BOSTON, MA and UNDISCLOSED LOCATION – APRIL 1 2026

Read more

By 2005, cloning was more routine, and efficiency had increased due to various optimizations, like histone deacetylase inhibitors to help address epigenetic issues. Dr. Wakayama, who had recently started his own lab at RIKEN, and his wife, Dr. Sayaka Wakayama, had an interesting idea: what if we cloned a mouse? And then took a cell from the cloned mouse and made another clone? And then did it again?

And again?

57 times?

Can’t stop, won’t stop cloning

Long-term experiments aren’t a new thing — in fact, they’ve been around for a while.

Some, like the E. coli long-term evolution experiment, have tracked genetic changes over time in a lineage of cells. But doing this in cloned mice is next-level difficult. With E. coli, all you need to do to make a new generation of cells is pipette a small drop into new growth medium. With cloning mice, the process involves biopsies, cell culture, egg retrieval, microinjection, and embryo transfer. There are only a few people in the world who can do this, and very few as good as those in the Wakayama lab.

And because DNA sequencing technology was still in its infancy when the project started, many of the tools that the project ended up using to analyze the mice were only developed partway through. This project, which started in 2005, persisted through lab moves, an earthquake in 2011, and the COVID-19 pandemic. Over 20 years of diligent effort, these researchers repeatedly cloned mice, creating a total of 58 generations. Due to the low efficiency of cloning, this took 30,947 individual attempts. On the 58th generation, the project ended. Not because the researchers got tired of it, but because the mouse cells just wouldn’t make clones.

So what happened?

In an interim report in 2013, the research group reported successful cloning for up to 25 generations with no decrease in the health of the mice or in cloning efficiency rates. Although the cloned mice had epigenetic abnormalities typical of clones (including enlarged placentas), these abnormalities did not worsen with time. Telomeres likewise did not shorten with increasing generations. However, in 2013 whole genome sequencing was much more expensive than it is today, and the researchers did not investigate the accumulation of genetic mutations over time.

It turns out that mutations were in fact accumulating. Not at a terribly high rate, but enough to matter. This brings us to the current paper.

In 2026, the Wakayama group reported their final results. Things appeared to go reasonably well up until around generation 40, but after that point, the efficiency started to drop, and by generation 58, the researchers simply couldn’t continue.

Although the mice that were born were healthy and lived to a normal mouse lifespan (Fig. 1D), the cloning efficiency by generation 58 was nearly zero (Fig. 1B).

With the benefit of advanced sequencing technology, the researchers found a steady accumulation of mutations over the generations. By generation 57, the cloned mice had acquired over 3400 single-base changes relative to the starting mouse, whereas 62 generations of natural reproduction (in an inbred mouse strain) had accumulated 752. In other words, cloning caused 3.1 times more single-nucleotide mutations per generation than natural reproduction.

On its own this increased mutation rate is not terrible. But unlike with cloning, harmful mutations that arise in sexually reproducing organisms can be lost through recombination. Cloning lacks this mechanism, so harmful mutations are more likely to accumulate.

On top of the single-nucleotide changes, there were also much larger issues with the genomes of the cloned mice. In particular, one entire X chromosome was lost at some point between generations 25 and 45, and never regained. Mice can tolerate having a single X chromosome reasonably well (much better than humans), but this is still not very healthy. There were also various other chromosomal deletions, inversions, and translocations. Between the structural abnormalities and the point mutations, it’s clear why the cloning process failed after 58 generations.

Conclusions

Although the media reports that this study showed “a big unexpected problem with cloning”, I don’t think this characterization is accurate. The accumulation of mutations was definitely a known issue (as far back as 1932) with asexual reproduction. Also, keep in mind that their first 25 generations were quite healthy, and even though the efficiency decreased in later generations, the mice that survived had normal lifespans.

A 3.1-fold increased mutation rate isn’t great, but I think if the researchers had been able to do whole genome sequencing for quality control at each generation (instead of only retrospectively), and only clone mice without severe genetic abnormalities, they likely could have avoided most of the negative effects they saw here. And the lack of epigenetic abnormalities accumulating over generations is also encouraging, both for cloning as well as for other technologies (like IVG) that create offspring from adult somatic cells.

Overall, I’d be worried if someone proposed to do human cloning for more than 25 generations. But that long term experiment would take a lot more than 20 years!

Subscribe now

Genetics of meiosis fidelity

Common variation in meiosis genes shapes human recombination and aneuploidy (Carioscia et al., Nature 2026). This study looked at data from an embryo testing company, Natera,1 which used SNP microarrays to determine embryo genotype and chromosome copy number. The researchers also obtained genotypes of the parents. This gave a sample size of 22,850 biological mothers and 139,416 embryos. Getting this data must have been a bureaucratic nightmare!

With the ability to examine genotypes of both parents and embryos, the researchers identified variants in meiosis-associated genes which impact the number of crossovers during meiosis and the risk of aneuploidy. Aneuploid embryos tended to have fewer crossovers, which makes sense given that a lower crossover rate would increase the risk of having zero crossovers (in which case the chromosomes wouldn’t stay together). Furthermore, women with high risk of having aneuploid embryos also had a shorter reproductive lifespan (later menarche and earlier menopause).

It would be quite interesting to see if these results replicate in Ovelle’s in vitro meiosis induction system.

Genetics of Epstein-Barr virus infection

Population-scale sequencing resolves determinants of persistent EBV DNA
I’ve been interested in Epstein-Barr virus (EBV) for a while due to its prevalence and ability to cause a wide variety of problems (autoimmune diseases, cancer, etc.)

For historical reasons,2 the EBV genome was included in the GRCh38 human genome reference assembly. Most researchers ignore this, but Nyeo et al. looked at EBV read counts in UK Biobank whole genome sequencing data, which comes from blood samples. They found that about 10% of individuals had a high burden of EBV DNA in their blood. Over 90% of the UK Biobank population is positive for EBV, so the people with a lot of EBV DNA represent those whose immune systems struggle to control the infection.

The genetics of persistent EBV DNA were strongly associated with the immune system. There were several significant genetic variants across the genome, but the most important were at the HLA locus (which will be no surprise to any immunologist). Understanding HLA interactions with EBV may help with vaccine development, and this paper makes a big contribution.

The researchers also looked at genetic variation within the EBV genome, but didn’t find any significant effects on pathogenesis (this was strong evidence against the results of several previous smaller studies).

Genetics of lifespan

Heritability of intrinsic human life span is about 50% when confounding factors are addressed

Human lifespan, like other complex traits, is heritable to at least some extent. Readers may recall that heritability of a trait is the proportion of total variance explained by genetic variance (as opposed to non-genetic variance).

Many previous studies found a rather low heritability for lifespan (about 20-25%). But it turns out that this is because they examined cohorts who were born from 1870-1920, when non-genetic variance of lifespan was much higher than it is today. This study (Shenhar et al. 2026) examined three cohorts of Scandinavian twins and one cohort of American centenarian-sibling pairs, and analyzed extrinsic (accident, homicide, infectious disease) vs intrinsic mortality.

The result is somewhat of a tautology: when ignoring many sources of environmental variance (extrinsic mortality), the heritability of lifespan increases to 50-55%. But this is still interesting, because extrinsic mortality is much lower today than it was 100 years ago, so the higher estimate of heritability is more relevant.

Hopefully extrinsic mortality will remain low across the world in years to come!

So, let’s discuss: what does an egg need to have in order to develop into a healthy baby after fertilization? How can we tell if an egg is good or not? And what does this mean for in vitro oogenesis?

Good genetics

At the most basic level, an egg must have the correct number of chromosomes. Chromosome pairing is established during meiosis I, and chromosomes are distributed into the embryo during meiosis II.2 Chromosome spreads3 are a simple yet effective method to check whether chromosomes are distributed correctly in individual cells. In metaphase of meiosis I, chromosomes should be present as tetrads: groups of four chromatids connected at their centromeres and at a crossover site. If chromosome pairing and recombination happened correctly, a spread at metaphase I will show exactly 23 tetrads per cell, with no chromosomes left unpaired.

This is a destructive test that can’t be used on embryos before implantation, but it’s a good way to do a quality control check to see if an in vitro oogenesis method is reliable enough to safely use.

Additionally, the chromosomes can’t have harmful mutations. Adult somatic cells often accumulate mutations throughout their genome, with this effect being worse in cells exposed to the environment (like skin)4 or cells which divide rapidly. Most of these mutations won’t matter, but some could be harmful if they disrupt important cellular functions. In order to check for mutations, whole genome sequencing can be performed on the starting stem cell lines. Starting with a low-mutation cell type, on average the best of five stem cell lines will have fewer mutations than natural reproduction.

Both the overall number of chromosomes, as well as any mutations, can be examined using pre-implantation genetic testing. A small number of cells are removed from the trophectoderm (the outer surface of the embryo) and sequenced. This is the same concept as what companies like Orchid Health use for embryo screening. Although this method is not perfectly reliable,5 combined with extensive sequencing of the starting cell lines, it can make the overall process safer than natural reproduction.

Good epigenetics

Epigenetic marks (chemical modifications to DNA or histones) are crucial for controlling gene expression throughout development. In particular, DNA methylation present in the egg can persist all the way into adult life.

So, in order to develop properly, an egg needs to have the correct epigenetics. This means completely erasing the epigenetic marks present in the starting cell, and also writing the egg-specific marks. This has been a challenge for the field: it’s likely that the low developmental potential of in vitro grown mouse eggs is largely due to epigenetic issues.

Examining epigenetics in natural eggs and embryos is challenging due to low numbers of cells, which makes it difficult to establish standards for what the epigenetics of in vitro grown eggs should look like. Some DNA and histone modification patterns have been characterized for human oocytes and embryos, but overall, more data are needed here. I am confident that recent advances in low-input epigenetic analysis will allow for improved datasets in the near future.

Good everything else

In order to develop properly, an egg needs to have grown to a large enough size, and have stored up the correct RNAs and proteins for embryonic development. It also needs to have enough mitochondria (several hundred thousand per egg). Egg growth is enabled by ovarian supporting cells, but for proper downstream development, the important thing is that the egg itself ends up looking like a natural egg, with the correct levels of RNA and protein expression.

