Summary

Measuring the quality of brain preservation, both in experimental settings and in each case, is an essential part of doing our best to preserve information in the brain. Without evaluating the procedures in empirical tests, it is too easy to throw up one’s hands and say that “our friends from the future” will solve everything.

There is debate about what type of metric is best to optimize for during preservation: structural metrics or viability/electrophysiological/functional metrics. While there are trade-offs to each approach, I prefer to focus on structural metrics because they are simpler, seem likely to be sufficient, and are possible to measure today.

Next, I address the practical question of how to assess structural preservation quality. As opposed to the more theoretical discussions in previous essays, in this essay I give my current best guesses about what structural features to preserve and how we can most effectively measure the quality of their preservation.

I focus on measuring cell membrane morphology because it is a potentially fragile structure, it is straightforward to measure, it is a proxy for the preservation of other gel-like structures in the brain, and it seems to be a high-information aspect of the biomolecule-annotated connectome.

A multimodal approach using microscopy data and biophysical reasoning, while also considering practical factors, is in my opinion the best way to evaluate different possible brain preservation methods. I explain my reasoning through a discussion of the achievements and limitations of the Brain Preservation Foundation prize, which was proposed in 2010 and finally won in 2018.

The no feedback problem

Since it was first proposed in the 1960s, one of the major debates about cryonics is how good it must be before it is offered to the public. More people are on board with the idea of long-term suspended animation: applying a technique known to be reversible at the time of people’s legal death with the goal of reversing the preservation when the cause of their death can be fixed. But cryonics, and brain preservation more generally, is not long-term suspended animation. By definition, we don’t know whether it is reversible. This makes the threshold for when it should be offered much less clear.

The first proponents of cryonics in the early 1960s, Ettinger and Cooper, had read the experimental data from cryobiology studies suggesting that cryopreservation made biophysical sense (Ettinger, 1964) (Duhring, 1962). Ettinger’s The Prospect of Immortality describes how the cryoprotectants glycerol and ethylene glycol can allow small organisms and cells to survive cooling to freezing temperatures. The book also describes evidence that liquid nitrogen temperature storage would likely allow maintenance for centuries. For Ettinger and the first cryonicists, despite the problems and likely damage that would occur, biophysical reasoning based on indirect experimental data was enough to start preserving people via cold temperatures right away. This position was known as “freeze now” (Shoffstall, 2016).

Ideally, prior to offering it to the public, there would be more direct evidence suggesting that a cryonics or brain preservation procedure preserves enough structure to make it plausible to allow for revival with long-term memories intact. Ideally, we would have evidence based on prospective human studies. But that data would take years to decades to gather and require a significant amount of funding that – historically – has not been available. And in the meantime, people who want preservation would legally die and not be able to access it. So, for many people, it seems reasonable to compromise on ideal research standards and allow people to offer the best methods that are available now. From a compassionate use perspective, there is a clear argument that people deserve the right to choose the best preservation methods available at the time of their legal deaths.

But this compromise causes a big problem, which Mike Darwin has called the “no feedback problem.” As an analogy, Darwin notes that if you go to a hospital for surgery, or bring your car to a mechanic, you and your family can get a reasonable sense of whether the procedure was successful. The knee replacement improves the pain from osteoarthritis or it doesn’t. The car is fixed or it isn’t.

In cryonics, there isn’t that same type of obvious feedback. This was especially true in the early days of cryonics in the late 1960s, when cryonics bootstrapped from nothing. At that time, getting people actually stored in liquid nitrogen dewars and keeping them there was a hard enough problem that there were effectively no metrics on preservation quality. People had little to no idea about how the procedures went and what the condition of the brain was. This caused all sorts of problems.

One example of the problems arising from the relative lack of metrics and feedback in cryonics is macroscale fractures. In 1983, multiple people who had been preserved were converted from whole body preservation to head-only preservation due to insufficient funding and the need to transfer them from one cryonics organization to another. This allowed Mike Darwin and colleagues to perform the first gross examination of the actual effects of human cryopreservation. They found extensive macroscale fractures throughout the tissues of the bodies. Although the team did not examine the brains, they reasoned that there was no reason for the brain to be spared from these fractures.

