CARTA: The Costs of Big Brains with Alex Pollen

0:25 Thank you for including me in this meeting.

0:28 I've always been inspired by CARTA as an organization and the people

0:32 and ideas you bring together around one of the fundamental questions:

0:35 how we became human.

0:37 We often discuss the benefits of large brains.

0:40 But today I want to highlight some of the costs and share

0:43 with you how we're using organoid models

0:45 to explore trade-offs in human brain evolution.

0:48 Dramatic changes in brain structure and function

0:52 evolved along the lineage leading to humans.

0:54 Our brains tripled in size in just the last few million years.

0:58 In my training, I focused on genetic, molecular,

1:02 and cellular mechanisms that could drive

1:05 cortical expansion in the human lineage.

1:07 However, one of the special properties of a million brain evolution

1:11 is that not all brain regions change equally as the brain expands.

1:16 The neocortex, underlying white matter, striatum,

1:20 and cerebellum have increased disproportionately compared

1:23 with the olfactory bulb and brain stem regions, including the ventral midbrain.

1:28 In my own group, we've become interested in how this process of unequal scaling,

1:33 combined with new functional requirements,

1:35 may drive cell-type-specific adaptations to our large brain cellular ecology.

1:42 We think some of these trade-offs between forebrain expansion

1:46 and connectivity may be most strongly manifest in the ventral midbrain.

1:50 Here, a small caudra of a few hundred thousand hardworking

1:54 dopamine-producing neurons project to vast

1:57 target regions in striatum and cortex.

2:00 Since our divergence from old world monkeys,

2:03 these target regions have increased in volume six to 17 fold,

2:07 but the number of source neurons has only doubled.

2:10 In addition, to these increased demands from unequal scaling,

2:14 the innervation density of these neurons has actually increased in humans,

2:18 even compared with chimpanzee in the medial caudate and nucleus eccumbents.

2:22 The dopamine-directed striatum hypothesis predicts

2:25 that this increased connectivity was

2:27 actually crucial for the evolution of cooperative behaviors in our lineage.

2:32 The neurons themselves also have extraordinary specializations.

2:37 Their exonal arbors spin nearly a meter, even in rodents,

2:41 creating enormous energetic demands

2:43 that increase oxidative stress and vulnerabilities,

2:47 particularly in nigrostriatal neurons

2:49 that undergo energetically expensive pacemaker activity,

2:52 and the production of dopamine itself further releases reactive oxygen species,

2:57 making these neurons particularly vulnerable

2:59 as connectivity demands increase with brain expansion.

3:03 As such, it may not be surprising that these neurons

3:07 are implicated in a number of disorders enriched in humans,

3:10 compared with other primates, most notably,

3:13 including Parkinson's disease caused by the loss of the nigral population.

3:18 Although these neurons represent

3:20 a phylogenetically ancient vertebrate population,

3:23 we reasoned that new demands in the human

3:26 lineage may have spurred recent adaptations,

3:29 and we sought to investigate how dopaminergic neurons have evolved

3:33 in response to brain expansion and new functional requirements and specifically,

3:38 we were interested in identifying molecular factors

3:42 that could inform the developmental origin of connectivity differences,

3:46 as well as exploring whether compensatory adaptations

3:49 evolved in response to the increased demands

3:52 these neurons face in the large brain cellular

3:56 ecology and how do we tackle this question,

3:59 when experimental access to human innate brains is limited for ethical reasons?

4:05 We were fortunate to have a talented postdoc, Sara Nolbrant, joined the lab.

4:09 Sara extended pluripotent stem cell models

4:12 of dopaminergic neuron differentiation and maturation from human

4:16 to across the primate phylogeny in 2D

4:19 and further adapted these to 3D organoid models.

4:23 This allowed us to compare dopaminergic

4:25 neuron development and gene expression between species.

4:29 To isolate species differences and mitigate batch effects,

4:33 Sara scaled the experiment across many individuals

4:36 in a pooled culture format within each species,

4:40 often described as a village in a dish.

4:43 But Sara also went further and generated inner species

4:47 organoids with cells from every individual and species represented

4:51 in a common garden environment in order to isolate

4:54 cell-intrinsic molecular differences between species

4:57 and further reduce batch effects.

4:59 Here's an example of what one of Sara's experiments looked like.

5:02 Here we've pulled together IPS lines from 17 individuals across four species,

5:09 exposed these to dual smatinhibition,

5:11 to generate neectoderm, activated wind signaling to form midbrain,

5:15 and applied tonic hedgehog to ventrize these.

5:18 By day 16, these cells form ventral midbrain patterns progenitors

5:23 that can be cryopreserved and reproducibly thawed to form midbrain neurospheres.

5:28 These neurospheres produce dopaminergic neurons

5:32 with axons expressing tyrosine hydroxylase in green,

5:36 the rate-limiting enzyme for dopamine production.

5:39 These axons wrap around the periphery and can

5:42 be matured in long-term culture over six months.