Due to the large size of eggs, single-cell transcriptomics and proteomics is actually not too difficult with them. Fun fact: the first-ever scRNAseq paper came out of Azim Surani’s lab all the way back in 2009! This looked at mouse eggs and early embryos, and the ~100 million SOLiD sequencing reads they generated probably cost them several tens of thousands of pounds. Today, proteomic, and especially transcriptomic, quality control data sets are readily accessible for eggs and early embryos. Transcriptomic data on early embryos can also shed light on epigenetic processes like zygotic genome activation. Overall, in vitro grown eggs should be as similar to natural eggs as natural eggs from different donors are to each other.

Implications for in vitro oogenesis

Any useful method for growing eggs in vitro must ensure that the eggs are consistently good. I emphasize consistency here because it’s often the case that a differentiation protocol works for certain stem cell lines, but not for others.6

At this point I want to mention a recent paper from Shoukhrat Mitalipov’s lab (Gutierrez et al. 2025). In this paper, the researchers took eggs from human donors, removed the chromosomes, and inserted nuclei of skin cells. The eggs were then stimulated to divide, and randomly distributed their chromosomes such that on average the resulting embryos ended up with 23 of them. This approach was an extension of their work with mouse eggs which I wrote about previously.

Although the researchers provided an accurate assessment of the method’s limitations in the paper itself, popular news media reported it as “researchers create human eggs from skin cells”. This can hardly be true if the researchers used human eggs as starting material! Plus, the method has no way to control chromosome distribution.

And furthermore, the epigenetics will correspond to a skin cell, not an egg cell.7 I am actually a big fan of the Mitalipov lab’s work, since they have uncovered some very interesting biology, but I am disappointed in the media response.

More broadly, any in vitro oogenesis company needs to have rock-solid quality control. A proof-of-concept human egg would be a great achievement, but to actually benefit patients, the method must be both scalable and reliable.

Thoughts on speed

The one big advantage of the Mitalipov lab’s method is its speed: taking an adult cell and putting it into a pre-existing egg means that a patient only needs to wait a few days for the egg to develop into an embryo, rather than weeks to months for growing a new batch of eggs.

Quality still comes first: no patient wants to use a procedure that has a substantial risk of making a baby with developmental issues. But assuming quality control is solved, the method that will win in the market will be the method that’s fastest (and relatedly, cheapest). And there is not necessarily a tradeoff between speed and quality: in fact, faster methods will actually be easier to optimize for quality due to shorter experiment cycle times. Methods that try to perfectly re-create natural ovarian development will be stuck waiting many months (and potentially several years) for their eggs to grow, and in the end, only 30%-50% of natural eggs develop normally after fertilization, which places an upper limit on how good their eggs can be.

At Ovelle, we focus on using regulatory factors to directly drive developmental processes such as epigenetic erasure, meiosis, and ovarian follicle formation and growth. This lets us fast-forward through development, achieving in weeks (and in some cases, days) what takes others months. At the same time, we continue to optimize the quality of our cells in gene expression, chromosome pairing, and epigenetics. We believe that this is the best way to make in vitro oogenesis work for everyone who needs it.

1

No reproduction is ever completely safe, but our goal at Ovelle is to have our eggs be safer to use than natural eggs. This is actually not a very high bar, given that natural eggs are >30% aneuploid, with the rate increasing at older ages.

How can DNA learn?

As I discussed on this blog back in 2023, DNA base pairing can be used to perform computations. Back then, I was writing about relatively simple computational systems, for example ones that could add two 6-bit numbers. These were built of logic gates, like an AND gate that releases a piece of single stranded DNA only when two other “input” strands are present.

The current paper goes far and beyond this. The authors came up with a clever system of making DNA logic gates that don’t interfere with each other. They scaled this system up to 100 input bits (implemented in about 1200 unique pieces of DNA),2 then showed that it could classify handwritten numerals from the popular MNIST dataset when they were presented as a 10x10 bit “image”. Most impressively, this classification was not hard-coded; instead, their chemical system actually learned the classification based on exposure to inputs. And once “learned”, the memory of patterns was stable for at least several days (or months if the DNA was frozen).

The basic structure of the DNA logic gates is based on a “toehold switch” that uses a strand of DNA to displace another strand of DNA based on sequence hybridization.

The current paper chained hundreds of these switches together, building on previous work but also inventing two new kinds of DNA logic gates (called an “activatable amplification gate” and “activatable transformation gate”). These activatable gates require a threshold concentration of an “activator” DNA strand before turning on, thus implementing the kind of nonlinear effects required for neural network based classification.

I am not sure how many actual applications this will have, since there’s no easy way to integrate this kind of short single stranded DNA with living systems (which generally operate with different kinds of signaling). And for everything else, electronics are vastly superior. Notably, the classification accuracy of this system was relatively low, just 53 to 81 percent (depending on the digit). Worse, the system can’t be reset after use, which means it can only be used once. This also means it doesn’t allow backpropagation, so training is very limited. Still, this system is an interesting proof of concept for information processing using chemical binding events.

How did the authors learn from DNA?

The more interesting thing about this paper is the story behind it. Building such a complicated system of DNA logic gates was not easy! As any veteran Factorio player knows, what starts as a reasonably simple and logical system can easily become a mess of spaghetti as more and more features are added. In this case, the DNA strands might end up looking a bit like literal spaghetti!

The authors started out on this project with an initial logic gate design, but kept running into problems. Every time they fixed one, another appeared. By the third year of research, they were probably starting to go a bit crazy. The situation was described candidly in the Supplementary Information:

We learned an important lesson from extensive rounds of design revisions discussed above (Supplementary Notes 4.1 to 4.8): Identifying challenges one by one and coming up with solutions for each challenge may lead to limited success. For complex molecular systems, a solution for one problem could give rise to another problem somewhere else in the system. This phenomenon could further cascade, and in the worst case scenario, forming a deadlock in a cycle.

At this point, they took a step back and identified seven interrelated problems with their design, any one of which might be solved on its own, but at the cost of making the others worse. So in the third year of research, they realized they had to throw everything out and start over.

Moving to Design 2 was a major decision – it meant that we must throw out three years of work on Design 1 and start over from scratch. This decision eventually led to a successful demonstration of learning in DNA-based neural networks. Beyond the science presented in this paper, we cannot over emphasize the philosophy that all challenges must be considered as a whole and that the most extraordinary courage is needed at the most desperate time.

With the benefit of three years of experience, they came up with a new design with extra features3 to resolve these problems in the first design. Fortunately, this ended up working a lot better.

The authors summarized their lessons as follows:

Two important lessons that we learned for engineering complex molecular systems are as follows. First, a failure mode of the debugging strategy is to focus on individual challenges. A solution for one problem may give rise to another problem somewhere else in the system. With further cascading, in the worst-case scenario, this debugging strategy may form a deadlock in a cycle. After understanding the failure mode, we arrived at an alternative strategy where all challenges are considered as a whole and solutions are devised to address the entire body of challenges simultaneously (Supplementary Note 4.9). Second, a waste of energy may occur if there is no approach to differentiate fabrication problems from design problems. For example, we discovered severe and uneven sample evaporation in source plates for a liquid handler, resulting in wildly inaccurate concentrations that directly affect the computation of the molecular system. Instead of just relying on a better sample storage method, we established a systematic approach to regularly evaluate the sample quality and reorder new strands whenever needed (Supplementary Note 5.13).

Besides the lesson about getting stuck trying to optimize an inadequate design, the authors also mentioned the importance of distinguishing design issues from execution issues. Often, execution issues can mask design issues, so ensuring accurate execution is critical for evaluating different designs. On the other hand, a design that is perfect in theory may be extremely difficult to actually execute.

What can we learn from this story?

A common failure mode in research, especially biology, is to get stuck in a local optimum. I’ve personally experienced this several times myself: I have a protocol that works a little bit, and I spend a few months trying to make it better while not realizing that it has fundamental limitations that make it unsuitable for what I’m trying to do. I’m just glad I haven’t spent three years this way!

The main lesson from the paper is: learn to recognize when you’re stuck, and have the courage to backtrack and start again. This can be quite difficult when you’ve invested a lot of time and effort. Your research can seem like “just one more experiment” away from finally working, but if you’ve been at that spot for months, perhaps it’s time to try a completely new approach instead of a variation on your existing one. Consider what experiment could address the fundamental problems as a whole, rather than trying to play whack-a-mole with individual bugs.

This requires both humility to admit your current approach might be wrong, and courage to abandon what seems safe and try something new. Of course, you shouldn’t give up too early either! As my friend Devon wrote, knowing exactly when to give up is one of the most important traits of a good scientist. But if you have a suspicion things aren’t going well with your project, it’s worth taking the time to consider the problems as a whole, and ask whether your current approach is really the best one.

Now, researchers from Ethan Bier’s lab have demonstrated a different approach. Instead of a gene drive causing sterility, they engineered a drive which makes mosquitos resistant to infection with the malaria parasite. Although this drive currently doesn’t work as well as the ones that cause sterility, in the future it could be a useful approach to combat malaria.

How does this work?

After a mosquito eats blood infected with malaria parasites, the parasites move from the mosquito’s gut to its salivary gland, whence they can then spread to the mosquito’s next victim. However, a variant allele of the mosquito FREP1 gene can prevent the malaria parasite from escaping the mosquito’s gut. The researchers engineered a gene drive to spread the resistant FREP1 allele through a population of Anopheles stephensi mosquitoes.1

The drive consists of a guide RNA expression cassette inserted into an intron of FREP1, close to the variant site of the resistant allele. In order to track the spread of the drive, it also contains a fluorescent protein expression cassette (either GFP or RFP). When supplied with a separate source of Cas9, the guide RNA will cut the wild-type FREP1 allele, and the mosquito will use the gene drive FREP1 allele as a repair template.

The researchers showed that the resistant FREP1 allele doesn’t negatively impact mosquito reproduction, but greatly decreases the number of malaria parasites able to infect the mosquito salivary glands. When fed with infected blood, 80% of wild-type mosquitoes carried malaria, but only 30% of resistant mosquitoes did. Finally, the researchers showed that their gene drive could spread through a population of mosquitoes expressing Cas9 in their germline cells.

How might this be better than other gene drives?

Compared to the other gene drives I’ve written about (which cause female-specific sterility), this gene drive would spread faster through mosquito populations since both males and females could spread it. Furthermore, there would not be any selection pressure for drive resistance among mosquitoes. Finally, in theory it would address the "mosquitoes are important for ecosystems" objection. However, in practice, people who hate GMOs will probably just invent a new objection to this gene drive.

