;01 that they require an inpatient psychiatric stay.
00:09:40;23 And these, again, are common targets for potential therapeutics.
00:09:47;13 One common measure that has been used in many clinical trials is the aberrant behavior checklist.
00:09:53;01 And it's a 58-item checklist that has a number of subscales that have been identified
00:09:57;04 based on factor analysis.
00:09:59;01 And this is just one of the 5 factors of this scale that's been called the irritability subscale,
00:10:04;17 because many of the items suggest that there may be underlying irritability --
00:10:09;13 irritable and whiny is actually one of the specific items on the survey.
00:10:15;24 But what you see is that there's a mix of other behaviors here as well that could include
00:10:21;04 mood lability, aggression, self-injurious behavior, and the like.
00:10:26;00 And we'll come back to this but, again, the fact that this out... this is often used
00:10:31;21 as an outcome measure in clinical trials speaks to how these behavioral problems
00:10:37;20 can also be targets for therapeutics.
00:10:39;25 So, where are we?
00:10:41;08 Well, we have therapeutic targets that could be the core features of autism.
00:10:46;06 And here we're talking literally about the two major criteria that form the basis for
00:10:51;26 the disorder.
00:10:52;26 So, can we imagine that there would be pharmacologic agents that would specifically target
00:11:00;14 social communication difficulties?
00:11:03;02 Perhaps there are behavioral intervention strategies that would target
00:11:05;20 social communication deficits.
00:11:08;19 Are there pharmacologic or other therapies that we might look at that are
00:11:11;27 specifically targeting repetitive, restrictive behaviors.
00:11:14;09 Again, those are the core features.
00:11:18;18 We also have therapeutic targets that include the psychiatric disorders that I highlighted
00:11:23;18 from the Kaiser database.
00:11:25;16 And there, again, multiple groups have described increased risk for psychiatric disorders in autism.
00:11:31;13 And then we have therapeutic targets that involve other behavior or other factors
00:11:37;23 that don't necessarily rise to a specific psychiatric diagnosis but nonetheless could be significant targets.
00:11:45;21 And moreover, none of these completely overlap.
00:11:49;03 And so we have a very complicated mix, potentially,
00:11:52;24 of targets that people are exploring in clinical trials.
00:11:57;25 So in summary, the illnesses are common, behavioral disturbance is common, and of course
00:12:03;23 the core features are common.
00:12:05;10 Now, medication is often used in the setting of autism.
00:12:09;25 In fact, multiple studies have looked at the prevalence of the use of psychotropic drugs,
00:12:14;26 and current evidence supports the use of medication for many of these targets that I've just highlighted.
00:12:21;17 There are as yet no drugs that have gained specific indications for the treatment of
00:12:27;01 core features of autism.
00:12:29;00 So, if we're asked today, what is the medication that you use for autism?, the answer is,
00:12:35;14 there is no such thing.
00:12:37;28 But there are lots of medications that are used in the setting of autism.
00:12:41;22 And age data, as I mentioned, indicates that these medications are used quite commonly.
00:12:47;08 In fact, virtually every study that's looked at the prevalence of medication use in autism
00:12:53;02 comes back with rates that vary from 30 to upwards of 70%, as was true in this study.
00:12:59;05 In fact, in this particular dataset taken from a Medicaid database in the... in Colorado,
00:13:10;19 over 70% of children had received at least one psychotropic medication by the age of 8.
00:13:16;12 And what you see here are the... the top three categories or classes of medication that are
00:13:22;04 most commonly used: antidepressants, stimulants, and tranquilizers, anti-psychotic drugs.
00:13:29;07 What I want to highlight on this slide is also the... the date.
00:13:34;04 So, this particular data comes from 2007.
00:13:38;08 And the reason I wanted to share the data from 2007 as opposed to more current data,
00:13:44;20 which absolutely exists, is to highlight the fact that these data have not changed in over a decade,
00:13:51;09 that these three classes of drugs have been and remain the most widely prescribed
00:13:56;28 drugs in the setting of autism.
00:13:59;06 And it's... it's actually a... an indictment, really, of the progress that we've made
00:14:06;15 as a field in terms of our identification of therapeutics.
00:14:10;04 And I want to talk for a minute about what some of the rate-limiting steps have been,
00:14:16;03 and highlight that irritability has been one of the most widely studied indications
00:14:26;01 for pharmacotherapy.
00:14:27;01 So, a number of studies have been focused on irritability as a target.
00:14:33;06 And so, if we pull out irritability for an example, I want to make another point
00:14:39;19 about the process of identifying therapeutics.
00:14:43;24 And we'll change that irritability circle to an 8-ball.
00:14:47;12 So, the... the point in billiards is that when you have the opportunity to sink the 8-ball
00:14:54;22 you have to actually declare where it's gonna go.
00:14:58;13 So, you have to say that it's going in the corner pocket, or the side pocket,
00:15:02;23 or you just have to... you... you need to declare a pocket.
00:15:07;15 And if it goes in that pocket you win.
00:15:10;22 And if it goes in a different pocket you lose.
00:15:14;15 This sadly is true of clinical trials as well.
00:15:18;18 So, what we need to do as investigators is to identify what our primary outcome is
00:15:25;06 before we start the trial, before we take that shot.
00:15:28;19 And if we find evidence that our primary outcome improves as a result of the intervention,
00:15:37;02 we win.
00:15:39;13 If we don't, that's too bad.
00:15:43;16 And it can take five years to complete one of these large clinical trials in the population
00:15:48;03 with autism.
00:15:50;01 There's a lot riding on getting that call right up front, before you go down this path.
00:15:57;27 And what happens is that sometimes you're right, but what I wanted to illustrate
00:16:04;24 on this slide, which was one of the pivotal trials that led to Risperidone getting an indication
00:16:11;18 for the treatment of irritability in autism, because irritability
00:16:16;26 was the... was the primary outcome.
00:16:20;00 And you can see that the improvement in red... so, the size of the bar correlates with the...
00:16:26;15 the size of the drop in point... in score of the irritability subscale that I
00:16:31;23 showed you earlier.
00:16:32;23 So, a significant change in the irritability... irritability subscale associated with receipt
00:16:38;01 of Risperidone.
00:16:40;07 But what you also see is that hyperactivity significantly improved.
00:16:45;06 If we had done this trial and not seeing the change in irritability, but that the signal
00:16:51;23 for hyperactivity had carried through as significant, it would still be a failed trial.
00:16:58;20 While we could say that in our secondary analysis we had identified that hyperactivity improved,
00:17:05;02 the... the best you could conclude at the end of the day is that we should do
00:17:09;04 another study and focus on hyperactivity as our primary outcome so that we could move the process forward.
00:17:17;09 So, one could argue about the threshold here, but the point is that the identification of therapeutics,