In hindsight, it is not surprising that fractures could occur, because this is a common phenomenon in cryobiology that occurs as a result of thermomechanical stress. That it took so long to anticipate or identify this basic type of severe damage is a sign of how little research – theoretical or experimental – had been done into preservation quality.

Thermomechanical stress fractures can occur due to differences in temperature in two locations of the material during cooling; this is a model of thermal stress simulation from (Li et al., 2017)

Another problem that arises from the “preserve now, research whether the method is reasonably likely to allow for revival later” compromise is that it creates incentives against doing methods research.

First, methods research becomes unnecessary. This is certainly true in pharmacologic research, where most research is performed prior to regulatory approval – once something is approved, there is much less of an incentive to do it.

Second, doing methods research on preservation procedures that have already been applied becomes scary. You might discover something suggesting that the preservations you have done thus far have been flawed. What if someone you loved was preserved with that method? This causes psychological resistance to empirical testing.

Third, from a cynical perspective, if members are already willing to pay for the procedure in the absence of research, doing methods research that might invalidate the procedure risks the loss of one’s income stream.

All of these incentives contribute to lock-in of whatever preservation methods were initially chosen. This happens in conventional medicine all the time: if a type of medical management or treatment starts to be performed before it is known whether it actually works, it can become entrenched even if it is unclear whether or not it is effective.

Part of the problem is the double-edged sword of hope. Hope can be profoundly motivating and it is the basis of most of the interest in brain preservation. But the desire to maintain hope can prevent forthright inquiry into the quality of preservation and long-term protection.

Surrogate outcomes for brain preservation

Right now cryonics is able to operate in many countries because of what many consider a wonderful property of a liberal society, which is that most things are considered legal by default unless the government says they are not. Brain preservation hasn’t been formally approved by the US FDA as a medical product so it is not a type of medicine but rather an anatomical donation to science, usually operating under the Uniform Anatomical Gift Act.

But let’s imagine that investigators were to run an actual human trial to test one or more types of preservation procedure and present the results to the FDA for approval as a medical procedure. The conventional outcome metric – revival – would be clearly impossible; otherwise it would be a trial of long-term suspended animation, not brain preservation.

An alternative metric that the FDA uses in the circumstance that the clinical outcome metric would take too long to adjudicate is called a surrogate outcome. This is a biomarker, such as a molecular, histologic, or radiographic measure.

Usually, a biomarker needs to be validated in separate studies to predict the actual clinical outcome metric before it can be accepted for use in such a trial. But in some circumstances, a biomarker may be accepted as a surrogate outcome for the approval of a medical product even if it hasn’t been validated to actually predict clinical outcomes. This is in the case that the biomarker is considered “reasonably likely to predict a clinical benefit,” which means that it is supported by sound mechanistic or epidemiological reasoning. An approval can be made on this basis if data is collected in the post-approval setting to ensure that the reasonably likely surrogate endpoint actually predicts the clinical outcome.

If we stretch the timescale of this “post-approval” setting to decades or centuries, then arguably one or more surrogate outcomes of preservation quality could be used for FDA approval of brain preservation. This seems quite unlikely to happen anytime soon, in part because of societal prejudices, although one could imagine that it might happen in a few decades.

For now, this exercise is most beneficial in helping us to recognize that, in the absence of clinical outcomes, surrogate biomarkers are absolutely critical in brain preservation. Arguably, by its own standards, this is how society ought to be judging brain preservation. Surrogate biomarkers of preservation quality need to be carefully chosen and diligently measured.

What surrogate outcome metrics would be required for society’s stamp of approval is a separate question from what standards individuals would require before they think that brain preservation is a worthwhile endeavor for themselves and/or their families.

What type of preservation quality metrics are we actually talking about? Let’s get a sense of them so that the discussion is more grounded. We can distinguish four types of metrics: performance metrics, gross neuroanatomic metrics, neuroimaging metrics, and tissue biospecimen metrics.

1. Performance metrics.

Performance metrics are measurable aspects of the preservation process. One obvious performance metric is a quantification of the amount of time the brain spent in various conditions during which it is likely to decompose, as well as the estimated temperatures during those time periods. This includes the amount of time with poor or no brain perfusion during the agonal period and the postmortem interval, weighted by the amount of time at cold temperatures – ideally, this time would be minimized, but it’s important to know either way.