What are the next steps?

This is a promising demonstration, but if I were a peer reviewer for Nature,2 I would have asked them to show a few more things:

1. As currently designed, the drive needs a separate source of Cas9 in order to spread, so it wouldn’t work in the wild. Can the drive be re-engineered with Cas9 (or a smaller Cas gene) in place of the fluorescent protein marker? Alternatively, can a second, separate gene drive spread Cas9 in a neutral safe-harbor site?

2. Does the drive work in A. gambiae? This mosquito species is responsible for most malaria deaths worldwide, so it’s a bit puzzling why the researchers chose to use A. stephensi instead. Did they try it in A. gambiae and it didn’t work? Or perhaps because they already had a strain of A. stephensi expressing Cas9 from a previous paper, it was easier to use that.

3. Is the reduction in malaria parasite load enough to be clinically meaningful in preventing malaria infection? For example, if someone gets bitten by 10 mosquitoes per day,3 does it matter if 8 or 3 of them carry malaria if all it takes is one infected bite?

4. How easily could malaria parasites evolve resistance?

Overall, I still think the best approach is to eliminate mosquito populations using sterility-inducing gene drives. But if that doesn’t work, then this could be a good backup option.

Thanks to Richard Fuisz and Eryney Marrogi for inspiring me to write about this paper. If you’re interested in gene drives, you should also check out Eryney’s post on them.

Recently, Mitinori Saitou’s lab at Kyoto University published a paper (Nosaka et al., Developmental Cell 2025)1 describing a method to mass-produce mouse eggs from stem cells, at up to 400,000 eggs per batch. Although the eggs are low-quality and cannot reach a fully mature stage, this is still a remarkable advance. Surprisingly, this method works without ovarian supporting cells, and this has important implications for in vitro meiosis and in vitro oogenesis.

How does this work?

The basic approach is as follows:

1. Starting from mouse pluripotent stem cells, differentiate them to primordial germ cell-like cells (PGCLCs) through treatment with signaling factors.

2. Take the PGCLCs and culture them on m220-5 feeder cells, which promote cell proliferation and epigenetic erasure. The PGCLCs also start forming intercellular bridges, which are important for meiosis to happen properly.

3. Add a short pulse of retinoic acid, followed by BMP (a signaling protein), to activate meiosis. The Saitou lab previously published a similar meiosis induction protocol for mouse cells. In the current paper, the main advance was finding that a short retinoic acid pulse was better than more prolonged stimulation.

4. After 13 days of culture, a lot of cells died off, but about 70% of the surviving cells completed prophase I of meiosis and arrested in the diplotene stage. Now the cells are dissociated (breaking up any intercellular bridges) and cultured as single cells for an additional 21 days, during which time they activate oocyte-specific gene expression and grow to a large size.

The researchers performed extensive optimization experiments to identify the best culture conditions for oocyte growth, identifying antioxidants and activators of important signaling pathways (PI3K/mTOR, cAMP, WNT, SCF, FGF)2 as key media components. The gene expression (at the RNA level) in the oocytes was quite similar to in vivo, and the oocytes grew to a size of 60 - 65 µm (compared to about 80 µm for a fully grown mouse egg). Upon stimulation, the oocytes resumed meiosis, completing meiosis I. However, they were not able to progress to meiosis II.

How good are these eggs?

The fact that they were able to generate eggs, even at an immature stage, without ovarian supporting cells is highly impressive. However, these eggs are not very good, probably due to the lack of supporting cells.

The biggest issue is that the eggs can’t do meiosis II, and thus can’t be fertilized. This is probably a result of the fact that they can’t grow to a fully mature size. The researchers also found increased mitochondrial activity in these oocytes compared to in vivo, probably as metabolic compensation for lack of supporting cells.

Additionally, there are epigenetic issues with the eggs. Although the DNA methylation erasure from stem cells to meiotic cells was pretty good (erasing all but 9.2% of genome-wide methylation compared to 2.8% in vivo), the re-establishment of methylation in female-specific imprint patterns during oocyte growth was defective. However, the X-chromosome activation status of the eggs was pretty normal.3 The researchers also examined histone modifications, and found them to be reasonably close to normal, implying the main defect was with DNA methylation.

What’s special about m220-5 feeder cells?

Although the researchers were successful in generating eggs without ovarian supporting cells, they still needed feeder cells to support their culture system. The cell line they used was m220-5, a variant of the m220 cell line. m220 cells were engineered all the way back in 1993, through transfecting Sl/Sl4 mouse hematopoetic stromal cells4 with an expression vector for a membrane-bound isoform of the growth factor SCF. In order to use these as feeder cells, they need to be mitotically inactivated, or else they will overgrow the culture. The straightforward way to do this is treating them with mitomycin C, a DNA damaging agent, but unfortunately m220 cells will completely die off when treated with mitomycin, instead of just halting proliferation. So, the Saitou lab screened 242 subclones of m220 (which had accumulated random mutations over prolonged culture) to find ones that survived mitomycin treatment. The result was m220-5.

Besides the current study, the Saitou lab also used m220-5 cells in their paper last year about epigenetic erasure in human PGCLCs. Feeder cells which overexpress membrane-bound SCF are a valuable resource.

The question now is: what besides SCF do the feeder cells provide? There are likely other important signaling pathways involved in promoting meiosis and egg growth. Anything which is expressed both by these feeder cells and by the ovarian supporting cells would be a likely candidate.

What does this imply for in vitro meiosis and oogenesis?

The first key takeaway is that meiosis can happen properly in 2D culture,5 even without ovarian supporting cells. Although some kind of feeder cells (e.g. m220-5) are still required, this is a lot easier than using ovarian supporting cells. This is good for iterated meiosis, because it’s a lot more scalable to do meiosis in 2D culture rather than an ovarian organoid 3D culture system.

The second key takeaway is that the eggs didn’t grow to full maturity without ovarian supporting cells. So, I think ovarian supporting cells will still be required to make high-quality, fully grown eggs. But perhaps the ovarian supporting cells could be combined with the egg precursors after the egg precursors arrest in diplonema.

Overall, this is a groundbreaking study that overturns previous ideas in the field: ovarian supporting cells are not actually required for meiosis nor for the early stages of oocyte growth. I’m quite surprised it didn’t end up in Cell instead of Developmental Cell – I guess even Mitinori Saitou might get harsh peer reviewers sometimes.

So, if things go well at Ovelle, maybe in a few years we’ll be eating a lot more caviar!

This was my first time at LessOnline, since last year it conflicted with a conference on meiosis. I’m happy to say it was a great experience, and I wish I had skipped the meiosis conference and gone last year.1

A burning itch to know

LessOnline was hosted at Lighthaven, a former hotel in Berkeley converted into a venue for events. The vibe of the space reminded me a lot of the MIT campus – a place where nerds can be themselves without feeling cringe about it, where it’s cool to write linear algebra or Elvish on the walls.

At the opening session, one of the organizers asked us to think of a word that characterized the people there. What I thought of was “curiosity”. It seemed that we all had a desire to know more about the world, not just because knowledge is power, but because knowing was interesting in itself.

I didn’t realize it at the time, but apparently curiosity is the First Virtue of Rationality. I guess I was in good company!

AI is a big deal

Although I don’t have access to the attendee survey results, I’d estimate that 80% of the attendees were working in software, of which 80% were working on AI. The general mood was pretty grim – a common conversation topic was “what’s your p(doom)”?

Daniel Kokotajlo, one of the main authors of the AI 2027 report (which predicted that superintelligent AI would be likely be developed by the end of 2027 and then go on to kill all humans) gave a presentation about this report. During the Q&A, when I asked him “What is the most important non-AI thing to be working on?”, he replied, “Prediction markets, to predict AI better.”

I tend to agree that superintelligent AI is likely within the next decade – and even if AI doesn’t become superintelligent soon, it will end up being “merely” as important as the Internet. However, there’s also a good chance (I’d estimate 33% or so, although this is a high variance estimate) that the current AI boom stalls out before reaching superintelligence. So I don’t think it’s worth living as if the world is going to end in 5 years, or neglecting to invest in biology research that may take 20 years to pay off.

Even outside of LessOnline, the Bay Area is really feeling the AI. I haven’t seen these kinds of billboards in Boston, but they’re everywhere around San Francisco:

Biology is cool too I guess

I gave a presentation about my work on meiosis and the research we’re doing at Ovelle to develop a system for growing human eggs from stem cells. This went pretty well, although it was difficult to find a balance between making things understandable for computer scientists who didn’t have bio experience, while at the same time not boring the other biologists who were there.

My friend from the Church lab, Devon Stork, gave a presentation about engineering bacteria to grow on Mars, which was super interesting. (You all should subscribe to his Substack!) I also enjoyed talking with Keoni Gandall about DNA synthesis, with Chase Denecke about gene editing, with Georgia Ray and Gavriel Kleinwaks about challenge trials, with Jeff Kaufman about pandemic prevention and early detection, and with his wife Julia about parenting. And I was glad to learn that my suggestion to Aella about working around her needle-phobia by using a jet injector was a huge success.

(There were a lot more interesting people too – if we talked and I’m not listing your name here, please don’t take offense!)

Other highlights

I learned about Anki and how to make cards using LLMs. Initially I was super excited about this but it turned out that AI-generated Anki cards had more errors than I realized.

Whoever came up with Rationality Cardinality is a genius.

The Fooming Shoggoths were at it again. The quality of the songs was somewhat variable, but I really liked “The Ninth Night of November”, which was about FTX and its tragic downfall.2

On the way home, I captured the essence of the Bay Area in one (slightly out of focus) picture:

I hope to be back next year, assuming AI doesn’t kill us all before then!

Last weekend I went to the LessOnline meetup in Berkeley. There were a lot of cool people there, and I hope to eventually write up a post focused more on that, but alas, running a startup leaves little time for blogging.

Anyway, one of the several things I learned about at LessOnline was Anki, which is basically a flashcard app that uses spaced repetition to help with memorization. I had previously heard of Anki, but dismissed it as too much work to set up. By the time I was done making all the cards, I would know the material anyway and there would be no point practicing it.

But it turns out that LLMs are really good at making Anki cards from academic papers (especially review articles). I can now upload a PDF to Claude and ask it to make 5 Anki cards about the key takeaways from the paper.