5:46 In addition, when we fuse these interspecies

5:49 ventral brain organoids to target regions,

5:52 we can see that these TH-positive axons

5:54 can actually project to the target regions.

5:57 As the organoids mature, we see spontaneous activity,

6:01 and we see the release of dopamine,

6:04 showing that we can realize functional maturation in the model system.

6:08 Next, we wanted to measure gene regulatory divergence

6:12 in these common garden environments of primate dopaminergic neurons.

6:16 To do that, we have cells from each species and individual mixed together,

6:20 and we needed to find a way

6:22 to assign these cells back to their individual species.

6:25 Nathan Schaefer in the lab developed a computational

6:28 tool kit that uses naturally occurring sequence variants

6:32 in the transcript to genotype the cells and assign

6:35 them to individual while also removing ambient RNA.

6:38 We call this package CellBouncer for checking

6:41 cell IDs and keeping out the riffraff.

6:44 Using CellBouncer, on a dataset of gene expression from these organoids,

6:48 we could identify diverse cell types grouped

6:51 in this projection based on transcriptional similarity,

6:54 with each dot representing a single cell.

6:56 Notably, cells from each species were

6:59 maintained across the long-term cell culture,

7:01 and we were able to maintain representation.

7:04 About a third of the cells highlighted in blue

7:07 in the chart below from each species belong to the dopaminergic lineage,

7:12 denoted by the expression of LMX1A.

7:16 In addition, we saw conservation of many marker

7:19 genes with a few highlighted here in dopaminergic

7:22 neurons from each species and has referenced

7:25 to a primary cell atlas of human brain development.

7:28 Before going on to compare gene expression between the species,

7:32 we wanted to gauge whether the cells mature at the same or different rates.

7:36 To do that, we could reconstruct the sequence of transcriptional changes

7:40 during dopamine neuron maturation using

7:43 an approach called pseudotime trajectory inference.

7:46 And applying this approach, we could see the expected genes come up,

7:50 but we could also score cells from each species.

7:53 When we did this, what really stood out is

7:55 the cells from Rhesus Macaque mature much more quickly.

7:58 At day 40, some of the cells have already

8:00 reached the day 80 stage in the other ape species,

8:03 whereas we observed more comparable maturation rates in human and chimpanzee.

8:08 By an orthogonal measure, the switch from neurogenesis to gliogenesis,

8:12 we could also see a similar pattern

8:14 where Rhesus Macaque switched to gliogenesis much earlier,

8:17 supporting the molecular staging of maturation.

8:22 We could then apply these homologous cell types

8:26 and quantitatively defined maturation stages to ask how have

8:31 homologous cell types diverged in gene expression using

8:34 a linear mixed model to account for this variation.

8:37 When we do that, we can see that gene expression is

8:42 largely correlated between species and dopaminergic

8:44 neurons and other cell types.

8:46 But we can also identify gene expression

8:49 differences that evolved specifically in the human lineage,

8:52 using the other species, rhesus and orangutan, as outgroups.

8:56 When we do that, we can further validate some of these examples,

9:00 like this inward rectifying potassium channel,

9:02 KCNJ16, that shows very specific expression in human dopaenergic lineage cells,

9:08 but not in other species.

9:10 In addition to looking at gene expression,

9:12 we were also able to measure chromatin accessibility in the same cells,

9:15 and we can see a concordant increase in chromatin

9:18 accessibility at the promoter of this gene in human,

9:21 further supporting the gene expression difference.

9:24 Using this approach, we identified many candidate genes,

9:28 including those that could influence connectivity, such as Neuritin 1,

9:32 a neurotrophic factor that promotes axon outgrowth,

9:35 with a human-specific increase both

9:37 in chromatin accessibility and gene expression,

9:40 specifically in the human lineage.

9:42 Across divergent loci, we observed an enrichment for human-specific

9:48 structural variants overlapping differentially accessible regions.

9:53 For example, at this GABA receptor locus,

9:57 we observed a human-specific insertion overlapping a non-coding cis-regulatory

10:03 element and appearing to disrupt and reduce the chromatin accessibility.

10:09 This was concordant with reduced expression of the nearby GABA receptor gene.

10:13 Looking broadly among the upregulated genes revealed that most

10:18 strongly divergent categories in immature neurons related to connectivity,

10:23 including a range of neurofilament genes and genes linked

10:27 to axonal transport that were upregulated specifically in the human lineage.

10:32 When we looked at more mature

10:34 dopaminergic neurons in our culture system, remarkably,

10:38 we observed that the most strongly upregulated

10:41 categories actually related to oxidative stress buffering,

10:44 consistent with our hypothesis that compensatory mechanisms may have

10:47 evolved in this lineage and this included a gene

10:51 catalase that was recently shown to be upregulated in adult

10:55 human cortical neurons compared with chimpanzee and Rhesus Macaque,

10:59 supporting the vivo relevance of some of these in vitro models.