Measuring the time period of decomposition is essential to improve outcomes. For example, in stroke care, there is a “door to needle” time that quantifies the amount of time it takes from when a patient enters the hospital (“door”) to the start of the infusion of tPA (“needle”) (Man et al., 2020). By tracking this, their goal is to reduce the time it takes to administer this therapy, because this leads to better outcomes (Saver, 2006).

Performance metrics can also be data gathered during the preservation procedure. Obviously, this will vary significantly based on the type of preservation procedure. For cryopreservation procedures, it includes perfusion outcomes such as the color of the venous effluent. It also includes the cooling curve, which can indicate temperature changes suggestive of ice nucleation (Tan et al., 2021).

Model of the time vs temperature curve during the cooling of water; (Tan et al., 2021)

Performance metrics are relatively easy and important to measure. But they don’t tell us all that much about actual brain preservation quality, and shouldn’t be used as surrogate outcomes by themselves.

2. Gross neuroanatomic metrics.

In many preservation procedures, such as those that take the brain out of the skull, create a skull window, or use a Burr hole, it is possible to view at least part of the brain tissue. If the brain can be visualized, then pictures or videos should be taken. The stiffness can potentially be measured via minimally invasive tactile sensors. In this way, proxies for the amount of decomposition and preservation quality, such as color changes, can be assessed.

Skull window creation to visualize a mouse brain; (Guo et al., 2017)

Another aspect of the gross anatomic condition of the brain that can be measured indirectly is the presence of fractures. It seems that this can be measured in part by acoustic measurements during cooling. The cryonics organization Alcor uses a device called the “crackphone” to measure fracture events, although it is unclear if it has been validated. While this is likely not specific to the brain, if there are fractures anywhere in the body, it is reasonable to assume that there are also fractures in the brain.

As with performance metrics, gross neuroanatomic metrics are important, but are not very dispositive of the preservation quality of the biomolecule-annotated connectome. So they could not be used as good surrogate outcome metrics by themselves.

3. Neuroimaging metrics.

In 2011, Alcor began to use CT scans of brain tissue to assess preservation quality. This was accidentally noticed to be possible in the course of a different experiment to measure fractures. Because of logistical considerations related to scanning the brain inside of a liquid nitrogen dewar, as far as I know, it has currently only been reported on people who opted for neuropreservation (i.e. head-only preservation).

Frozen brain tissue has a lower density on CT because ice is less dense than water (Sugimoto et al., 2016). Tissue cryopreserved with the Alcor’s cryoprotectant has a higher density on CT because of its sulfur content. This allows very important feedback about the degree to which the brain has actually been cryoprotected and whether ice has formed.

A CT scan can also be used for other measures of brain tissue quality. For example, it can measure the amount of brain shrinkage, which is an almost inevitable consequence of current methods that use cryopreservation by vitrification.

In my view, neuroimaging is a helpful surrogate outcome metric, because it actually measures brain tissue, but it is not sufficient on its own, because it cannot visualize the biomolecule-annotated connectome.

4. Tissue biospecimen metrics.

Ultimately, engrams are not encoded in the morphologies of macroanatomic features such as the size of the cerebral ventricles. Instead, engrams seem to be encoded by microanatomical structures that make up the biomolecule-annotated connectome, such as synapses and dendritic spines. Measuring these microanatomic features requires examination of tissue biospecimens. As a result, examination of tissue biospecimens is clearly the gold standard for surrogate outcomes.

There are two major potential sources of brain tissue biospecimens: whole brains that are not preserved with the long-term goal of revival and small biopsy samples from those that are.

When new methods are first being developed, it clearly makes the most sense to test these methods on brains where there is no intention of potential revival. These brain samples could either be from animal studies or from people who want to donate their brains to science but do not want to be potentially revived themselves. Because these whole brains can be fully dissected and evaluated, this source should make up the lion’s share of the information about the expected microscopic outcomes of a brain preservation procedure.

The other option is to perform a very small biopsy from the brain or upper spinal cord of people who are preserving their brains and do have the long-term goal of revival. At least in the initial application of brain preservation procedures in practical settings, it seems crucial to measure the tissue quality via biopsy samples during or after the preservation process.