I’ve recently been learning more and more esoterica of meiosis and developmental biology. Reality has a surprising amount of detail,1 and being able to reason through complex problems requires me to actually remember these details instead of having to look them up every time. Anki is extremely useful for this. I can also share my deck with my other team members at Ovelle to help everyone learn.2

The main risk of using LLMs to make Anki cards is memorizing incorrect information. Out of the first 100 cards, I caught two errors (one was using yeast gene names for human meiosis, the other was because the model didn’t know what the TTM complex was). So I do have to be vigilant, and I wouldn’t use AI-generated cards on topics where I couldn’t tell if something seemed wrong. This method is most useful for solidifying information that I already know vaguely, rather than learning completely new things.

In parallel, I recently have grown tired of Duolingo. I’ve kept up a streak for slightly over 5 years (starting during COVID lockdown) but I reached the maximum level on the Polish course in late 2022, and it hasn’t really taught me anything since then. I do think the “streak” mechanic is brilliant for incentivizing me to practice every day, but at this point, I’m more Goodharting my streak rather than actually learning things. If I want to learn more Polish, I’m better off including that in my Anki deck (or practicing with an LLM) instead of using Duolingo.

Fortunately, I can also set up Anki with a streak mechanic, and although I don’t think I can officially “transfer” my Duolingo streak, I can always mentally add +1835 days. So, starting tomorrow, I’ll be setting aside Duolingo for good, and fully switching over to Anki. Here’s to further learning!

I’ve previously written about gene drives and their potential to fight malaria by eliminating the mosquitoes that spread it. In 2023, I also wrote about some of the technical and political considerations about gene drives, and concluded that political and regulatory issues, rather than technical problems, were the main roadblocks.

This year, in observance of World Malaria Day, I talked with Dr. Michael Santos, the director of GeneConvene Global Collaborative, an organization working to provide technical and regulatory advice to implement genetic biocontrol strategies for diseases such as malaria.

We discussed the overall concept of gene drives, their potential use against malaria, and challenges that need to be addressed before deployment. Overall, Dr. Santos thinks that the first trials involving release of gene drive mosquitoes into the wild are still 3 to 4 years away, with key obstacles including managing resistance in diverse populations of mosquites, and obtaining approval from regulatory bodies.

If researchers or policy workers want to get involved, Dr. Santos recommended checking out the resources on the GeneConvene website, and also reaching out to the Outreach Network for Gene Drive Research, an organization which connects scientists and regulators to help inform gene drive policy.

Below is a (lightly edited) transcript of our conversation:

Metacelsus:

It's great to meet you, Dr. Santos.

Michael Santos:

Great meeting you as well, and please feel free to call me Mike.

Metacelsus:

I've been interested in gene drives for quite a while. I think I found out about them around 10 years ago. Just for their potential in malaria suppression and improving global health. I've been doing a bit of writing about them over the years on my blog, and it's great that you could talk to me today.

World Malaria Day is coming up, and we should try to focus on methods that are able to prevent this terrible disease. Because right now, it's one of the biggest killers of people around the world.

Can you, for readers or listeners who aren't familiar with the topic area, explain a bit more about how gene drives work, and how they show promise in preventing malaria?

Michael Santos:

Yeah, absolutely. Setting the World Malaria Day context is great because right now, we're still in a situation where a child dies every minute due to malaria, and progress in reducing malaria deaths, which has been impressive from 1990 through several years ago, has plateaued recently. WHO and others have highlighted the need for new ways of attempting to control and ultimately eliminate and hopefully eradicate malaria.

That's the context for why people are exploring these genetic biocontrol approaches, one of which is gene drive. Gene drive is a process where a gene or a genetic element increases its prevalence in a population of organisms over time, more so than you would expect, based on its fitness effects.

This is a natural phenomenon. In fact, it's almost the 100 year anniversary of the first observation.

Metacelsus:

You mean like, drosophila P elements and things like that,1 or other natural gene drives?

Michael Santos:

I mean in Drosophila obscura. I think in 1928 was the very first observation of sex bias that I don't know exactly how well was understood at the time, but that was the first observation of what would later be understood as selfish, genetic elements. The P element would later be identified. Transposable elements identified by Barbara McClintock in maize, and in others.

The beginnings of the understanding of this natural genetic phenomenon are quite old and go back quite some ways. This observation that it's possible for genes or genetic elements to have this property has been understood for a long time, and it's observed in many different organisms and through many different mechanisms.

Metacelsus:

Yeah, there’s definitely a lot of diversity in the ways that this can happen.

Michael Santos:

If anyone is interested in going deeper, we have some explainers on our Youtube channel and some more information at the GeneConvene Virtual Institute that summarizes some of the different categorical mechanisms of gene drive. But this observation that you can have these genetic elements that become more prevalent in a population over generations.

Since about 1960, people have suggested that maybe this could be harnessed, either natural or engineered analogs, for the control of vectors that transmit diseases. So this is also quite an old idea.

But in the decades since then there's been a lot of progress, and particularly recently, but a lot of progress over that time. And you know first transgenesis to begin with, and then ultimately, these genetic engineering tools, like CRISPR-cas9, that a lot of people are familiar with, but others that preceded it, that have made it easier to engineer this gene drive phenomenon into organisms like the mosquitoes that transmit malaria.

Metacelsus:

CRISPR, in particular, was a big advance because you can more precisely target genes that you want to have as part of your gene drive. I think the more recent gene drives in the literature have nearly all used these CRISPR systems that can precisely target the site that they're going to copy the gene drive into.2

Michael Santos:

If we look at where things have progressed to now, this idea that you could get these genes to increase in prevalence over time. People identified a couple of different mechanisms through which you could limit the spread of a disease like malaria. That's in the case of malaria transmitted just through this vector.

So one is reducing the numbers of the vectors and the other is making a modification in the vector such that the pathogen can't be transmitted. In the case of malaria mosquitoes, there have been investigations both into sex distortion, making the offspring all or almost entirely male, that would eventually lead to a nearly all male population and thus a dramatic reduction of the population, or reducing fertility or something in that pathway that would just directly reduce the number of mosquitoes.

Then the other approach, the malaria parasite has a particularly complicated life cycle and goes through several life cycle stages inside the mosquitoes. And engineering the mosquitoes with elements that are expressed in the mosquito after it takes a blood meal that interferes with the parasite's life cycle is another way of potentially preventing the spread of malaria.

Both those approaches right now have been engineered into malaria vector mosquitoes and those approaches have been demonstrated in laboratory studies. They've demonstrated both the drive that it increases in prevalence over generations, as well as whichever effect they were trying to introduce, either a suppression of female fertility or an inhibition of the ability of the parasite to complete its life cycle inside the mosquito, and that has been achieved with CRISPR-cas9 based homing drives, although in earlier stage work. People have demonstrated a number of different kinds of mechanisms are possible to be implemented leveraging these tools that we have now.

Metacelsus:

In terms of your background and what you've been involved in this area, can you tell me more about how you got started in this? And also, what you're working on now with GeneConvene?

Michael Santos:

I started working in this area in about 2015, when I was at the Gates Foundation in the health program there. The team that I was on was one of the funders of this research. I got involved on the research funder side of it originally, and then in 2019, I moved from the Gates Foundation to the Foundation for the National Institutes of Health, where I am now, which is a nonprofit organization that was created by law in the US to support the mission of the NIH, and manages many different scientific partnerships, and has actually been involved in this particular area for 20 years, going back to the big injection of funding that occurred with the grand challenges in global health program in around 2004, 2005.

Since 2019, I've been at FNIH, part of this program called the GeneConvene Global collaborative, which I lead now. GeneConvene itself is an initiative, a program within FNIH that supports informed decision making about genetic biocontrol approaches for public health.

One of the main areas of focus for us are these gene drive approaches, specifically for malaria and sub-Saharan Africa. We support that mission in a few different ways. One is primarily convening people to identify and address key questions. Although there are elements of gene drive applied to malaria that have precedent, obviously there's genetic modification of organisms, there is classical biocontrol approaches, there are other malaria control approaches.

Metacelsus:

You mean like sterile insects?

Michael Santos:

Yeah, sterile insects as an example of genetic biocontrol achieved through radiation impact on the genetics. But classical biocontrol. For example, the introduction of a natural predator from the home range of what's now an invasive species. Then obviously conventional vector control, like insecticide-based vector control. Genetic biocontrol of malaria vectors draws from all of these different areas. So there are opportunities to bring people together and identify the specific questions for gene drive approaches, and help inform them. We also do provide technical advice and technical services, especially around regulatory science, which we'll talk about a little bit more later. And then also a key focus for us is on strengthening capacity and sharing information. A big goal for us is to help all stakeholders that are interested have a greater awareness, have a deeper understanding, provide resources. At a wide range of scales, from helping people understand the basics through helping people understand the technical depths. That's what we do and what I'm delighted to be part of.

Metacelsus:

Sounds like you're doing some good work there. As you're partnering with other organizations, are there any promising organizations that are working in this area? I'm aware of Target Malaria, which is doing a lot of things, and I believe Burkina Faso. Do you know of any other gene drive research that's going on either in academia or in nonprofit organizations that you think should be highlighted?

Michael Santos:

There are a lot of different partners that work in this space in many different ways. One of the really important ones to highlight is the African Union and the African Union Development Agency-NEPAD. The African Union convened a high-level panel on emerging technologies to evaluate different technologies for their potential to contribute to development goals within Africa. One of the areas they identified and prioritized was supporting African countries to develop their capacity around research and evaluation of gene drive approaches for malaria. The African Union Development Agency then supports different African countries and regional groups in order to build their capacity for that. They're a really important partner for us around the capacity strengthening activities I talked about.

There's also the African Genetic Biocontrol Consortium, which we provide support to, which helps convene people who are parts of different organizations from these different stakeholder groups of vector control, from one health or from biosafety within Africa. Organizations like WHO play an important role in providing their perspective on new tools.

I know you wanted to know on the science side too, so in terms of the groups that are most advanced in developing these technologies specifically for malaria there is Target Malaria, as you mentioned, which is a a multinational research collaboration which so far has been primarily pursuing this female fertility reduction approach. There are two other research consortia that are pursuing these population modification approaches of trying to stop the parasite from being transmitted by the mosquito. One called the University of California Malaria Initiative, and another called Transmission Zero. Within malaria mosquitoes, those are the most advanced programs. For controlling invasive rodents, there's a collaboration called GBIRd, which focuses on, so far, developing gene drive for mice, but with an interest long run in other invasive rodents. Those are the research collaborations that have gone farthest. But in academia, nonprofit research institutions globally, there's work in lots of different organisms for lots of different applications, for golden mussels, for New World screwworm. People have looked in ticks. People have looked in other disease vector mosquitoes besides the anopheles that transmit malaria. This kind of work at an academic earlier stage level is happening in lots of different areas.