00:17:26;06 and moving them forward to a place where you can actually get them
00:17:32;13 into the... clinical use, because of indications that the FDA or other body is conferring
00:17:43;04 on the drug, require at least three of these pivotal trials, and... and you can
00:17:49;13 imagine that this process could unfold over a number of years, and that there could be drugs
00:17:54;21 for which there may be a positive signal that are not moving forward because that first bet
00:17:59;15 on the primary outcome was wrong and that 8-ball went into a side pocket
00:18:07;13 instead of the corner that was called.
00:18:09;24 So, that's a problem, just in terms of the structure, moving forward.
00:18:15;28 Another area that I think is important to comment on, as I showed you earlier about
00:18:21;17 whether co-occurring psychiatric disorders that... you know, that they can be targets
00:18:28;16 of treatment in the setting of autism.
00:18:30;04 So, we have a situation, here, where we have ADHD that's in the setting of autism.
00:18:37;26 And we have ADHD in the general population.
00:18:41;08 And a question that logically could be asked is, if we already know what works for ADHD
00:18:47;26 -- we have dozens and dozens of clinical trials that support the use of stimulants, for example,
00:18:54;00 and other drugs as well -- why can't we just use those therapeutics for the ADHD in autism?
00:19:00;18 Do we really need to do independent studies of ADHD in autism in order to move the field forward?
00:19:08;24 And the answer is mixed.
00:19:10;28 So, in one of the pivotal trials associated with... or, the largest trials associated
00:19:17;13 with stimulants in the setting of autism, here, the Research Units on Pediatric Psychopharmacology,
00:19:25;01 or RUP trial, revealed that in about half of cases methylphenidate was effective.
00:19:33;17 So, 50% of the children who received active treatment, or methylphenidate, responded.
00:19:40;15 And 19% of those who received placebo responded... responded to placebo.
00:19:47;24 And so this is a significant outcome.
00:19:51;08 Interestingly, nearly a fifth of those kids in the trial had to drop out because of
00:19:59;11 adverse events associated with the receipt of the active treatment in the autism group.
00:20:05;14 Now, we can compare this to a trial of ADHD alone.
00:20:08;22 The Multimodal Treatment of ADHD Study, which was the largest of its kind.
00:20:16;10 And in this study, what you see is that the response was much better: 70% of the kids
00:20:22;19 with ADHD responded to methylphenidate.
00:20:26;02 The placebo response was half of what it was in the autism study.
00:20:32;10 And I can't...
00:20:33;13 I couldn't even put the words of the text in that tiny little slice that has 2% there,
00:20:38;25 which were the dropouts due to adverse events.
00:20:42;24 And so what... what I think clearly is represented here is that, in fact, we probably can't just
00:20:49;03 walk straight over from ADHD to ADHD in autism, that there are some very important differences
00:20:57;05 in terms of how these populations are responding to the receipt of the active treatment,
00:21:04;20 particularly in terms of the ability to even tolerate the medicine itself.
00:21:10;24 One of the things that has also limited our progress as a field is that studies like these end
00:21:20;24 when you call in the jury at the end of the five-year clinical trial.
00:21:27;02 And so what you're left with is these data.
00:21:32;20 And what we need as a field is to take a look at this 20%, for example, and ask, well...
00:21:40;12 so, they clearly are very different within this universe of people with autism.
00:21:45;22 They've demonstrated that they are very sensitive to this drug.
00:21:50;14 We have a physiological probe if you will that we have just evaluated.
00:21:56;18 So, what should we use in this situation to treat the ADHD?
00:22:02;04 This arguably is the most informative outcome from this trial, and sadly, you know,
00:22:09;00 people disappear and go their separate ways after the... after the trials are complete.
00:22:13;15 But going forward we need to anticipate that we have another therapeutic lined up that
00:22:20;12 we can look at for those who don't respond to the first intervention.
00:22:25;22 So, there have also been clinical trials that have looked at core features, and I wanted
00:22:32;16 to highlight one.
00:22:33;16 I told you earlier that antidepressants were among the most widely prescribed drugs
00:22:39;19 in the setting of autism.
00:22:40;28 They still are.
00:22:41;28 They have been for the past decade.
00:22:44;01 The largest clinical trial that we did looking at the use of an antidepressant
00:22:49;14 targeting repetitive behaviors in autism revealed no difference between active treatment and placebo.
00:22:56;03 What you can see here in terms of the curves is that they are completely overlapping,
00:23:03;25 so no hint, even, of a difference favoring active treatment.
00:23:10;06 So, this obviously was a trial that didn't yield a positive outcome in terms of supporting
00:23:17;05 the use of antidepressants for repetitive behaviors, but... but how do we reconcile
00:23:22;15 the fact that these drugs are still among the most widely prescribed in this population
00:23:27;21 when we have no data that supports their effectiveness, at least for repetitive behaviors?
00:23:35;04 And so one of the other areas that this illustrated, in terms of problems with clinical trials,
00:23:40;24 is the placebo response.
00:23:42;22 So, the higher the placebo response, the more difficult it becomes to identify the separation
00:23:49;07 between that and the active treatment.
00:23:53;12 With a large placebo response in a trial, you have to enroll significantly more subjects
00:24:00;21 in order to be able to capture, you know, a smaller difference between active treatment
00:24:05;05 and the placebo.
00:24:07;13 And what I've shown you on this slide are the... the numbers... the placebo response
00:24:13;24 numbers across a number of autism trials, where you see that it's not uncommon at all
00:24:19;12 to see placebo responses in the 20% range.
00:24:24;03 And in some studies, when that response rate goes upwards of 30 or, you know,
00:24:28;13 in some cases even 40 or 50, the chance of seeing a signal with your active treatment becomes almost nil.
00:24:37;16 So, there have been active efforts in the field to look at ways to mitigate or
00:24:43;27 reduce the placebo response.
00:24:45;12 And one of those has been to look back and see what factors might lead to a greater likelihood...
00:24:52;24 likelihood of placebo response.
00:24:54;03 And one of the things that we've identified is the illness burden.
00:24:58;23 So, if you come into a trial looking at irritability, for example, and your irritability scores
00:25:05;27 are relatively low, the likelihood of response to a placebo is significantly greater than it is
00:25:12;07 if you come into a trial with much higher irritability scores.
00:25:18;26 And there's another thing.
00:25:20;24 And that is that the populations that are coming into these clinical trials are
00:25:24;28 very heterogeneous.
00:25:26;02 So, we have some conditions -- and you'll hear more about this from Dr. State --
00:25:32;22 that are associated with identifiable genetic factors that contribute risk.
00:25:39;06 And some disorders are very much associated with a greater likelihood of autism:
00:25:44;22 Fragile X syndrome, for example, where the prevalence of autism in Fragile X may be upwards of
00:25:51;05 30% of the population.