11:04 Although we can't age these organoids for decades,

11:07 we can expose them to oxidative stress acutely

11:11 and then directly examine whether human-specific regulatory networks,

11:15 such as those we just identified, influence the stress response.

11:20 Here, we can really leverage the phylogeny and Addish

11:23 approach by evaluating these cross-species differences in a controlled

11:27 common garden environment where cells from both human

11:30 and chimpanzee are represented together in the same state.

11:33 By adding rotinon, which blocks the electron transport chain,

11:38 we could recapitulate the selective vulnerability of these neurons,

11:42 the specific loss of TH in these pictures,

11:44 and then we could measure gene expression across individuals using CellBouncer.

11:49 When we do that, we see that the major

11:52 axis of variation relate to the stress response,

11:55 principal component 1, and to species differences, principal component 2,

12:00 with parallel response trajectories in human and chimpanzee.

12:04 These response trajectories include conserved bienergetic stress genes.

12:10 For example, heat chock proteins, early response genes,

12:14 and mitochondrial genes are upregulated in both human and chimpanzee,

12:18 while neuronal marker genes are downregulated.

12:22 In addition to these conserved responses,

12:25 we could also see that the oxidative stress stimulation

12:29 unmasked species-specific differences in the response to oxidative stress,

12:35 including BDNF, which was transiently upregulated in humans, but not chimpanzee,

12:40 and the mitochondrial calcium uniporter MCU,

12:44 which was transiently downregulated.

12:46 We were excited about this because both of these genes,

12:50 the activation of BDNF and the repression of MCU,

12:53 have been shown to promote dopaminergic

12:55 neuron survival in Parkinson's disease models,

12:58 consistent with evolved neuroprotective effects in the human lineage.

13:02 Given the baseline increase we observed in oxidative

13:06 stress buffering genes and these human-specific stress-dependent responses,

13:12 we next asked whether the trajectory

13:13 of oxidative stress response differed between species.

13:18 When we did this, we found that all seven chimpanzee individuals

13:22 had stronger responses along this molecular

13:25 trajectory than all eight human individuals,

13:28 suggesting that even in this simplified

13:31 in vitro model with young dopaminergic neurons,

13:34 the stress response may be blunted

13:37 in human cells by compensatory molecular changes.

13:41 To conclude, we generated interspecies

13:44 ventral midbrain organoids using a phylogeny

13:48 and a dish approach that produce dopamine

13:51 and resemble primary brain atlas dopaminergic neurons.

13:55 By comparing gene expression and chromatin

13:58 accessibility in a controlled environment,

14:01 we were able to identify human-specific changes

14:04 and candidate genetic mechanisms driving these alterations.

14:08 The top-ranked categories in immature

14:11 and maturing dopaminergic neurons related to changes

14:15 in connectivity by changes in neurofilament genes and exonal transport genes,

14:21 as well as changes in antioxidant activity and oxidative stress buffering,

14:26 including peroxyrdoxins and catalase.

14:29 Further, by applying acutely

14:32 in this experimental model oxidative stress stimuli,

14:35 we could unmask human-specific neuroprotective responses and see that the human

14:40 cells had a blunted overall response

14:42 compared with chimpanzee across many individuals.

14:45 Coming back to our early question of whether dopaminergic neurons have

14:50 evolved molecular changes in response to trade-offs in human brain evolution,

14:54 we see evidence that there is, in fact,

14:56 an increase in buffering of oxidative stress

14:59 response that may partially compensate for the dramatic

15:02 changes in connectivity that have evolved

15:04 in our larger brains and with new functional requirements.

15:08 Even though this is a simplified system,

15:10 it shows us how we can investigate cell type evolution and the response

15:15 to trade-offs in otherwise inaccessible human

15:18 and ape cell types using organoid models.

15:20 In the future, we'd be interested

15:23 to extend this approach beyond long-range connectivity

15:27 that I focused on here to many other potential costs of large brains.

15:32 For example, large brains take longer to develop,

15:36 and humans particularly have a protracted period of development,

15:39 including a human-specific stage of early childhood.

15:42 Where there's dramatic changes in neuronomorphology.

15:46 In addition, large brains create new challenges for vesicle trafficking,

15:51 bioenergetics, and the reorganization of neurons,

15:54 reducing average connectivity between them creates modularity

15:58 and computational constraints that could be investigated.

16:01 In closing, evolution involves trade-offs,

16:04 and we can combine organoid models and single-cell genomics to learn

16:08 from how cells and systems have

16:10 responded to these costs from molecular perspective.

16:12 I want to thank a number of people

16:15 in my group who really made this work possible.

16:17 Sara Nolbrant established this system.

16:20 Janelle Wallace led the Gene Expression Divergence analysis,

16:24 and Jingwen led regulatory network analysis I didn't have time to talk

16:28 about while Nathan Schaefer and Bryan

16:30 Pavlovich established the CellBouncer package,

16:33 and Brian reprogrammed many of the ape lines that we used in this study.

16:38 Thank you for your attention, and I'm so grateful to be part of this symposium.