The fundamental problem of the great majority of so-called “in mice” studies is their use of exactly 1 (one) fully inbred type of mouse — C57BL/6J. Would you trust N=1 clinical trials?
In this particular PNAS study you have to be hard-core reader and rummage around in the Methods section to uncover this awkward detail:
“Mouse strains used were C57BL/6J (Black 6, Jackson Laboratory; JAX stock #000664), JAX stock #003725 (Gabrd−/− mice on a C57BL/6J background…”
No wonder translation fails so often. It is not the mice, it is scientists cutting corners!
Some of us working with mice use many different genotypes of mice to ensure somewhat more robust results that have a better chance to generalize across mouse genomes and perhaps even human populations.
This is not new news. Here is a classic on this topic in Alzheimer’s research published in Neuron in 2019:
One more point: There is a great deal of variation in expression of key genes/proteins mentions in this recent PNAS study, in particular parvalbumin and the GABA receptors.
For example parvalbumin protein levels in the hippocampus of different strains of mice at different ages vary over a 20-fold range. Here are the hard data from GeneNetwork.org;
And GABA receptors also have high levels of variation in hippocampus and other brain regions. Here is a good paper that compares mouse and human GABA receptor variation.
This is comment is needlessly inflammatory and is not informative of how biomedical research is conducted.
First, it’s important to point out to the lay reader that the reason researchers frequently use mice strains in the first place is not “to cut corners”; it as an attempt to manage the incredible variability inherent in biology, especially when one moves to complex systems like living mammals. If the mice in a study have the same genetic background, then it far more likely the effect under study can be observed, and not otherwise buried in the noise as it would if wild mouse strains were used. Using a common mouse strains helps compare studies across different areas of research too. It’s the same reason why so much research is done using HeLa cells. Your post gives the lay reader the impression the researchers were being irresponsible; rather than just using a common strain to demonstrates their point.
Second, you apparently weren't being too much of a "hard core' reader (or least explainer to the lay reader) because if you had you would highlight this bit in the Methods section that list the multiple mouse models used:
1. C57BL/6J
2. Gabrd-/- mice on a C57BL/6J background
3 Pvalb-IRES-Cre mice
4. B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J
5. ApoE4-TR:5xFAD mice
These are not stock C57/Bl6 mice and do carry specific mutations.
Finally, as the second half of your post acknowledges and the paper demonstrates, different strains are being used in AD research. There are entire centers dedicated to developing specific mouse strains, many with humanized genes, to attempt to capture specific mutations known to be present in humans, some on C57 background, but many others that are not. Jackson labs has developed several as has several other prominent groups in the field [1].
My comment was directed more at those criticizing mouse models than this particular paper. Techniques are lovely; results are hopeful. But I need to see this generalized before talk about translation in HN or elsewhere, or yet another discussion about lousy mouse models and the necessity of “in mice” in every title.
From my point of view my comments were needfully and purposefully inflammatory to break the fever.
Hard to believe, but scientists (myself included) can be exceedingly stubborn and resistant to necessary paradigm changes such as incorporating genetic diversity into studies.
The principal reason mouse models are critiqued in HN and at NIH is that they are regarded as translationally ineffective.
Many of my colleagues often say that mice are poor models of human diseases X, Y, and Z.
To be blunt they usually do not really know what they are talking about. What they mean is that B6 is a bad model. But B6 is regarded as THE MOUSE. Is any one human a great model of human traits and diseases? Why do we allow this for experimental work. It truly is N=1.
We also have a community-wide fixation on moving all mutations and transgenes on to the glorious or notorious C57BL/6 strain.
Without background genetic variation we have fragile models that often do not generalize across strains of mice.
“Our current models are mannequins we dress up with sparse data from a few experiments and send down a publication catwalk, praying they don't toddle and trip. What we need are powerful, diverse, and well dressed models that can navigate a landscape, not a catwalk. As a community we have not begun to grapple with this issue, but one certainty is that we will require extraordinarily well structured data sets for humans and for experimental models, to make accurate predictions.”
As to your request that I review the Methods section more carefully — done. And I do not see any errors of commission on my part.
All of the lines that you list above are C57BL/6J (B6) or congenic mice on a nearly pure B6 background.
The first is the reference B6 inbred strain.
#2 is a B6 congenic (B6.xxx)
#3: The Pvalb mouse is at least a five-generation backcross to B6, and therefore also nearly congenic. This line is likely to have been backcrossed for more generations to B6 at JAX; to the point that it is likely to be >95% B6.
#4: The nomenclature of this line clearly states that it is a B6 congenic. The typical minimum number of backcrosses to list as a cogenic is 10, but this is a popular line of mouse and may have been backcrossed for even higher purity.
#5: Finally, the APOE4-5XFAD mice are 97% C57BL/6J.
“…mice homozygous for APOE4, and hemizygous for the 5XFAD transgenes, on a background strain 97% C57Bl/6J and 3% SJL; mice were then inbred between 5xFAD+ and 5xFAD- resulting in littermates E4+/+:FAD+ and E4+/+:FAD−.”
The obvious major caveat is that mice are not people, and that mouse models of what someone thinks are Alzheimer’s-like diseases are not people with Alzheimer’s.
I wish breakthroughs in Alzheimer's were like breakthroughs in battery tech. Batteries are getting dramatically cheaper and denser every year with a clear path for 10+ years of improvements.