00:25:52;13 Rett syndrome is another neurogenetic disorder that's been associated with autism symptoms.
00:25:59;14 And that's like comparing apples and oranges.
00:26:02;13 When we bring... if we were to do a clinical trial and... and allow people in with
00:26:08;03 Fragile X and Rett syndrome, we would not expect them to respond in the same way to therapeutics.
00:26:15;17 Increasingly, molecular genetics is highlighting that there are glutamatergic abnormalities
00:26:21;20 in autism.
00:26:23;16 And that abnormalities can end up either in the place of hyper-excitability or hypo-excitability.
00:26:34;10 And your therapeutic is going to move you in one of those two directions.
00:26:38;16 But if you have a mix of some people getting better while other people are getting worse,
00:26:42;21 it gets washed out in your overall signal.
00:26:45;00 So, you can see here the beginnings of identifying how important it is to know what the underlying
00:26:52;06 genetic background is for some of these trials.
00:26:57;05 And in fact, where we're at -- going back to the wetlands -- is, while we can identify
00:27:04;24 the universe of people with autism, we need to be much better about being able to identify
00:27:09;28 where we're starting out in terms of the underlying pathophysiology or severity, because the pathway out
00:27:16;05 is going to be very different depending on where you're starting.
00:27:21;10 And so there have been multiple attempts to focus on heterogeneity.
00:27:27;02 And I want to...
00:27:28;05 I want to speak about that in a moment here, but, again, back to the wetlands.
00:27:33;01 You can well imagine that if your goal is to extricate yourself from being stuck
00:27:37;11 in the wetlands somewhere that your path out is gonna be very different if you're,
00:27:42;09 you know, stuck in one of these areas in the... in the midst versus one of these major canals or
00:27:48;24 you know, further out near the exit.
00:27:53;17 In fact, the... the process that we've... that we've been using over the last decades
00:27:59;28 is not unlike the Marco Polo game, where you, with successive approximations -- and certainly,
00:28:07;08 clinically, this is the strategy -- you try something.
00:28:10;20 You say Marco, and you listen for where the Polo is coming from.
00:28:14;24 And that moves you closer to getting to a better place.
00:28:19;18 But if you wanted to replicate that pathway at the end of the day, put up an algorithm
00:28:23;13 for "this is how you find Polo or Marco", whoever is... whoever you're after in this game,
00:28:29;20 that is no way to... to move forward in a reliable, reproducible manner.
00:28:36;13 And so people are looking at biomarkers, biomarkers that can help parse the heterogeneity in autism,
00:28:44;10 identifying individuals that are high risk for autism because of underlying genetic
00:28:48;10 or other factors, for example; help us narrow in on a specific subset of individuals with autism;
00:28:54;01 identify biomarkers that suggest who's going to respond to this treatment that
00:29:00;02 you're interested in within the universe, etc.
00:29:04;14 And some of those that are moving forward as active candidates involve EEG abnormalities,
00:29:09;19 for example.
00:29:10;16 And this is just one example, but multiple investigators have shown that there are differences
00:29:17;03 in EEG power that characterize some individuals with autism.
00:29:22;26 And so might we use this as a potential sorting function?
00:29:28;11 Not a diagnostic measure to say who has autism and who doesn't, but a sorting function for
00:29:33;20 who gets into our trial or not, who were interested in.
00:29:36;27 And just as one example where an EEG was obtained pre- and post-intervention, this for a psychosocial intervention,
00:29:49;11 the Denver Start Model shows kids' EEG profiles who are typically developing
00:29:58;19 and kids with autism who have received community intervention.
00:30:02;16 When you look at those who received the Early Start Denver Model, you see that
00:30:08;07 their EEG profiles match those with typical development.
00:30:11;22 And so one could potentially use this as even an outcome measure to get an early signal
00:30:19;22 back on a trial that doesn't require perhaps months of exposure to your therapeutic of interest,
00:30:27;21 and potentially accelerate the drug development process.
00:30:32;02 With autism, I think there's also interest, increasing interest, in other biomarkers
00:30:37;05 like sorting biomarkers, as those I've just highlighted with EEG, or process biomarkers.
00:30:43;01 If you're way upstream and thinking about, is this a drug that may be worth looking at?,
00:30:51;19 before you go into a clinical trial you might want to establish that the drug actually
00:30:56;18 is reaching its target in the brain, and look for a measure of target engagement,
00:31:03;06 that would save you all of the years of Marco-Polo in order to get to the... to the outcome
00:31:11;04 that you're hoping to achieve.
00:31:12;10 Obviously, if the drug doesn't engage its target, it's not something that you want
00:31:16;25 to move forward.
00:31:18;04 So, where are we at?
00:31:20;01 Well, I keep coming back to the complexity here, because there are any number of places,
00:31:28;04 potentially thousands... a thousand genes, potentially, that might discriminate one group
00:31:34;07 from another.
00:31:35;13 But as you look at the wetlands, carefully, you also see that there are pathways
00:31:41;10 that exist, that take you out of the wetlands, potentially.
00:31:47;09 So, there's been increasing attention on pathways in the brain, functional pathways where...
00:31:54;14 where some of these genes that are associated with autism are converging.
00:31:59;24 And is that a future target that we... that we can explore?
00:32:05;01 The answer of course is an unequivocal "yes".
00:32:09;13 And that's what you'll hear in Dr. State's talk, that as molecular genetics is
00:32:16;21 bringing us new tools and new potential targets that it's going to change the game dramatically
00:32:23;11 in terms of the opportunity to develop therapeutics from the inside-out, as opposed to
00:32:32;04 the Marco-Polo compass that I'm showing you here.
00:32:34;20 So, our progress to date has been limited both by our ability to capture change and
00:32:42;00 to identify outcomes of interest, as well as narrowing in on our targets.
00:32:48;26 Although the diagnosis of autism spectrum disorder is very reliable,
00:32:53;19 like... like the wetlands, it's very heterogeneous.
00:32:57;22 And the boundaries around the disorder are going to be redrawn.
00:33:00;20 It could very well be the case that what we currently think of as autism becomes a disorder
00:33:08;07 that includes autism, ADHD, anxiety symptoms in one group, and includes autism, sleep disturbance,
00:33:17;19 and hyperactivity in another group, and that those core features of autism become
00:33:23;17 much less important, as they are part of a... multiple way...
00:33:29;09 multiple different ways of expressing what the underlying pathophysiology is.
00:33:33;21 So, where the field is taking us is GPS, that molecular genetics is gonna hand us the ability
00:33:42;06 to know exactly where we are in the wetlands.
00:33:45;23 And that is going to accelerate the development of more specific therapeutics that move us out
00:33:51;16 and where we need to be.
00:33:53;21 It's a really exciting time in the field of autism therapeutics development and one
00:33:59;12 that is long overdue.
00:34:00;18 Thank you for the opportunity to present today.