Metacelsus:

It's about 10 years since the first CRISPR-based gene drives were actually published. In those 10 years there's been millions of deaths of malaria, especially in Africa. Why do you think there hasn't been deployment yet of gene drives to combat mosquito vectors of malaria?

Michael Santos:

The process of getting to the implementation of gene drives as a malaria control tool has lots of different steps to it. To ensure that everyone knows that the current stage is that these approaches have been studied in cages in bio secure facilities. But not in the environment yet. There has not yet been an application put into a regulatory authority to conduct a field trial for gene drive. Yet. It is anticipated that some of these groups that are most advanced in their research, like I was talking about, are likely to do that within the next 2 or 3 years. It's a question then of whether regulatory authorities will approve those field trials, and whether they'll be funded.

Metacelsus:

Which regulatory authorities will these be? You mentioned, the African Union. Will it be the actual governments of those countries?

Michael Santos:

The way genetically modified organisms are regulated is not uniform throughout the world. However, most countries in the world are signatories to a treaty called the Cartagena Protocol for Biosafety that's under the UN Convention on biological diversity. In those countries there is an authority that is specifically responsible for assessing biosafety, and the scope of those authorities includes genetically modified organisms generally. It's the case that these mosquitoes that are genetically engineered with gene drives would fall under their jurisdiction. An organization that is interested in doing a field trial would go through the process that's described in their country in those regulations in order to put in an application, and then receive a decision potentially with conditions about what's required to do as part of that study, and potentially move forward if they're approved.

Metacelsus:

Gene drives in theory would spread beyond any country, especially if they're in an area with a lot of mosquitoes. How is that going to be handled in terms of the applications?

Michael Santos:

That's a key topic of discussion. One of the reasons that it's really valuable that the African Union Development agency is involved in the case of the primary vectors of malaria in Africa. Those vectors exist in many different countries in Africa, although generally not outside of Africa. For example, there's been a lot of work in the West Africa region through this organization called ECOWAS, which is the French acronym for the West Africa region. In these countries, there have been representatives from the different ministries and authorities have been convening together to develop kind of regional perspectives on best practices, which they published in a series of guidelines that are publicly available through AUDA-NEPAD. Although these are country-level decisions, there's a strong emphasis on bringing together countries at the regional level and to the continental level under the AU, so that there's an opportunity for transparency and collaboration. In some cases, for example in West Africa, there's actually a mechanism for making regional biosafety decisions, so other regions are looking at that too.

Metacelsus:

In terms of the biggest regulatory or political challenges that's holding back the use of this technology, what would you say the main issues are that need to be addressed?

Michael Santos:

Right now, none of the research development groups have even progressed to the point of applying yet. If we look at the factors behind that, that's because, you mentioned the first demonstrations were about 10 years ago, but people, when thinking about malaria control, look at this through a lens of what they call a target product profile or preferred product characteristics. And looking at the performance that you would want to see in order to be confident that as an intervention against malaria it could be successful, so there has been a fair amount of work, actually quite a lot of work since those initial demonstrations in trying to improve the performance of the gene drives.

For example, in the case of a fertility reducing mechanism, there's really strong selection pressure for functionally resistant alleles. So, eventually exploring different mechanisms to try to minimize the formation of resistance in the population has been a key frontier for additional work over those 10 years. So, part of it is getting to the point where researchers feel confident that they actually have a construct that is suitable for use. But to do a field trial, you need to then do some baseline characterization work of your field trial site. There's, of course, community engagement work with the communities that's part of that and preparing for that. Then, in order to put in an application to an authority that's going to assess the biosafety risk, there's a number of experiments and studies they're looking to have done and see as part of that application, so that they can evaluate it and make their decision. Similar to if you were approaching a clinical trial for a new medicine, the way that you would put together a package of information to present to the regulator.

Metacelsus:

What types of experiments would those be? Large cage studies?

Michael Santos:

So many different domains, but thinking through the different categories of risk. Human or animal risks are things like allergenicity and toxicity, basic kinds of studies, whether the transgenes result in the production, in some cases they deliberately result in the production of new proteins. Whether you see any allergenicity or toxicity effects with a mosquito bite or ingestion of the mosquito. There are questions like stability of the construct. So, what can you demonstrate in there. As you say, you might do it through case studies of following the construct through multiple generations and being able to show its stability, at least over the period of time that you can study. But there are many different things to think about, and people have done work on the different potential pathways to harm. So, the sort of science of risk assessment is to talk about what the protection goals are, what the potential harms are, what the pathways to those harms are, and then ask whether there are experiments or studies you can do in order to inform your understanding of the probability or the magnitude. That's the kind of work that goes into these dossiers, as well as characterizing what the research study you would do would be, like actually designing the field trial.

Metacelsus:

In terms of the technical barriers to achieving that level of data, what do you think are the most pressing technical issues currently with gene drives? Is it resistance? Is it actually being able to spread efficiently? Is it scaling up the production of mosquitoes? Because if these mosquitoes have low fitness, it might be hard to actually produce enough of them. What do you think is the main roadblock in terms of the technical capabilities?

Michael Santos:

I think a lot of the development focus again, like a lot of the reason why time has elapsed between the first demonstration and now is improving performance on the efficacy side. Especially managing resistance to the gene drive in the case of population suppression, but also looking at the performance for population modification approaches of preventing the transmission blocking effect. For example, not just how it performs against a laboratory strain of the malaria parasite, but how it performs against the wider diversity, genetic diversity of parasites that are encountered in the field. That's some research that's happening right now. So, understanding and being confident of the efficacy has been a key driver of scientific development. There are questions about the performance in the field that it's unknown whether there will be technical barriers or not until those experiments are done, and as you say, there are certainly important questions downstream around implementation. but those aren't necessarily gating for people getting to the point of conducting the first field trials and looking at the performance in the field, which will be, if field trials move forward, a dramatic step up in the understanding of the performance of these approaches.

Metacelsus:

What is GeneConvene working on in this space in terms of addressing these issues either on a technical or political level?

Michael Santos:

We work both on the side with the developers of helping to inform their decisions how they're moving forward, and on the governance side. On the developer side, we've been working quite closely with the 3 groups I mentioned that are working in malaria on the way that the first field trials will be designed. There are properties of gene drive that are intended to persist and spread that create complexities for designing a trial where you're attempting to measure the impact of the intervention. That's been technical work. There's been great collaboration among the modelers that are working within each of those groups as they're developing their modeling and their statistical analysis plans. That's one of the kinds of areas. Another kind of area is thinking about what kind of monitoring and surveillance will happen as part of the trials and after the trials. We held a workshop last year, and we have a report from that that will be coming out soon looking at specifically environmental monitoring. There are environmental risk questions.

Metacelsus:

Monitoring the spread of the gene drive, or monitoring the effects on other species?

Michael Santos:

Both are important. But monitoring the malaria vector mosquitoes, at least there's maybe a little bit more experience, because there's a lot of other malaria vector studies and vector control intervention studies that do that. But then, looking at the effect on what people call non-target organisms, organisms that you're not trying to affect with the vector control intervention because there are questions that people ask like: if you suppress the population of these malaria vector mosquitoes, does it have any impacts in the ecosystem, any impact on biodiversity, any impact on the ecosystem services? That you would want to see, so we just launched a webinar series earlier today on, specifically, this topic of environmental monitoring where researchers that work on different aspects of this will be talking about some of the different approaches. We have a couple talks coming up on environmental DNA called eDNA. So, this is on the technical side. On the governance side, we work a lot in regulatory science and working through the mechanisms of the UN Convention on Biological Diversity to provide new guidance on how to do environmental risk assessment for engineered gene drive mosquitoes that was endorsed late last year at the Conference of the Parties, and then working with partners, like the African Union Development Agency, on training for regulators and for other stakeholders to familiarize them with these technologies, to answer their questions and to help strengthen their capacity to perform rigorous risk assessments if, in the future, these applications come to them.

Metacelsus:

I'm curious. There are potentially impacts of gene drives that could hypothetically be negative, like affecting other species or, you mentioned, causing some sort of allergic reaction. If you had to pick one thing that is most on your list of things you're worried about, in terms of gene drive effects or any reason that they might not be a good idea to use, what is your biggest concern in this space that you want to watch out for?

Michael Santos:

I think the biggest concern probably is that they just don't work very well in the field, because if they don't deliver benefit for malaria, then there's no risk benefit calculation at all. Like if the gene drives simply do not work. In terms of if they end up being efficacious at suppressing the transmission of malaria, but at the same time risks to other important areas are identified, then that's exactly why countries have governance processes that can weigh the different areas of value, and collect stakeholder perspectives on this and come to some opinion about how they're going to balance these different needs. That's a really important part of having these governance systems in place because that's a judgment that countries, regions and continents need to make for themselves through their mechanisms. What we want as much as possible is for people to be able to make those decisions in an informed way, both to have good information to make that decision with, and to have the capability to evaluate that information and have an informed way of weighing potential benefits and potential risks.

Metacelsus:

When you look at the impact of malaria, especially in Africa, both on number of lives lost, but also on lost economic growth, it's pretty large. I would think that it would be in the interest of many of these governments to deploy this technology. Which do you think will be the first country to allow a gene drive trial if you had to guess?

Michael Santos:

First to say on the first point, I think the importance of the challenge of malaria and the potential of this technology is why the African Union picked it out as an emerging technology for specific capacity building focus. In terms of where, the countries where the research is most advanced include Burkina Faso, Sao Tome and Principe, Uganda, and Tanzania. Those countries are probably the most likely countries within which applications could go in to conduct field trials. Doesn't mean for sure that one of those countries would be the first place that it would happen.

Metacelsus:

Sao Tome and Principe are specifically islands, right? So at least in theory, if you did something like this on an island, you might not have spread to the mainland, or is that likely to happen regardless?