As discussed in a previous essay, performing a small brain biopsy is a common neurosurgical procedure. It is exceptionally unlikely that this biopsy alone would lead to any significant damage to one’s memories. But it is likely that it would help with organizational accountability and methodological improvements that might allow for the consistent preservation of long-term memories.

I don’t want to take the idea of performing brain biopsies lightly. A brain biopsy is widely known as one of the most delicate diagnostic procedures in medicine. Biopsy procedures have often been resisted by cryonics organizations and by many cryonicists opting for preservation. It’s also true that performing central nervous system biopsy procedures will not always be possible depending on the resources of the organization.

Ultimately, if someone doesn’t want to preserve their brain with an organization that performs biopsies, then that is their choice. But when it is possible, performing tissue biopsies would allow stakeholders to get a much higher level of direct feedback about the microanatomic quality of the brain preservation procedure.

The viability vs structure debate

We have established that revival is not an appropriate outcome metric for brain preservation, because brain preservation is by definition not long-term suspended animation.

What should be the surrogate outcome metric, then? This has been a hotly debated topic over the years. There are two major camps: those who think it is best to optimize procedures for viability preservation metrics and those who think it is best to optimize for structural preservation metrics.

Viability or structure are, of course, not themselves brain preservation methods or even metrics. They are categories of metrics. The debate is over whether we should research and/or use brain preservation methods that optimize for the category of viability or structural metrics.

Let’s define these terms. First, what is viability? It can be a slippery term.

For example, consider fetal viability. This was defined in Roe v. Wade as the “interim point at which the fetus becomes … potentially able to live outside the mother’s womb, albeit with artificial aid.” But this clearly depends upon factors that affect the quality of medical care for that fetus, including the level of available medical technology, which is likely to change over time and will differ in different areas of the world. Randomness in outcomes due to our limited understanding of biology also plays a role in fetal viability, as it does in all areas of medicine. The “potential” point of fetal viability could be defined as the point when all fetuses will be able to survive outside of the uterus, or when fifty percent will, or when one fetus has been shown to survive, or when no fetus has yet survived, but a statistical model based on the available data suggests that a fetus has greater than a probability ε (epsilon) of surviving, where ε is some extremely small number. It’s hard to know where to draw this line.

Although the topic is much less controversial, similar challenges are found in discussions of viability following cryopreservation. For example, one definition of cell viability is whether cells can sustainably divide in cell culture. But the cell division metric in cryopreservation is tricky for a reason that Goetz and Goetz pointed out in 1938: this type of cell viability experiment only tells us whether at least one cell has survived the process (Alexander Goetz et al., 1938). But the survival of one cell could be due to random factors that don’t affect most of the cells in the tissue. Our question in brain preservation is whether all or nearly all of the cells are viable following the procedure.

Another problem is that we might call cells treated with certain chemicals “non-viable” today, correctly, because they would not be able to divide and multiply in cell culture. However, this doesn’t mean that technology won’t advance in the future to render them able to divide and thus viable. Thus, viability metrics are dependent on our current understanding and technology. This potential for improvements in future technology is the hope that underlies all of brain preservation.

Instead of cell division, contemporary metrics for viability tend to assess other functional outcomes, such as the presence of active metabolism in cells or electrophysiological responses in the brain tissue. The most useful viability metric relevant for cognitive functions seems to be retaining global electrophysiology functions across the brain, such as those seen on electroencephalography (EEG).

Structural metrics, on the other hand, are easier to define. We can define a structural metric as an observable, static property of brain tissue, for example as seen under a microscope or using a biomolecule profiling method.

Before we go further, I should say that another approach is to optimize for both structural and viability methods at once. However, there are deep methodologic trade-offs between optimizing for these categories of metrics that will be explained in later essays, so I do think that one must choose given our current technology.

Arguments for focusing on the preservation of viability metrics

First, we already know that preservation methods which retain measures of viability, such as vitrification by cryopreservation, can be reversible on small biospecimens via simple rewarming. We can use cryopreservation methods to preserve and revive human cells, human tissues, and small animals. Vitrification by cryopreservation has already been shown to preserve a type of long-term memory in C. elegans (Vita-More et al., 2015). There are many people walking around today who were previously cryopreserved as embryos.