Part 2: From Genes to Pathology in Autism Spectrum Disorders

00:00:07;19 Hi.
00:00:08;19 I'm Matthew State.
00:00:09;19 I'm the Oberndorf Family Distinguished Professor and Chair of the Department of Psychiatry
00:00:13;10 at the University of California, San Francisco.
00:00:15;08 And I want to talk to you today about the hunt for genes in autism, and then about
00:00:19;19 the process of moving from gene discovery to understanding something about the pathophysiology of
00:00:25;18 autism spectrum disorders.
00:00:26;21 So, you've heard already from Dr. King about the defining characteristics of autism.
00:00:31;04 I won't belabor them here.
00:00:32;20 I just want to point out that we've known for quite a long time that genetics
00:00:38;09 plays a very important role in autism.
00:00:40;15 But despite knowing that, when I began to work in this field in the late 1990s,
00:00:47;13 one of the things that drove my interest was the lack of information about the molecular/cellular/circuit level
00:00:53;24 pathology of autism.
00:00:57;14 And that is a fundamental obstacle to identifying novel forms of treatment.
00:01:03;08 And consequently, I became very interested early on in using genetics as a way to address
00:01:08;11 that problem.
00:01:09;15 Now, one of the things that I was interested in was focusing on a particular kind of genetic variation,
00:01:15;12 rare mutation, as part of the strategy.
00:01:18;17 And the reason that I was interested in doing so is that I knew that there was
00:01:23;06 a pretty reliable relationship between rare mutations and large biological effects.
00:01:29;01 And because of the strategy, which was to use gene discovery to begin to identify
00:01:35;19 a thread to pull on to understand biology, our presumption was that the larger the effects
00:01:40;08 the better.
00:01:41;08 So, we focused on rare mutation because we presumed that they would have large effects.
00:01:45;22 One of the reasons that we focused on genetics as opposed to some other approach is that
00:01:49;26 it also gives you some important potential insights that may be difficult to find
00:01:54;18 with other methodologies.
00:01:56;00 One of the most important is that... remember that genetic mutations, if they're in the germline,
00:02:00;01 and these mutations... we were looking for germline mutations... that if they're
00:02:04;16 in the germline they're there before the brain is built.
00:02:07;22 And that gives you some traction on the temporal unfolding of risk, which is... can be
00:02:14;04 a difficult problem in autism, and one that genetics, if you can be successful... one that genetics
00:02:19;00 can answer.
00:02:20;01 And obviously it... if you can find genes, it's taking you down to the molecular level.
00:02:25;05 And then hopefully would allow you to go from molecules to cells to circuits.
00:02:28;14 So, that was the strategy that we were interested in using, focusing on rare mutation as
00:02:32;21 an entree into biology, but there's one additional twist on this that I want to describe for you,
00:02:39;16 which is that we were not only focused on rare but we were also focused on de novo mutation.
00:02:44;20 And here, just to remind you, that a small amount of mutation in every generation is
00:02:48;20 introduced newly.
00:02:49;20 So, the vast majority of variation is transmitted from generation to generation, but there is
00:02:54;13 a small amount new in each generation.
00:02:56;22 And it has particular characteristics, if it's related to disease, again,
00:03:01;10 that help us potentially identify things with very large effects.
00:03:04;06 And the idea here is that de novo mutation... there's very little time for natural selection
00:03:09;12 to act on de novo mutation.
00:03:10;28 So, if the child is born with a de novo mutation, they can have very severe neurodevelopmental problems.
00:03:18;17 But as long as they're able to survive through the pregnancy, you would be able to see
00:03:23;05 a relationship between a de novo mutation and this severe phenotype.
00:03:28;03 And you would imagine that after generations of that, it would be unlikely that that would
00:03:32;15 be transmitted.
00:03:33;15 So, de novo mutations give you a particular opportunity to find very large effect variations.
00:03:38;17 And we were quite interested in that.
00:03:40;13 And that fortunately coincided with a clinical observation that families... many families
00:03:46;16 with autism look as though there might be a de novo mutation, because no one else
00:03:50;15 in the immediate family seems to be affected, and then there's an affected child.
00:03:53;26 And that... we call that a simplex family, if we... if overtly there's only a single
00:03:58;08 affected kid.
00:03:59;08 And finally, a very important piece of information arose in about 2006/2007, when several labs
00:04:07;19 were able to show, looking at submicroscopic changes in chromosomal structure, that it
00:04:12;26 looked like de novo mutation in... in chromosomal structure was... was more frequent
00:04:19;22 in kids with autism than without.
00:04:21;23 And that gave us a sense that if we could look, potentially, at the genome in higher resolution,
00:04:26;08 if we could sequence the DNA in every gene in every individual, that we
00:04:33;02 might be able to identify even higher-resolution genetic abnormalities that would take us to
00:04:39;05 the specific genes that were contributing to ASD.
00:04:42;06 But then as a consequence of a particular patient cohort put together by the
00:04:47;20 Simons Foundation in New York, called the Simons Simplex collection, we were able to study
00:04:52;27 families that not only had a single affected individual but in most cases also had unaffected siblings.
00:04:59;27 And the key here is that, in addition to seeing what was different between the parents and the children,
00:05:04;26 we were also able then to compare the affected child to the unaffected child.
00:05:11;01 And we asked the simple question, within the same family, if there's a de novo mutation,
00:05:15;16 is it more likely to land in a child with autism than you would expect by chance?,
00:05:22;08 with the idea that that would give us a signal that de novo mutation was contributing to autism.
00:05:27;27 So, as I said, the technology that we were able to use by about 2011 was called whole exome sequencing.
00:05:33;19 And this allowed us to read the DNA code of each one of the about 21,000 genes in
00:05:38;23 the genome for every family member.
00:05:40;27 And it's this level of resolution and throughput that really made an enormous difference
00:05:45;11 in our being able to be successful in gene discovery.
00:05:48;08 So, using that methodology, we were then able, in 2012, to publish a paper that looked at
00:05:56;14 about 240 families.
00:05:58;24 Now, they're called 238 trios, only to... it should say, actually, quartets not trios.
00:06:05;07 And quartets meaning that there were both two parents as well as two children:
00:06:10;07 one affected and one unaffected.
00:06:12;14 And we call the affected the proband, and we call the unaffected the unaffected sibling.
00:06:17;17 And for all the diagrams... well, on this page anyway, the red represents the affected child
00:06:22;08 and the blue represents the data from the unaffected siblings.
00:06:26;26 And what we're showing here is that, when we start to look through the genome at
00:06:31;12 de novo mutations, what we find is that if you look at mutations that are not expected to
00:06:37;19 have any biological function at all, we don't see any difference between probands and siblings.
00:06:43;28 In fact, the siblings have a slightly higher rate, but it's not statistically significant.
00:06:48;01 But then, importantly, if we begin to look at variations in the genetic code that would
00:06:52;08 be predicted to alter the function of the underlying protein, or the encoded protein,
00:06:58;11 we began to see consistently that there's a higher rate in... in affected individuals
00:07:04;12 versus unaffected individuals, and this is statistically significant.
00:07:08;00 Now, there's several things that I need to add to clarify this result.
00:07:11;26 The first is, I've told you it's de novo mutation.
00:07:15;02 More specifically, what we were looking at is germline de novo mutation.
00:07:19;12 What that means is that the mutation was introduced in sperm or an egg just prior to conception,
00:07:25;08 is there for the first cell divisions, and consequently is present in every cell in the organism.
00:07:31;08 That's what we were looking for through our sequencing.
00:07:33;15 Now, cancer, as I'm sure you're aware, may result from... also from de novo mutations,
00:07:39;03 but often what we're looking for there are what are called de novo somatic mutations.
00:07:43;27 They're... they're mutations that happen in DNA after the early cell divisions in an organism,
00:07:49;24 leaving some cells to have the mutation of interest and others don't.
00:07:55;15 In this case, again, this is germline.
00:07:57;13 So the mutation that we're interested in would not be present in any of the cells in the parents,
00:08:02;13 and would be present in all the cells in the kids, so, germline de novo mutation.
00:08:07;07 The second thing that I want to clarify here is that what we are looking at, the rate of
00:08:13;16 de novo mutation.
00:08:14;22 You can see it's the rate of de novo mutation... is somewhere in the neighborhood of
00:08:21;00 1-1.5 examples per exome.
00:08:23;13 So, of all the sequence we refined, essentially one event in... in an individual, if it was
00:08:30;13 a mutation that was expected to affect function at all.
00:08:34;02 If it was the most damaging type of mutation, which is a nonsense mutation