Michael Santos:

People don't know for sure. This is something people study. They look at the population genetics of the mosquitoes on the islands and on the nearest mainland and the mainland locations that are connected by transit and try to understand the rate of gene flow. I think those populations are relatively well isolated. Having said that, gene drive is different than regular gene flow, because you have the ability of the gene to increase in prevalence from a very low initial percentage, even if the mosquito that makes that journey is not particularly fit in the local background. There's uncertainty about this. The oceans do create a geographic barrier, but the sort of height and strength of that geographic barrier in this specific use case isn't known for sure.

Metacelsus:

There are a few designs for drives, I believe they're called daisy drives, that will only copy themselves for a certain number of generations. It does add complexity to the system, so it's more complicated to design and implement, but would that be something to evaluate if you wanted to do a more limited initial trial?

Michael Santos:

As you say, there's a range of mechanisms that can impose some control, spatial or temporal, through either mechanisms where the drive would decay over time or mechanisms where there's a threshold to the drive occurring or not. These approaches definitely are being studied. I mentioned there's a lot of research that's happening in this space, and there are people who are working to implement some of these different kinds of methods. For example, some of the different gene drive mechanisms themselves have intrinsic higher thresholding effects than homing drives which have nominally no threshold to their growth. People are working on those. It may be the case that when we get to the point of countries making regulatory decisions, maybe countries will not want to move forward with self-sustaining low-threshold gene drives initially. Maybe they will decide that they would prefer to have more intermediate steps.

The different research programs are approaching this differently, so the Target Malaria program has already done a field study of genetically sterile mosquitoes in Burkina Faso. So that proposal was put in. The regulatory authority approved it. They moved forward. They've put in a proposal to do another study of a construct that is self-limiting, that's not a gene drive. It will decay out of the population over time, but is not genetically sterile, so the construct will persist for some number of generations. So these are sort of stepping stones toward a self-sustaining gene drive. Transmission Zero in Tanzania is planning to evaluate a split drive, which is similar to the simplest iteration of a daisy drive, where you separate the drive mechanism from the effector mechanism so that it will get temporarily boosted and then decay away. They're also planning to proceed with this kind of stepwise approach. So, there are different ways to approach this, depending on what researchers propose and what regulatory authorities approve.

Metacelsus:

You mentioned that it'll take about 2 or 3 years for these organizations to even be at the point where they're applying to do the gene drive field trials. How long do you think it will be before we see the first gene drive field trial?

Michael Santos:

It's tough to say because I don't know what decisions regulatory authorities will make. If the first applications go in in a couple of years and the first applications are approved, then maybe gene drive trials are beginning in four-ish years or something like that. Of course, the alternative is applications don't go in on that timeline, or the first applications are not approved, in which case the timelines could be longer. We sort of know around the minimum time, but the distribution extends beyond that in a way that we can't really constrain.

Metacelsus:

Presumably, these trials would also involve monitoring, which would go on for a period of time afterwards. How how often would gene drives need to be re-released after the first trial? I guess it would depend on resistance and things like that, but is this likely to be a one-and-done type of situation, or will we need to periodically update drives as resistance evolves to some of the initial versions?

Michael Santos:

That the kind of question that field trials will be really important to inform, because we don't know. It's assumed that for almost anything, at some point resistance will become important. The timescale on which that happens has a pretty wide range of plausible values. If you look at computer models of the introduction of gene drives that have properties similar to what are observed in the laboratory cages, and that assume that those similar properties will occur in the field, then they could sustain for a long time, but we don't know for sure that the performance will be the same. In the timeline, there's some sort of stochasticity to the mechanisms that are associated with resistance arising. It may also be different in different locations. In many places, mosquito populations are highly seasonal in their abundance, so there could be, certainly, when people study this with modeling, there are certain properties that make it more or less likely that the gene drives disappear from local populations and the local areas then recolonized by wild type mosquitoes without the gene drive. So, in those cases you would, in principle, need to re-release. But how common that actually happens, in practice, will only be understood once there are field trials.

Metacelsus:

Has there ever been any research into gene drives in the malaria parasite? I know that has both asexual and sexual reproduction stages, or other, beyond mosquitoes, other approaches.

Michael Santos:

I am not aware of anybody looking at it specifically in the parasite. People have looked at gene drive-like mechanisms in viruses, so in pathogens is the case. I'm not aware of any work specifically in the malaria parasite. There are many different applications that people are looking at for gene drive, besides mosquito-borne diseases. Of course, other vector-borne diseases. The control of agricultural pests. For example, the New World screwworm is controlled in North America with sterile insect technology. In South America, researchers in Uruguay, have proposed the possibility of gene drive approaches for controlling it, and, in conservation, like I mentioned, in the application of removal of invasive rodents from islands. That's probably the area that there has been the most attention. But people have speculated about other conservation applications, like squirrels in in the UK and others, where the research is much farther away from being able to implement it.

Metacelsus:

Before we end, is there anything else you'd like people to know about GeneConvene or about gene drives in general?

Michael Santos:

We're really committed to raising awareness and understanding of these approaches. I mentioned this before, but just to reiterate, at the GeneConvene Virtual Institute, we provide a lot of information. There's a very large, frequently asked questions section that goes through many of the different questions that people ask about this, including many of the different kinds of risks that people are interested in. We have a YouTube channel where you can see how someone makes a transgenic mosquito, explanations of different kinds of gene drives, or the basics.

On the website, we have a piece on the history if people want to see the kind of seminal papers on the history of the thinking about this approach. I definitely encourage anyone who's interested in learning any more about this to check this out. We have a newsletter, which you can subscribe to, a webinar series, and lots of different resources. And if people want to learn more and can't find a resource, please let us know. In that same vein, I want to thank you. We really appreciate people who are committed to communicating and helping more people learn about and understand this. I really appreciate the work that you're doing and the opportunity to talk with you today.

Metacelsus:

You’re welcome. If there are other things besides communication that you're looking for, like people getting involved with, how can people help? Are you looking to hire smart PhD biologists to work on gene drives, negotiators for political treaties, or things like that?

Michael Santos:

A great thing for people to check out would be the Outreach Network for Gene Drive Research. That's an organization that includes representatives from many different labs and organizations that are working on gene drive in different applications across different fields, so especially for people that are coming more from the research side. One of the things they do that you just mentioned was help people who are interested in engaging in some of these processes. Scientific experts can provide input into some of these discussions, negotiations and governance processes, because that is a really important role for researchers to play. When you go to these meetings, there are a lot of scientists there that are supporting the discussions. I can make sure you have the link for their website and organization. So, people can check that out as well.

Metacelsus:

All right, thank you. It's great talking.

Michael Santos:

I enjoyed it. I definitely appreciated the opportunity. Thanks for connecting.

My family’s beloved black cat, Lucy, had died. Devastated, I begged my parents: we need to get another Lucy. They must have felt the same way, because a few days later, we visited the Minnesota Humane Society, where there was a crowd of cute kittens up for adoption. One stood out to me, a black kitten with green eyes. The Humane Society staff had named him Spitfire, due to his high energy. But after bringing him home, we decided to name him after an icon of rebirth: Phoenix.

Over the years, Phoenix and I became best friends, sharing many special moments. In high school, he often helped me with my homework:

And even in college, I still saw him whenever I visited my parents. At the beginning of 2022, my partner Ula and I adopted Phoenix, welcoming him into our apartment. At this point, he was already showing his age, and, like many old cats, he had developed kidney disease. Still, he still had the zoomies from time to time, and really enjoyed his cuddles. Ula gave him a new nickname, “The Bean”, because he was “a cute little bean”

For someone who hasn’t met him, it’s hard to explain why Phoenix was the best cat ever. Of course, I’m biased. But Phoenix truly had the best of both playful energy, and calming cuddles. When he sat on your lap and purred, you felt so relaxed that you just couldn’t do anything except pet him. He also had a charming personality, including a morning routine that involved asking for fresh water in his fountain, then either sitting on the windowsill in the summer, or burrowing under the blankets in the winter. Whenever we were sick, Phoenix would always try to help us feel better. I’ve met plenty of other cats, but none of them were like Phoenix. For 17 years, he was truly a family member and a constant companion.

Unfortunately, Nature has a way of destroying everything dear to us. Domestic cats typically live for 13 - 17 years, and Phoenix was no exception. By the end of his life, he was suffering from kidney disease, arthritis, and a heart murmur. We tried to give him the best quality of life he could have. Starting in May 2024, we even injected him with subcutaneous fluids every day (100 mL lactated Ringer’s), which he tolerated remarkably well. He hated injections at the vet, but apparently trusted us enough to be injected at home. I think he understood that the fluids made him feel better.

At his last checkup in January, the vet found that his blood urea level was 5 times higher than the unhealthy threshold, and remarked that we must be taking good care of him, because he still appeared to be doing well in spite of his kidney disease. In fact, what finally did him in was not his kidneys, but a rapidly growing nasal tumor. We first noticed some swelling on Phoenix’s face around February 13, and a week later, it had progressed to the point that he was refusing food and water. We had previously discussed end-of-life plans for him, but I was surprised how quickly he declined in the end. On Friday, February 21, 2025, we had a vet come visit our apartment to put Phoenix to rest. He was just a month away from his 18th birthday.

Only mostly dead

I had often talked about preserving Phoenix’s cells after he died. After all, with a name like Phoenix, it was only fitting that one day he might be re-born as a clone. I knew that Viagen Pets had successfully cloned cats starting from skin fibroblasts. The only catch was that the cloning cost $50,000, and I really couldn’t justify spending that much when there were plenty of better causes to fund. But, if I could preserve his fibroblasts, there was a chance that I could clone him in the future when the technology was more accessible. Plus, I needed to practice fibroblast isolation anyway, because some of Ovelle’s future work will involve using skin-derived fibroblasts to generate induced pluripotent stem cell lines.

Isolating fibroblasts requires taking a skin sample, and I didn’t want to hurt Phoenix while he was still alive. After all, he had already lost the tip of one ear! So a post-mortem sample was the only option. In the last few weeks of Phoenix’s life, I read up on fibroblast culture methods, and found a helpful protocol from JoVE, a journal which publishes tutorial videos for experimental procedures.2 I wanted to try out the protocol using my own skin, and I got all the materials to do so, but at the end, Phoenix declined so rapidly that I didn’t get a chance to test out the method first.

Following a guide from Viagen, I took some skin samples starting about 45 minutes after death. This was more difficult than I expected. The hardest part was removing all the fur! I also was very emotional — I’ve dissected several mice over the course of my research, but it’s really different when it’s your own pet. I transported the samples to the lab in tubes of lactated Ringer’s solution, and set up the cultures. Following the JoVE protocol, I cut up the skin samples using two scalpels and put the pieces into 6-well culture plates with a small amount of growth medium.