There is a sense in which maybe all that needs to be done is to scale the existing cryopreservation procedures to the size of the human brain or whole body. This seems like an engineering problem that may not be all too difficult to solve.

Second, functional viability tests such as electrophysiology have the potential to “screen off” the need to guess about what the structural components of meaningful neural activity are. In other words, if meaningful neural activity can be directed shown to be present on reanimated brain tissue, then we know that this must be retained by the brain preservation procedure. This could potentially render several of the earlier essays about the structural correlates of engrams superfluous and avoid the seemingly endless debates about this topic. There would be far fewer unknown unknowns in brain preservation if one were able to demonstrably preserve functional viability metrics such as global electrophysiological patterns.

We wouldn’t even need to understand how a viability-optimizing preservation method works, if it works well enough. In medicine, we are often able to use interventions to improve health outcomes before we understand how those interventions work mechanistically – arguably, this is the more common scenario.

Third, in brain preservation procedures that focus on viability, it is easier to see how to iteratively improve upon those methods towards long-term suspended animation. Because a long-term suspended animation procedure is likely to be much more appealing to society than an uncertain brain preservation procedure, as research advances closer to this point, it seems likely that more societal interest and investment will follow.

Fourth, brain preservation methods focusing on viability are more easily translated to other areas in biomedicine, such as reproductive medicine, organ transplantation, tissue engineering, agriculture, and many other fields. As a result, there is broader societal interest and investment in preservation methods focusing on viability metrics.

Arguments for focusing on the preservation of structural metrics

First, while it would be great if long-term suspended animation were possible, it’s not. Long-term suspended animation advocates often seem to think that it is right around the corner, but people have been saying that since the 1960s (Prehoda, 1969). It’s hard to tell how far away we are. We can’t yet reversibly preserve a single human brain region with functional properties intact, let alone a whole human brain, let alone a whole human body. There’s no guarantee that such a procedure would ever be practical, especially in realistic cases with agonal or postmortem damage.

Second, many commonly used viability metrics are potentially irrelevant to the preservation of long-term memory recall. For example, individual cells could theoretically retain their metabolic activity – a viability metric – even if the structural connectivity patterns between them that seem to be necessary for long-term memory information is lost. The information for memories has much more to do with patterns of connectivity between cells than it does with ribosomes, but the former is not strictly required for cell metabolism while the latter is.

Even if local or global electrophysiology could be produced on reanimated tissue, it would still be hard to tell if engrams are preserved. Anything short of actually reviving an organism with long-term memory recall intact would require guesswork based on our knowledge of neuroscience about whether engrams are still present. And any procedure to accomplish that would meet the criteria of long-term suspended animation. So in the absence of long-term suspended animation, any procedure that optimizes for the preservation of viability metrics is going to be shrouded in uncertainty, just as with procedures that optimize for the preservation of structural metrics.

Third, structural preservation metrics are more straightforward than viability metrics. The idea behind a microscope is pretty simple. You literally look at the tissue and see whether the most likely substrates of engrams look like they normally do.

Fourth, preservation methods optimizing for structural metrics have the potential to be much cheaper. For example, it’s much more plausible to imagine a method that optimizes for structural preservation being compatible with room temperature storage. This could help significantly with financial access to brain preservation.

Summary of arguments for focusing on viability or structural metrics

Consideration when using procedures that optimize for this class of metrics Viability metrics Structural metrics
Difficulty in defining the goal of the procedure High; “viability” is a notoriously slippery concept, and reproducible measurement of viability is challenging Moderate; most of the difficulty is in choosing which structural metrics are best to focus on
Reversible preservation procedures possible on small biospecimens Yes; e.g. in embryos, brain cells, ovarian tissue, and small animals No, methods that best optimize for structural metrics are not reversible with today’s technology even on small biospecimens
Need to debate about what is required to retain the information for valued cognitive functions, such as engrams Today, yes; theoretically, in the future, may be able to avoid this debate, if reversible preservation can be achieved with an excellent proxy of memory recall function intact Yes, unless an actual revival method becomes possible
Ability to iteratively improve the method until long-term suspended animation is possible Potentially, yes; methods preserving viability are much more plausibly on a path towards this, although whether it will ever be possible is still unknown No; much less plausible
Societal incentive to develop procedures aside from brain preservation High; numerous potential applications of related technologies in organ transplantation, tissue engineering, reproductive medicine, etc Low to moderate; helpful for certain niche pathology applications or biology studies, but not as much broader societal interest
Potential to achieve low-cost methods Low; it is difficult to imagine how it could be possible without an expensive procedure and continuous low-temperature storage for the long-term High; it is possible to imagine a procedure compatible with room temperature storage
Example of high-quality outcome metric on brain tissue Retaining global electrophysiologic functions across the brain, such as those seen using electroencephalography (EEG) Retaining nanoscale-level topological arrangements across the brain, such as those seen using volume electron microscopy
Already a procedure to achieve this high-quality outcome on large brains today No Yes; aldehyde-stabilized cryopreservation