00:08:38;11 -- so, these are stop codons or frameshift mutations, or mutations that affect a canonical splice site --
00:08:44;20 we see an even lower rate of de novo mutation, so 1 in 10 individuals.
00:08:50;24 So, you might ask the question, is this... what are we looking at?
00:08:53;22 Is this simply that there are more de novo mut... there's a chromosomal fragility or
00:08:59;19 a sequence fragility that leads to many more mutations, kind of randomly distributed throughout
00:09:06;11 the autism genome?
00:09:07;11 But in fact, what you looking at in the next panel here is that when you ask the question,
00:09:14;09 are children with autism versus their siblings more likely to have multiple mutations?,
00:09:19;18 this is the expected distribution in the unaffected versus the... sorry, it's the observed distribution
00:09:27;06 in the unaffected versus the observed distribution in the affected of multiple de novo mutations.
00:09:33;13 And when you compare those, there's no difference.
00:09:36;07 So you're not looking at a situation where there... kids with autism have 1, 2, 3, 4, 5
00:09:42;02 de novo mutations that are leading to a greater amount overall.
00:09:46;13 What we're really looking at is kids with autism potentially having just 1 mutation
00:09:52;00 that is contributing to their being affected, and that's why we see the increased rate
00:09:56;06 in affected versus unaffected individuals.
00:09:58;18 Now, science marches forward and... and we were extremely lucky to be able to then
00:10:05;07 study the Simons Simple Collection as it grew.
00:10:08;12 So, this was a cohort of individuals that was the product of a collaboration nationwide
00:10:16;04 with more than a dozen clinical sites that very carefully phenotyped individuals with autism,
00:10:22;09 found these unique or distinctive types of families that only had one affected child.
00:10:30;27 And as you saw, by 2012, we only had about 240 of those families to be able to study.
00:10:36;07 By 2014, we had more than 2,500 families to be able to study.
00:10:41;26 And again, I used trios and that should say quartets.
00:10:47;24 And... and what you're looking at in each one of these... now we have a little bit different
00:10:52;16 color scheme, but again the red is affected, the green is the unaffected sibling.
00:10:57;28 And the take-home here is that if you look at mutations that have a significant impact
00:11:03;10 -- at least putatively on protein function, we're interpreting stop codons, etc --
00:11:10;15 that we find that there's a statistically significant increase.
00:11:13;08 Now, LGD down here is a term that was coined at about this time by a number of authors,
00:11:19;04 meaning likely gene disrupting mutations, and that's taking missense mutations
00:11:24;10 that are thought, based on a variety of metrics, to be functional and then pooling them with
00:11:29;27 mutations that are nonsense mutations.
00:11:32;27 So, these LGDs -- likely gene disruptive, or likely gene damaging mutations --
00:11:39;19 are more common in kids with autism than without.
00:11:41;20 So, that's an interesting finding.
00:11:43;20 So, what we can say is that, actually, in relatively small samples, only 200 or so kids,
00:11:49;17 we can see the effect.
00:11:50;19 We can then replicate the effect in a much larger sample.
00:11:53;26 The effect size is about the same, and quite significant for a contemporary genetic study.
00:11:59;27 If that's as far as we could go, though, it would be disappointing.
00:12:02;17 Because we don't just want to be able to say, there are more of this type of mutation
00:12:07;22 in kids with autism than those without.
00:12:09;10 What we want to do is to be able to leverage that information to identify specific genes
00:12:14;14 -- that's why we're sequencing the coding portion of the genome.
00:12:17;10 And I'm gonna jump forward to 2015 and... and get to sort of the punchline, which is
00:12:22;20 a highly, highly successful way of identifying, reliably and reproducibly, genes for autism.
00:12:29;01 Now, I've told you the story, and it may sound like this was my laboratory out in the wilderness
00:12:33;16 doing this, but in fact that wasn't the case at all.
00:12:35;25 There were multiple laboratories that were pursuing a similar strategy.
00:12:39;19 The first paper that I showed you from Nature in 2012 actually published with two...
00:12:45;26 two papers from other labs in the same issue of Nature.
00:12:49;16 The second one that I just showed you from 2014 was a consolidation of the laboratories
00:12:56;04 that were involved in those initial studies and then... and then, in fact, two large consortia
00:13:01;20 that sprang up.
00:13:02;24 So, there was an enormous amount of work going on not just in our laboratory but across
00:13:07;00 the country, really around the world, using similar strategies.
00:13:09;27 And... and the benefit of that is what we saw, which was tremendous convergence.
00:13:14;01 People got to the finish line at essentially the same time, and we were able to compare results
00:13:17;14 and see that unlike the early days of psychiatric genetics, where one lab
00:13:22;24 would find a gene and then two weeks later there would be a contradictory result in the genome,
00:13:27;22 we began to see tremendous convergence where, regardless of whether you were
00:13:31;19 in Boston or in San Francisco or in Washington state,
00:13:34;18 that we were getting the same answer.
00:13:36;27 By 2015, Stephan Sanders -- who was at the time in my lab, now he's an assistant professor
00:13:43;22 at UCSF -- was lead author on a paper where we took all of the available data from these studies,
00:13:49;21 put it all together, and then provided... sort of consolidated information on
00:13:55;08 what we knew by 2015 about the genes that were contributing to autism spectrum disorder.
00:14:01;12 And that's the table listed here.
00:14:02;21 So, there are a couple of important things.
00:14:04;20 One is just to say as a Gestalt... you know, I told you I worked on this since 1997,
00:14:09;28 and the first 10 to 15 years really... you know, my sole goal in life was to be able to
00:14:16;21 give a lecture and identify one gene with autism that I had confidence about.
00:14:20;18 So the idea of, now, that we've gone from sort of wandering in the desert, as late as 2010,
00:14:26;19 with regard to specific genes contributing to autism... a handful of findings,
00:14:32;05 but very, very hard-won successes... now, to where not only do labs have a tremendous amount of agreement,
00:14:37;23 but I'm showing you a table that has 71 reliable, reproducible risk loci or genes.
00:14:48;13 And what I'm showing you here... the table is broken down to show the confidence
00:14:52;13 that we have, the statistical bins, if you will, that we've laid out, so that if you want to
00:14:57;15 be very certain that you're looking at an autism gene you can rely on a false discovery rate
00:15:02;11 that's quite low, less than 0.01, but if you're simply doing bioinformatics kinds of experiments,
00:15:07;06 not expending a lot of money to make a model system like a mouse or whatever,
00:15:11;25 then we can also begin to show you other genes that don't quite cross that threshold
00:15:15;22 but have significant evidence.
00:15:17;13 And for those of you who are very interested or have your favorite gene in this paper published
00:15:21;16 in Neuron, we included at the time -- and it's still a pretty good resource --
00:15:26;16 a table of every gene in the genome providing the statistical, the cumulative statistical evidence,
00:15:31;18 of its association with ASD, with autism spectrum disorder.
00:15:35;05 So, you can look and see where your favorite gene is, at least at that period of time.
00:15:39;16 And these are now just being updated.
00:15:42;03 If you have a real interest, you're welcome to contact the lab if there's
00:15:45;08 a particular gene you'd like to know where it stands now.
00:15:47;24 Now... so as I said, this is... you know, for someone who's been in this game for
00:15:51;13 a long time, remarkable to be able to tell you that we have, you know, all of these reliable
00:15:55;14 autism genes.
00:15:56;20 And in fact, one thing I also want to sort of point out is that... that our hope
00:16:01;08 that they would have large effect sizes, that did bear out.
00:16:03;19 So, that group that I showed you that have the highest level of confidence, of FDRs of 0.01,
00:16:08;10 cumulatively they have 20-fold increases in risk in the affected individual
00:16:14;07 versus the unaffected individual.
00:16:15;16 If you read genetic studies, most genetic studies that you will read now are of
00:16:19;06 common transmitted variation, where if you have a 20% increase in risk that's seen as kind of
00:16:24;24 a large... you know, that's about as big as it gets for complex disease.
00:16:29;09 So, for a common disorder like autism, to be able to find reliable risk genes that have
00:16:35;21 20-, 30-, 40-fold increases in risk, it's exactly what we were looking for.
00:16:39;11 Now, the downside is they're quite rare.
00:16:41;11 That limits some of the studies that we can do.
00:16:43;08 But the initial strategy that I laid out for you, where we were gonna try to leverage
00:16:47;26 very large effect sizes, it looked like by 2015 that we clearly had an avenue to do so.
00:16:54;17 So, when you look through that gene list, it's hard not to just say, wow, what do
00:16:58;27 those genes do?
00:17:00;15 And interestingly, you could, even without sort of a lot of fancy kinds of statistical analysis,
00:17:06;06 begin to see that it didn't look like these were a random assortment of genes, right.
00:17:12;24 So, we can anticipate that there are going to be hundreds, potentially maybe a thousand genes