Again, this was a bit trickier than I expected, and I had trouble cutting the pieces to be small enough before putting them into culture dishes. Following a different guide, I also cryopreserved a few skin pieces just as a backup in case the culture didn’t work.

The hardest part of the fibroblast culture was that I would have to wait for several days to know whether or not it worked. Importantly, the skin pieces needed to attach to the bottom of the culture plates, and if I was too impatient and moved the plates around too much, it might disrupt the attachment, causing the cultures to fail!

So, just to be sure, I took some additional samples the next morning (after keeping Phoenix’s body refrigerated overnight). This time, knowing what to expect, it went much more smoothly. Afterwards, we brought Phoenix’s body to the vet to be cremated, but I kept a few skin samples (in RPMI medium with antibiotics) and gave them to one of my friends who had previous experience with fibroblast culture.

Thankfully, despite my anxiety and lack of practice, all of the cultures3 were successful! I noticed the first signs of growth on February 25 (4 days after starting the first cultures), and soon the cells were spreading across the bottom of the culture plates. The pattern of growth basically matched the description in the tutorial video: the first cells to appear were keratinocytes,4 but soon, fibroblasts overtook them and grew across the plates.

A new beginning?

In a few days, I will freeze the cultured fibroblasts. As Hayflick famously discovered, fibroblasts can’t be cultured indefinitely, and furthermore, mutations or epigenetic abnormalities could accumulate in cell culture. Thankfully, I have plenty of liquid nitrogen storage available, so a few extra vials of cells won’t get in the way of my research. And after this experience, I’m definitely more confident in my ability to culture fibroblasts from skin, which will come in handy for generating new iPSC lines.

In the meantime, I want to design a really nice urn for Phoenix’s ashes. I have an idea for something like a phoenix egg, with a winged cat sitting on top.

And who knows, maybe 25 years from now, we’ll have a Phoenix Junior. The second time around, I want to do some gene editing to make his kidneys stronger. And maybe also make his eyes express luciferase . . .

The nanotechnologist Eric Drexler popularized the term “gray goo” to refer to self-replicating nanobots that go out of control and digest everything to make more copies of themselves.

But it turns out we already have self-replicating nanobots: they’re just called cells.

The Azolla event

Around 49 million years ago, the Earth was a greenhouse, atmospheric CO2 was ~3500 ppm, and the Arctic Ocean was a freshwater lake. Then something interesting happened.

Aquatic duckweed-like ferns of the genus Azolla thrived in the Artic Ocean, creating massive blooms and sequestering about 0.9 – 3.5 trillion tons of carbon over about a million years in total. By the end of this period, atmospheric CO2 was only 650 ppm.11. Although not all of this reduction in CO2 is directly attributable to Azolla, these blooms certainly helped push the Earth towards its current glaciated state.

Notably, there have currently been calls to re-create a similar event to fight global warming. This might backfire horribly.

A thought experiment

March, 2053. Ajinomoto-Bayer-Cargill have just opened the first pilot plant for the environmentally sustainable production of L-glucose, a promising artificial sweetener. Decades of research projects on opposite-chirality biomolecules have finally come to fruition, leading to the engineering of a mirror-image variety of the marine photosynthetic microbe Synechococcus elongatus. Now, S. elongatus2 will capture CO2 from the air, and fix it into valuable chemicals. Ajinomoto-Bayer-Cargill can sell not only artificial sweeteners, but also carbon credits! What could go wrong?

July, 2053. Algal blooms are common in the Gulf of Mexico around this time of year, but this one is different. The algae die from overcrowding, but don’t seem to be decomposing at all. And the standard PCR tests for the usual algal suspects are all coming up negative. Scientists quickly determine that the bloom is due to the mirror-image S. elongatus, which apparently escaped somehow from Ajinomoto-Bayer-Cargill’s plant. Fingers are pointed, but nobody is ever able to determine the cause of the leak.

October, 2053. A crash effort has produced a mirror-image virus targeted to kill the S. elongatus, but by now the rogue organism (with a doubling time of 2 hours under ideal conditions) have spread throughout all the world’s oceans. Since they can fix both carbon and nitrogen, and have no natural predators, minerals such as phosphorous and iron are their only limitations, and they quickly sequester all available nutrients. The virus is humanity’s only hope to bring the S. elongatus under control.

April, 2054. The virus has actually made things worse. It kills most of the S. elongatus, but some still persist at a lower population density. The dead cells release nutrients (phosphorous, iron, etc.) back into the water, and carbon fixation continues. Too bad the fixed carbon is all unusable by normal life forms.

July, 2054. The ocean still tastes salty, but by now it is also faintly sweet from all the L-carbohydrates. All marine life has died out, outcompeted by the S. elongatus. CO2 levels are already down to 350 ppm with no signs of stopping, and Earth is clearly headed for a snowball state.

September, 2060. The last survivors of humanity huddle around a steam generator in the Arctic while children work in the coal mines to keep the boilers fed.3

Key takeaways

The use of synthetic biology to create engineered human pathogens is scary, but even worse things are possible if synthetic organisms can outcompete natural ones in ecosystems. Mirror-image photosynthetic organisms are a good example of this. The technology is not developed enough to cause immediate concern today, but research on mirror life poses risks for the future.

Right now, efforts to optimize photosynthetic organisms have resulted in “domesticated” strains with improved performance under laboratory or agricultural conditions but worse performance in natural environments.However, as the field of synthetic biology progresses, we should be aware of potential harmful effects of our creations on ecosystems, especially if they involve something that, like mirror life, is highly distant from any natural precedent. The broad range of potential threats means that targeted mitigation efforts (such as testing for a particular species of concern) may be inadequate. The Nucleic Acid Observatory is a good example of an effort to detect biological risks in an unbiased manner. We should be funding this and things like it.

Developing robust containment systems (such as synthetic auxotrophy) may mitigate inadvertent leaks of engineered organisms, and these systems should also be a high priority for research. However, these might provide a false sense of security because even if inactivation by mutation or recombination is prevented (which is not trivial!), whatever safeguards are engineered may be removed by a bad actor.

We should also consider the downsides of dual-use technology that could have a high potential for harm if abused. This issue has already been raised for viral engineering, but also applies more broadly to efforts to engineer fitter organisms.

Finally, we should be especially wary of any plans to intentionally release engineered organisms to globally alter ecosystems. Gene drives to eliminate specific species of mosquitoes are fine,4 but a synthetic Azolla event is not something I ever want to see.

As a reproductive biologist, I find it frustrating when people (on both sides!) make mistakes resulting from a lack of understanding of the biology behind these issues. For example, pro-transgender people often dismiss the differences in athletic capabilities between cisgender and transgender women, and anti-transgender people often confuse puberty blockers with cross-sex hormones.

So in order to help people make better informed decisions, I will provide a biological overview of the following topics:

• Definition of terms, including intersex and transgender

• Basic developmental biology of sex

• Prevalence and potential causes of transgender identity

As with my previous post, I will attempt to remain as neutral as possible, and stick to scientific facts instead of my personal opinions. And where the science isn’t fully settled yet, I will be sure to point this out, instead of confidently proclaiming something which might be wrong.

Given the length of this post so far, I’m splitting the second half into a separate post. Part 2B will focus on:

• How puberty blockers work and their potential for side effects

• Effects of cross-sex hormone therapy and other transgender medical procedures

• Athletic performance of transgender women vs. cisgender women

Definition of terms

• Sex: A biological trait, based on the production of male gametes (sperm) or female gametes (eggs).

  • Disorders of sex development (also known as differences of sex development)1: Developmental abnormalities affecting sex characteristics.

  • Intersex: People who have sex characteristics that are not fully male or fully female, due to disorders of sex development.

• Gender: A social trait, based on identification as a man or woman (or occasionally a third gender).

  • Cisgender: People who have gender matching their sex.

  • Transgender men: People who have female sex (also known as “assigned female at birth”) and male gender.

  • Transgender women: People who have male sex (also known as “assigned male at birth”) and female gender.

  • Gender nonbinary: People (of either sex) who do not identify with either male or female gender.2

  • Gender dysphoria: Psychological distress resulting from a mismatch between a person’s gender identity and sex.

• Gender-affirming care: A broad term that can refer to a variety of medical treatments aiming to reduce gender dysphoria, including puberty blockers, hormone therapies, and surgeries.

  • Puberty blocker: Medication that prevents the production of sex hormones (thus preventing puberty).

  • Masculinizing hormone therapy: Medication that increases levels of male sex hormones and decreases levels of female sex hormones.

  • Feminizing hormone therapy: Medication that increases levels of female sex hormones and decreases levels of male sex hormones.

  • Gender-affirming surgery (also known as “sex reassignment surgery”): A variety of surgical procedures, intended to take the anatomical structures present in one sex, and change them to become more similar to the structures present in the opposite sex.

Let’s talk about sex

In mammals, sex is determined by the presence of a Y chromosome. Individuals with Y chromosomes are genetically male and individuals without Y chromosomes are genetically female. During fetal development, the SRY gene on the Y chromosome directs the gonads to develop into testes instead of ovaries. Testes and ovaries are different in their primary sex characteristics: testes make sperm and ovaries make eggs. Additionally, testes and ovaries produce different levels of sex hormones. Ovaries produce high levels of estrogen and progesterone, and low levels of testosterone, whereas testes produce moderate levels of estrogen and progesterone, and high levels of testosterone.3 These hormones cause secondary sex characteristics to develop throughout the rest of the body. For example, estrogen causes breasts to grow, and testosterone causes body hair to become thicker.

The production of sex hormones is regulated through the hypothalamic-pituitary-gonadal (HPG) axis. Puberty begins when the hypothalamus starts to secrete pulses of the hormone GnRH to the pituitary gland.4 In response, the pituitary secretes hormones (FSH and LH) which travel through the blood and act to increase the production of hormones (including progesterone, estrogen, and testosterone) by the testes and ovaries. These hormones5 act in a feedback loop6 to control the rate of GnRH, FSH, and LH release.

Beginning during fetal development and sharply increasing during puberty, the ovaries or testes produce sex hormones (estrogen for both males and females, and testosterone for males), which regulate important processes such as bone development, muscle growth, and brain development. This results in males and females having biological differences in traits such as basal metabolic rate, grip strength, and maximal respiratory rate.