Inclusionist perspective

This is an important debate with real consequences. But as with many debates within cryonics, it often ends up so heated that the groups involved forget that they are both tiny compared to general society that at best doesn’t care about cryonics and at worst is trying to impede it. So as an inclusionist, I try to respect both perspectives, and often wish that we could argue about it less. I think both perspectives have merit.

It is often said that brain preservation procedures optimizing for structural preservation metrics would only be compatible with the revival method of whole brain emulation. Because all potential revival methods are highly speculative at this point, it is hard to know anything for sure about this topic, but this does not seem to be true. For example, both pure cryopreservation and aldehyde-based approaches seem to be compatible with the potential revival method of molecular nanotechnology-based repair. In fact, one of the key conceptual innovators in molecular nanotechnology-based repair, Eric Drexler, was among the first to propose an aldehyde-based approach for preservation, in his 1986 book Engines of Creation (Drexler, 1986).

Even though structural preservation metrics are not as good with brain preservation methods that attempt to optimize for viability metrics, some make the argument that they may be good enough. We know there is some leeway given the likely capabilities of future technology to infer the original states. For example, when asked about distorted cell membranes in cryopreserved brain tissue at the conference Biostasis 2020, one researcher pointed out that aldehyde-stabilized cryopreservation had better ultrastructure preservation than pure cryopreservation approaches, but they thought it didn’t matter because there was still sufficient structural information present in the brains preserved with pure cryopreservation approaches. Personally, I do not agree, because I am uncertain about what degree of structural brain preservation is “good enough,” but I can see why others might share this perspective.

Viability or structural preservation in research or practice

I find the arguments for pursuing brain preservation procedures that optimize for viability metrics in a research setting to be reasonable. But in terms of what type of metric we should optimize for when preserving brains in practice today, to me there is no contest. All that we can achieve on human-sized brains today is structural preservation. The quality of outcomes on viability metrics that I have seen with all currently available brain preservation methods are very poor. It seems misguided to me to think about improvements in viability-focused preservation technology – which might be possible in the future – when we are talking about what we can achieve with preservation today.

As a result, I prefer to use structural preservation metrics to compare between current methods for preserving brains. I like that they are straightforward, relatively easier to measure, and possible to accomplish today. So now let’s shift gears and talk about the most useful structural preservation metrics to target.

What structural information should we focus on preserving?

In this section, I provide my current best guesses about what makes the most sense to try to preserve. These are provisional and subject to change based on new information.

The abstract structural features I focus on are:

Synapse shape and myelin shape are widely studied special cases of cell membrane shape. I will consider these separately because there is often a literature on them in particular.

The biomolecule features I focus on are:

For the first five biomolecule features, which are different types of biomolecules, I focus on their composition and location information content. Biomolecular conformation states, as discussed in a previous essay, should be predictable based on composition and location information; however, they may be more liable to be lost in a brain preservation procedure, so they are considered separately.

I consider chromatin separately because it is a combination of biomolecules, including nucleic acids, proteins, small molecules, and lipids (Fernandes et al., 2018). It is also a widely studied structure in cells.

On a lower level, chromatin is primarily made up of histone octamers (which are proteins) and strands of DNA, which are organized into nucleosomes:

Low-level chromatin and nucleosome organization principles; (Morgan, 2007)

On a higher level, chromatin is organized in the nucleus into territories by chromosome, organized into primarily open/transcriptionally active (compartment A) and transcriptionally repressed (compartment B) regions, and further organized into topologically associated domains (TADs) (Bak et al., 2021):