00:17:16;24 that are contributing to autism, which is a remarkable statement.
00:17:19;24 And you could imagine that they might be so distributed in terms of function that
00:17:23;18 you would need 100 or 200 different treatments, and... and that would be unfortunate.
00:17:28;18 But right away, we could see that that would... that was unlikely to be the case,
00:17:33;24 because you could already tell that these genes began to kind of fall into known biological functions
00:17:39;17 or pathways.
00:17:40;17 It was suggesting that these are consolidating around a much smaller number...
00:17:46;05 we don't know exactly the number, but a much smaller number of sort of biological processes.
00:17:50;03 And there were not 1000 different, truly distinct paths.
00:17:53;28 And here what I'm showing you are kind of three that popped out from these high-confidence genes.
00:18:00;03 The one that is kind of most prominent, that you just kind of fall off a log when you're
00:18:05;01 looking at the gene list, particularly for those high-confidence [genes], about half of them
00:18:08;06 will fall into genes that encode proteins that are involved in synaptic signaling.
00:18:14;26 And the other half fall into genes that are involved in regulating gene expression,
00:18:19;22 so are involved in chromatin modification, and here is a diagram showing sort of the cumulative list of
00:18:26;17 genes by about 2014, and now we've added to this.
00:18:32;11 Now, at one level, again, this is remarkable to be able to say, you've got...
00:18:36;13 I've got this many genes for autism, and I have a good idea that they're consolidating
00:18:40;00 around these interesting pathways.
00:18:41;20 Sort of a glass-half-empty view of this might be that you've spent all of this time and energy,
00:18:47;01 and what you're telling me is that genes for a neurodevelopmental disorder are
00:18:51;04 involving neural and development.
00:18:54;09 We've got synapse genes and we've got chromatin modification genes.
00:18:58;08 I think that... you know, at some level that... that reflects a challenge that we still have
00:19:04;12 in the field, which is I can give you a broad category but unless I can tell you something
00:19:09;21 more specific than that I don't think that's going to take you very far.
00:19:13;18 But I think it's also important to point out that even though they fall into those broad categories,
00:19:17;17 and I can take you back just one slide and say that... you know, these are...
00:19:23;24 it's... it's not just a gemisch.
00:19:25;09 You're looking at specific pathways.
00:19:27;21 And that may give us some traction on some molecular specificity to autism risk.
00:19:34;10 And the same is true here.
00:19:35;11 It's not all of chromatin modification.
00:19:36;27 There are certain genes that are carrying risk, and so far we've seen that there are
00:19:40;09 others that are not.
00:19:41;12 So, there is, I think, some pathway here to begin to leverage directly these the...
00:19:47;20 the known functions of these genes, to begin to take the step after gene discovery to understand biology.
00:19:52;20 And I have to say that I was so preoccupied for so long simply with being able to
00:19:57;15 find genes that by the time we got to that finish line we hadn't really thought that deeply
00:20:03;20 about what was gonna happen after that.
00:20:05;21 And I showed this slide for about 15 years, with the kind of... now, in retrospect,
00:20:12;26 quite simplistic... simplistic idea that what would happen is we would find a rare mutation carrying large effects
00:20:17;06 -- and we've been able to do that -- that we would study that in vitro and in vivo,
00:20:21;01 in model systems and in vitro, that we would get a handle on what pathways
00:20:27;20 were involved -- what molecular pathways were involved -- and then that should be able to
00:20:31;16 lead us to hypotheses about novel treatments.
00:20:34;12 And... and this seems like a good idea, but when you actually have the genes in hand
00:20:39;22 you begin to realize pretty quickly that this is overly simplistic and there's some significant challenges.
00:20:45;14 This was crystallized for me, actually, right after we discovered... we identified that
00:20:51;09 we would be able to be successful in this way, we found the first gene that we knew
00:20:56;08 was going to be reliable and reproducible, and we knew we had a path forward.
00:20:59;11 It was a gene, a sodium channel, SCN2A.
00:21:02;00 And... and literally ran down the hall -- this is when I was an investigator at Yale --
00:21:07;18 because there was a laboratory down the hall, Steve Waxman's lab, that knew everything about
00:21:13;26 sodium channels, and we figured we have the gene and then this is the laboratory that's gonna, like,
00:21:17;23 get us all the way around this circle to pathway very quickly.
00:21:22;19 And so, you know, very excitedly, we went into the office.
00:21:25;17 I was with a postdoc.
00:21:27;15 And... and I was, you know, sort of taken aback a little, because he didn't seem
00:21:32;04 all that excited.
00:21:33;04 And so when I pushed him on, you know, why he wasn't terribly excited about the fact
00:21:36;14 that we were giving him this, you know, kind of first autism gene and a sodium channel,
00:21:40;09 he said, well, you know, that's important information but before, you know, we really
00:21:45;10 can have a successful collaboration I need to know some things.
00:21:48;17 I need to know when in development this is having its effect, and I need to know
00:21:54;10 where in the brain is having its effect.
00:21:56;23 And the reason I need to know that is because sodium channels do a lot of different things,
00:22:01;01 and they do different things at different times in different places in the brain,
00:22:04;15 and so if we just kind of randomly look at this we'll get some sort of phenotype.
00:22:08;21 I'll be able to show you something, an electrophysiological signature, or it'll change migration.
00:22:14;08 But I'm not...
00:22:15;08 I don't know whether that has anything to do with autism.
00:22:17;01 And I... and I had two kind of reactions, simultaneously.
00:22:20;18 One was like, how in the world am I gonna tell you... we know so little about autism.
00:22:24;21 I can't tell you when and where.
00:22:26;06 But at the same time, I realized that he was telling me some really important about
00:22:30;15 the challenges of moving to genes to do biology that I really hadn't thought about.
00:22:35;05 So, the diagram that I showed you actually is sort of borrowed from my mentor and collaborator,
00:22:41;06 Rick Lifton, who had done similar studies to understand blood pressure and to work in
00:22:46;09 the kidney.
00:22:47;12 And there, actually, it really was almost enough simply to be able to identify the pathway
00:22:53;04 and then to be able to move from that to identify targets.
00:22:55;18 But, you know... it's embarrassing to say as a child psychiatrist that, in fact,
00:22:59;28 all that Steve Waxman was telling me is that the brain is a lot harder than a kidney.
00:23:04;11 And the developing brain is a lot harder than just the brain.
00:23:07;08 So, the issue is that brain development is highly dynamic, that we really don't know
00:23:13;23 a ton, honestly, about how the brain is built.
00:23:16;14 We don't have a full census of all the cells that go into the brain.
00:23:19;11 We know something about the kind of topographical arrangement in the brain, but it's...
00:23:26;18 it is pretty primitive.
00:23:28;22 And so what we've got is sort of a moving target about which we don't know a tremendous amount,
00:23:33;08 this organ.
00:23:34;08 And at the same time, we're identifying genes that, like SCN2A, as Steve was pointing out,
00:23:39;13 have multiple different effects -- they're highly pleiotropic.
00:23:42;18 So now you've got a problem, which is that you may be able to identify a pathway,
00:23:46;13 you may be able to see a phenotype, but there's no guarantee that the phenotype that you're seeing
00:23:51;00 is really relevant to the pathology of autism spectrum disorders.
00:23:54;28 So, in a... shortly after this conversation with Dr. Waxman, my lab took a turn.
00:24:03;05 And the turn that we took was to try to answer what he saw as a simple question, what I originally
00:24:07;19 thought was gonna be impossible to answer, which is, when and where are these genes having
00:24:12;12 an effect?
00:24:13;12 Is there some point of convergence in the risk for autism that would allow us to come back
00:24:17;28 and lay out a series of experiments that wouldn't tell you on... not... it would tell you more than
00:24:24;04 what the function of the gene was, to begin to tell you when and where it
00:24:29;10 was having its impact on disease pathology.
00:24:32;01 And so... so then the question is how we were gonna do that.
00:24:36;15 And again the... you know, I was extremely fortunate to have a brilliant collaborator
00:24:40;19 down the hall, Nenad Sestan, who remains at Yale.
00:24:43;22 Now, Nenad was interested in building a map of the molecular landscape of the developing
00:24:49;20 human brain.
00:24:50;20 And surprisingly... at least, surprisingly to me as a human geneticist, that really had
00:24:55;06 not happened, again in part... not too dissimilar from autism... from genetics, is that we
00:25:01;20 didn't really have the tools in hand.
00:25:03;23 But by about 2010, he could begin to ask the question, if we look across all the genes
00:25:08;27 that are present in the human organism, can we begin to map out their trajectory of
00:25:14;01 when genes are turned on and turned off in brain development?