Notably, male and female brains are different. Even before puberty, sex is evident just from an MRI brain scan. Some traits, such as the structure of a particular region of the hypothalamus known as the sexually dimorphic nucleus, show clear distinctions between males and females. Other traits show more overlap between the sexes, although still with some differences.

Disorders of sex development occur in about 1 in 5,000 people, usually due to genetic mutations. These include:

• Gonads fail to develop. This can happen in both XY and XX individuals for a variety of reasons. No sex hormones are produced, and sex characteristics are mostly female but puberty does not occur.

• SRY deletion: an XY individual develops ovaries and female characteristics.7

• 5α-reductase deficiency: an XY individual develops testes, but does not produce enough of the highly active form of testosterone. Sex characteristics can be ambiguous. However, testosterone production increases during puberty, often leading to the development of male characteristics later in life.

• Androgen insensitivity: an XY individual develops testes which produce testosterone, but the body cannot respond to testosterone and develops female secondary sex characteristics.

• Congenital adrenal hyperplasia: the adrenal glands produce excess testosterone.8 In XX individuals this can result in the development of some male characteristics (with the degree depending on the testosterone levels).

• Kallmann syndrome: the pituitary does not receive GnRH,9 meaning sex hormone levels are low and puberty does not occur.

• Sex chromosome aneuploidies (X0, XXY, etc.) also affect sex development, but male or female characteristics still follow the presence or absence of a Y chromosome.

Intersex individuals are people with biological sex characteristics that are not within the normal male or normal female developmental ranges. This is relatively rare, occurring in approximately 1 in every 2000 to 4500 people.10 Notably, most transgender people are not intersex.

Prevalence and potential causes of transgender identity

Statistics on the prevalence of transgender identity rely on self-identification through surveys, and may be unreliable due to people not wanting to disclose their transgender identity, or due to cisgender people giving spurious responses. Here, I will look at sources which analyze data from the CDC’s Behavioral Risk Factor Surveillance Survey, which is a relatively high-quality yearly phone-based survey from the United States. With that disclaimer out of the way, let’s look at the survey results:

In the United States as of 2017-2019 (the most recent range for which complete data are available for youth ages 13-17), 0.5% of the adult population (~1.3 million) and 1.4% of youth ages 13 to 17 (~0.3 million) identify as transgender. There are slightly more transgender women than transgender men.11 Although rates are roughly similar among different racial groups,12 they vary by geographical location, with a 4 to 5-fold difference among the states with the highest and lowest rates.13 In the 5 years leading up to the 2017-2019 survey period, the overall rates remained stable over time.

Looking at data from more recent years (2020-2023), the prevalence of transgender identity has sharply increased among young adults aged 18-24.14 This increase has largely been driven by individuals of female sex identifying as nonconforming/nonbinary or as transgender men, and the increase has taken place in different states of the USA at similar rates, regardless of whether the states were politically liberal or conservative.

The causes of transgender identity (and its recent increase) are a topic of intense debate. There is not currently a clear answer, and it is likely that multiple factors contribute. Potential causes can be divided into two categories: biological and social.

Genetics are an important component of many human behavioral traits, but mainly due to political difficulties as well as its low prevalence, there have not been any large-scale genome wide association studies of of transgender identity. Smaller studies examining correlations between different types of twins15 have estimated the genetic heritability,16 and found results ranging from 0% to 84% with most studies in the 30-60% range. For comparison, the heritability of homosexual behavior was recently estimated in a large, high-quality study as as 8 to 25%. Overall, I expect there to be some genetic component to transgender identity, but genetics are only one of many contributing factors.

Environmental factors may also play a biological role. These include things like endocrine-disrupting chemicals,17 hormone levels during fetal development, or just random variation in brain development.

In terms of brain structure, pre-treatment transgender individuals typically match their biological sex, but more research is needed in this area to uncover potential associations between brain structure and gender identity. One study from 2009 examined the brain structures of 24 transgender women prior to feminizing hormone treatment, comparing them to 30 control cisgender men and 30 control cisgender women. The transgender women had brains that generally matched their male biological sex. One brain region looked more similar to female brains, but given the relatively small sample size it is hard to draw definitive conclusions from this study. A 2015 imaging study of adolescents (55 transgender boys, 38 transgender girls, 44 cisgender boys, and 52 cisgender girls) found that brain structure largely followed biological sex, with only “subtle deviations” in transgender individuals. Overall, high-quality studies in this area are lacking, especially given that thousands of participants are typically required to draw reliable conclusions about associations between brain structure and traits,

Given that biological parameters (genetics and the developmental environment) have not changed much over the last decade, it seems likely that social causes are behind the recent rise in transgender identity. For example, greater societal acceptance of transgender identity could cause more transgender people to “come out” as transgender.18 Likewise, if transgender identity is viewed as desirable among certain social groups (such as young people), this might induce more of them to identify as transgender, continuing a social trend.19 Or, if cisgender people enforce rigid expectations of gender identity, people who don’t fully conform to them might identify as gender nonbinary. There is not much concrete evidence here, but one or more social causes are likely important.

After looking at the evidence, I think the following combined biological and social model (inspired by the liability threshold model of genetics) may be a good way of thinking about the causes of transgender identity, although this is highly speculative:

1. Every individual has a different innate propensity for gender-nonconforming identity and behavior. This propensity is largely biological. When the propensity is above a certain threshold, the person will identify as transgender.

2. This threshold at which someone identifies as “transgender” or “gender nonbinary” is largely socially determined, and has changed in recent years.

Again, I want to emphasize that the causes of transgender identity are still an area of active research and heated debate. Hopefully scientists will learn more over the coming years about this.

Stay tuned for the next part of this post, covering transgender medical treatments and athletics.

Subscribe now

4

What causes the start of puberty is still an area of research, but it seems that leptin signaling (which relates to body fat percentage) plays a role. The average age of onset of puberty has decreased over the last 150 years and continues to decrease today. Most studies of this have been in girls, since menstruation is a clear-cut marker of female puberty, and the onset of puberty is harder to measure in boys.

The first part of this guide will focus on embryos and fetal development, and the second part will focus on sex development, puberty, and transgender-related issues.

Embryos and in vitro fertilization

Does a human life begin at conception? The answer depends on what we consider to be a human life. However, scientifically speaking, embryos do not have a 1-to-1 correspondence with people.

1. Most human embryos die before developing to a baby. The fraction varies, with the age of the mother’s egg being the most important factor, but even in couples with peak fertility, more than half of naturally conceived embryos die.

2. Up until the developmental stage of gastrulation (discussed below), it’s possible for an embryo to split in two, forming identical twin humans. There have even been rare cases of identical triplets. Therefore an embryo does not correspond to a unique human being.1

3. A pre-implantation embryo has no more sensory awareness than a skin cell or liver cell.

For many infertile couples, in vitro fertilization (IVF) is the only way to have biological children. IVF involves collecting eggs from a woman’s ovaries, adding sperm to fertilize them, and then transferring the resulting embryos (ideally one at a time) to a woman’s uterus2 until a pregnancy is established. The number of embryos created during a cycle of IVF is limited by the number of eggs retrieved, which depends on the woman’s age but is typically about 10-20. This is roughly as many embryos as a couple will produce in 1-2 years of trying to conceive naturally (at a rate of 1 embryo per 28 days).

Timeline of human pregnancy

In medical practice, pregnancy is measured in gestational weeks, the number of weeks counting from the date of the woman’s last menstrual period.3 For a woman with typical fertility, ovulation occurs two weeks into the menstrual cycle, so the time of fertilization is already gestational week two. The earliest sign of pregnancy is typically a missed period.4 Even in ideal circumstances, this is not possible to notice until an additional two weeks (gestational week four). Since many women have irregular menstrual cycles, they might not realize they are pregnant until gestational week five, six, or even later.

Key developmental milestones include:

• Gastrulation happens at gestational week 4–5 (14–21 days post fertilization).5 This is the developmental process that creates different lineages of embryonic cells that will go on to form different organs. After gastrulation it is no longer possible for the embryo to split and form identical twins.6

• Heart formation happens around gestational week 6. The heart is the first organ to form in the fetus, so detection of a fetal heartbeat does not necessarily indicate that the brain or other organs have developed. At this stage the embryo is about 3 mm (roughly the size of a peppercorn). For images see here.

• Brain formation begins around gestational week 7 but the activity of the brain at this stage is still quite limited. The earliest plausible stage at which a fetus could feel pain is gestational week 12 although some researchers have argued this capacity does not develop until gestational week 24. Although it is impossible to ask a fetus whether it is actually experiencing pain, the biological structures required for pain response are definitely established by week 24, and some aspects of these structures begin to form earlier (around week 12).

• Viability of the fetus outside the uterus depends on the level of medical technology. There are current examples of fetuses as young as 21.5 gestational weeks surviving premature birth. However, babies born this premature have a low survival rate, require prolonged stays in a neonatal intensive care unit, and generally have severe long-term issues.

• Full-term birth happens at gestational week 38-40 (anything earlier than week 37 is considered preterm).

• For a more detailed overview of developmental stages (including pictures) I recommend the Carnegie collection.

In the USA as of 2021, 45% of abortions were performed at ≤6 weeks of gestation, 81% were performed at ≤9 weeks of gestation, and nearly all (93.5%) were performed at ≤13 weeks of gestation.

About 15% of all recognized pregnancies end in miscarriage (this statistic excludes early miscarriages that happen before a woman realizes she is pregnant). When a woman has a miscarriage, she may need medical help to remove the fetal tissue from her uterus so that it does not cause complications such as infection.

Ectopic pregnancies can occur when an embryo implants in a location outside of the uterus. This happens in about 2% of pregnancies and requires medical removal of the fetus to prevent life-threatening complications such as bleeding. An ectopic pregnancy cannot result in a live birth.

Final thoughts

Obviously, questions of IVF and abortion are very politically inflammatory. But political opinions don’t change scientific facts – and if legislators or voters don’t understand the facts, they may not realize the consequences of proposed legislation. For example, banning abortion after a fetal heartbeat is detectable is banning it at a much earlier developmental stage than most people realize.

If any politicians (from any party) would like scientific advice on these issues, or have a question about something I haven’t covered here, I’d be happy to help. You can reach me at metacelsus [at] protonmail.com

Share