00:25:16;22 And so he started to lay out this map of the human brain transcriptome, the molecular landscape
00:25:22;26 of a developing human brain.
00:25:24;05 And he did it by taking 15 periods of development.
00:25:27;06 They're kind of arbitrarily divided here, but they start at four weeks post-conception
00:25:33;04 and they go all the way to over sixty years of age.
00:25:36;00 And he took postmortem tissue from brains at each one of these stages, and then
00:25:45;02 he dissected the brains into 16 identifiable, established anatomical regions.
00:25:52;21 And then he did expression analyses.
00:25:53;24 And he began to just build, as I said, this map, across development, of gene expression.
00:25:59;27 And we had several long conversations about how we might be able to leverage that to answer
00:26:03;22 this question of when and where.
00:26:05;13 And the first idea that we had was to take the genes about which we had highest confidence.
00:26:10;17 So, these were the first ten genes that came out that exceeded our most, kind of,
00:26:18;03 stringent statistical threshold and were already obviously being identified in multiple different laboratories.
00:26:27;00 So, we took those genes and we put them together with these maps of gene expression,
00:26:31;27 and we built co-expression networks.
00:26:33;16 So, we used the genes to seed these... the networks that are shown here.
00:26:41;24 And so, for each one of the periods and for each one of the brain regions,
00:26:47;25 we built a co-expression network.
00:26:49;26 Now, that will tell you something about how the biology of these different autism genes
00:26:54;11 may be connected, but, again, we wanted to understand something about autism risk.
00:26:58;18 So, we set out to try to determine, how would we select among these co-expression networks
00:27:04;18 ones that might be relevant to autism?
00:27:06;24 And what we did here is we took the next set of genes -- so these are an additional,
00:27:11;19 larger set of genes that don't quite meet the threshold for high confidence but have evidence that
00:27:16;01 they're related to autism -- and we hypothesized that if a network was relevant for autism --
00:27:22;27 one of these co-expression networks -- at a particular time in a particular place in
00:27:26;24 human brain development that we would begin to see these additional autism risk genes
00:27:31;27 show up in those networks.
00:27:34;13 And so we... we, you know, sort of developed a statistical test to be able to see
00:27:39;02 whether or not we had enrichment of these autism risk genes within any of the networks to a degree
00:27:44;22 that was greater than expected by chance.
00:27:47;01 And this top panel shows that, in fact, we took brain regions in here... these are
00:27:52;24 16 brain regions that are collapsed in an unsupervised hierarchical clustering of expression data
00:27:59;11 into just kind of four anatomical areas of the brain.
00:28:03;28 But what you can see... let's see if I can get there... is that there was really the...
00:28:08;24 the deeper the red, the more significant the finding.
00:28:11;24 We found two periods right next to each other, one that was very highly statistically significant.
00:28:17;24 It was enriched for additional autism risk genes.
00:28:21;04 Built on the first ten genes that we identified, but then enriched for this next set of
00:28:26;10 risk genes.
00:28:27;10 And that was in prefrontal cortex and primary motor cortex.
00:28:34;03 And so, here is just a drawing.
00:28:36;23 I've now... you know, we sorted through more than 50 spatiotemporally defined
00:28:42;02 co-expression networks.
00:28:44;15 And this is the most enriched one that corresponds to this finding, here.
00:28:49;04 And again, without taking you through the detailed methodology, just... this is a representation
00:28:55;05 from a review that we did in 2014... is that once we knew that this network was the
00:29:00;27 one that we were interested in, we could begin to ask questions about... about additional characteristics.
00:29:07;15 So, we knew already that this network was enriched, and in this period 3-5, which is
00:29:12;08 mid-fetal cortex or mid mid-fetal brain, and then we're looking at prefrontal cortex,
00:29:17;08 and that's represented down here.
00:29:20;25 But then we could go further and ask additional questions about that network.
00:29:24;02 So, we asked, if you looked at cortex, was that network present in all six layers of cortex?
00:29:30;11 And in fact it was not.
00:29:31;11 It was present, really, only in this enriched network, was present in deep layer cortex,
00:29:36;26 layers 5 and 6.
00:29:39;10 And... we also asked the question whether or not it was in all cell types in the cortex.
00:29:43;05 And we came up with the answer that it wasn't.
00:29:44;25 It was only in glutamatergic neurons in layers 5 and 6.
00:29:48;23 So, when you think about this, it now was really more than anything a proof of principle
00:29:53;03 kind of experiment.
00:29:54;08 But it's taken us from what you might think of as a static data set, which is the...
00:29:59;24 these individual mutations that are contributing to autism risk, it's overlaid functional data
00:30:04;06 about the developing human brain, and that's allowed us now to identify or at least hypothesize
00:30:10;02 that a particular point of convergent biological vulnerability for autism for the genes
00:30:17;10 that we're looking at are mid-fetal human brain deep-layer cortex and glutamatergic projection neurons.
00:30:23;12 And this diagram over here is just meant to sort of point out the slight kind of twist
00:30:28;25 on how we were thinking about understanding pathology.
00:30:31;14 So, as opposed to saying... you know, what exactly is going wrong?, we're asking,
00:30:37;23 when and where is it going wrong?
00:30:38;24 And this is pointing out that the genes that we're looking at are in different cell compartments,
00:30:44;10 they're involved in a variety of different functions from chromatin modification to synaptic functioning, etc.
00:30:49;13 What's bringing them together is not a pathway, not a protein-protein interaction.
00:30:53;24 What's bringing them together is the spatiotemporal distribution of risk.
00:30:57;20 So, I just want to summarize here.
00:31:01;08 We're certainly not at the end of the story, but I hope that I've been able to show you
00:31:04;27 that exome sequencing has transformed gene discovery, particularly leveraging these interesting
00:31:12;22 characteristics of de novo mutation.
00:31:15;05 These are large effect mutations.
00:31:16;26 They are present in the coding portion of the genome.
00:31:19;06 And consequently give us a fairly direct pathway, now, to understand them at the bench.
00:31:23;23 A challenge that we face, continue to face, is how we can then begin to constrain
00:31:31;12 the next set of experiments in a way that's gonna tell us not only something about the biology
00:31:35;23 but that's gonna really begin to hone in on pathophysiology.
00:31:39;16 Again, when you see a phenotype, when you manipulate something in a model system
00:31:44;01 for an autism gene, if it's pleiotropic there is a difference, potentially, between the
00:31:48;02 biology that you might describe and the pathophysiology that we want to understand.
00:31:53;01 Already I've shown you that there is convergence, at least for a subset of autism genes.
00:31:58;05 This is unlikely to be the answer for all autism genes.
00:32:00;25 In fact, I would bet against it.
00:32:02;11 But there is a convergence among our data set, but then again from other laboratories
00:32:08;10 in mid-fetal development.
00:32:10;04 And the last point that I'll make is not a political one but important nonetheless,
00:32:14;05 which is, in fact, it's that genetic heterogeneity of autism, I think now, that really gives us
00:32:19;24 the tremendous opportunity to be successful in identifying novel treatments.
00:32:24;28 If you could imagine that after all this work we ended up with one gene
00:32:29;00 -- say, like the Huntington's gene --
00:32:31;05 and that would be the thing that we would absolutely have to understand
00:32:34;12 and would have to turn out to give us an avenue to treat, in fact in autism we have
00:32:40;15 quite a different story, where we have many different potential targets that are coming up simultaneously.
00:32:46;08 They are fortunately consolidating around a smaller number, a manageable number,
00:32:51;18 of biological processes, and what look like a very manageable number of developmental time points
00:32:58;26 and anatomical distributions, that we really might be able to leverage this heterogeneity
00:33:05;12 in a way that will lead to the ultimate success in this research strategy.
00:33:11;20 This work was done by... by wonderful folks in the lab.
00:33:16;03 I particularly want to highlight the work of Stephan Sanders and Jeremy Willsey.
00:33:20;06 Stephan's name and Jeremy's you saw throughout the presentation.
00:33:23;17 They're both assistant professors now at UCSF and really did the lion's share of the work,
00:33:29;23 from developing the analytic tools to writing the papers.
00:33:34;11 And this involved a collaboration with multiple cloud collaborators, both at Yale, at UCSF,
00:33:39;17 and around the country, including other labs that were seminal in identifying
00:33:46;23 the role of de novo mutations to autism spectrum disorders.
00:33:49;20 But most importantly, none of this work would be possible, either without the support of
00:33:55;13 our funders, but more importantly the participation of the families that contributed to the
00:34:00;04 Simons Simplex Collection and the other autism cohorts that... that I've talked about in this presentation.
00:34:08;26 So, thank you for your attention.

Videos in this Talk

Part 1: Developing Autism Spectrum Disorder Therapeutics: Treatment Targets and Tools from 30,000 Feet
Audience:
- General Public
- Student
- Researcher
- Educators of H. School / Intro Undergrad
- Educators of Adv. Undergrad / Grad
Part 2: From Genes to Pathology in Autism Spectrum Disorders
Audience:
- Student
- Researcher
- Educators of H. School / Intro Undergrad
- Educators of Adv. Undergrad / Grad

Speakers: Bryan King, Matthew State

Total Duration: 1:09:13

Recorded: August 2017

All Talks in Human Disease

Talk Overview

In the first lecture, Dr. Bryan King introduces the autism spectrum disorder (ASD) and defines the clinical criteria that characterizes ASD. Although there is no universal drug that is used to treat ASD, there are multiple medications used in the setting of autism. King outlines the different drugs used to treat the core behavioral features of autism as well as psychiatric disorders associated with autism (e.g. ADHD and anxiety), and discusses the challenges and problems with designing clinical trials to study drugs to treat ASD, specifically with the heterogeneity of the ASD population.

In the second lecture, Dr. Matthew State overviews the hunt for genes associated with autism and explains how studying de-novo rare mutations in the germline has advanced the understanding of the genetics of autism spectrum disorder (ASD). Studying a cohort of families with one affected individual and one unaffected sibling, they were able to map multiple genes involved in ASD. Because of the complexity of the developing brain, the same gene could act upon different regions of the brain at different times during development. Therefore, after identifying genes associated with ASD, they characterized the region and time during development where these genes are expressed and more likely to have an effect that is associated with ASD.

Speaker Bio

Bryan King

Dr. Bryan King is a professor of Child Psychiatry and Vice-Chair of Psychiatry at the University of California, San Francisco. He obtained his B.S. degree from University of California, Irvine (1979) and received his medical degree from the Medical College of Wisconsin (1983). He continued his medical training with an internship in internal medicine (1984),… Continue Reading

Matthew State

Dr. Matthew State is the Oberndorf Family Distinguished Professor and Chair of Psychiatry at the University of California, San Francisco and President of the Langley Porter Psychiatric Hospital and Clinics. He obtained his bachelor’s degree from Stanford University (1984) and received his medical degree from Stanford Medical School (1991). He continued his medical training at… Continue Reading

Part 1: Developing Autism Spectrum Disorder Therapeutics: Treatment Targets and Tools from 30,000 Feet

Part 2: From Genes to Pathology in Autism Spectrum Disorders

Talk Overview

Speaker Bio

Bryan King

Matthew State

Leave a Reply Cancel reply

Find a Video

Topics

For Educators

Educators Recommend

Techniques

Free Courses

Career Development

Our Favorites

Connect with Us

Talk Overview

Speaker Bio

Bryan King

Matthew State

Playlist: Bench to Bedside

Related Resources

Reader Interactions

Leave a Reply Cancel reply

Footer

Find a Video

Topics

For Educators

Educators Recommend

Techniques

Free Courses

Career Development

Our Favorites

Connect with Us