Alzheimer's treatments today maybe slow cognitive decline by 3 months and cost a fortune.
The paper itself and the article on it both state it is in mouse models of disease within their titles.
Seems like many here bring up this point for no reason other than to repeat it over and over as a low-effort dismissal. It is exhausting to read these threads.
At the moment I am writing this, there is only a single comment about the actual article.
I do wonder if medicine would do better if we simply skipped the efficacy tests in animals (due to the very low accuracy of animal models of diseases not well understood) and simply did animal safety tests, then moved straight onto human trials.
Obviously the vast majority of treatments would be ineffective - and this in turn should let you design multi-treatment trials, where say a trial of 100 patients are used to test 50 substances, with each person receiving simultaneously tiny doses of ~25 candidate treatments.
Then the patient outcomes are used to identify which of the 50 likely have some effect, and a trial is done of just that substance.
I work in the side of academia that feeds drug leads to pharma and have seen this process.
The big problem aside from time is cost and scale. You might start with either a million compound in silico 10000 compounds in a high throughout enzyme screening program. You might get 500 hits with reasonable affinity which you put those in cell cultures and get 50 compounds. Then you go a small scale mouse study and may get 5-10 decent hits that don't have very obvious tox issues. Then you do a large mouse study of those for efficacy. You will get 1-2 compounds that may be suitable to go into people and there is maybe a 10% chance you get through phase III and onto the market. Usually there is also a step in monkeys before it goes into people.
The main issue here is that at each of those steps the cost scales 10-100x so you really, really need cut cost as much as possible and eliminate non promising candidates. Realistically if a molecule doesn't work in mice sure it could work in humans, but you'd be better off just doing the mouse study so you only need to run one trial instead of 10 and use the savings to invest in other trials.
Human trials are insanely expensive. One of my colleagues works at a company that pays well over $100k for a single primate for preclinical testing. Doing it in people is even more expensive with the massive costs of just organizing and insuring a trial. Comparatively it is very easy and cheap to inject a couple dozen mice with a drug and watch them for 6 months.
Also mixing trial candidates is a really bad idea, CYP liver enzyme reactions are a real thing and are often very different for humans and animals. Doing it with one drug is dangerous enough, but mixing 25 where one may inhibit decomposition of another is a recipie for dead people. You'd also need a very very large number of people (likely thousands) to deconvolute the statistical noise.
I appreciate you listing some of the complexities and issues of human trials. I learned some new information and perspectives that I have not considered before.
However, I am not entirely convinced such trials could still not be efficiently and ethically conducted in some way.
If I can have little "Organ Donor" checkbox on my license, then I do not see why I could not have a little "Alzheimer's Experiment Donor" checkbox either. I'm not saying it would be easy to conduct, and obviously there would have to be more stipulations than I have mentioned.
Alzheimer's is an ultimately fatal disease and ain't a pretty way to go out. What do these sufferers have to lose? You can't kill a deadman.
Rationally what you are saying makes sense, but the problem with progressive diseases like Alzheimer’s is not just defining the point at which quality of life is so poor you might as well throw spaghetti at the wall with the hope of helping society if not the individual, but also getting people and their caretakers to admit they’ve reached that point.
Then there is potential legal liability if someone disputes a doctor’s choice after the fact.
I mean for AD it could perhaps be useful (of course it would not work for most other diseases) but legally that is very dicey because of informed consent. TBH we are not even sure if AD is one even one disease or if it multiple masquerading as one. We aren't even sure if amalyoid beta is a major contributor or cause. Or if it is an effect of the disease. The field is fraught with controversy and even seminal papers have been retracted. But getting back to that lack of understanding of the root cause this is why mice are useful, if we have a reproducible AD like phenotype in a transgenic mouse line that gives us some fundamental understanding of the underlying disease process in a way that likely wouldn't be possibly in a possibly polymorphic human disease population.
You can also dose the mice at the same point in the disease process which is very useful in getting statistical significance for an effect.
I also don't think you'd get enough people and they would be spread out all over the country. Hard to administer your proprietary drug in 100 different nursing homes. Much easier to walk down the mouse facility in the basement
To add to your point, over the least decade animal models are increasingly being seen as models of specific mechanisms, rather than as complete models of a human disease.
To relate this back to the AD field, researchers have developed mouse models to better understand the specific steps in the development of a form of angiopathy seen in humans, cerebral amyloid angiopathy; that is, amyloid plaques that develop on blood vessels. This is being done in the hope that these models will elucidate the basic biology of CAA, and thus help researchers make more informed no/go decisions about CAA therapies in humans.
This is really good point. People underestimate he degree failed trials have value in understanding the basic science behind diseases, especially mechanistically complex ones.
But you are dying already. Vanishing to be a stumbling around hull, foreign to the world. If not medicated, you would be in constant panic of being lost in a foreign place. Fast death with a chance of healing seems preferable if chosen out of ones own free will before it errodes away?
Tiny doses can generally avoid this. When you eat a banana, there are probably over 1 million different chemicals in there. There are probably a few molecules that exist nowhere else on earth.
Yet nobody is worried about the safety of a banana, because the concentration of the really rare chemicals is low enough to not matter.
The efficacy of most new drugs is barely feasible to detect at a small trial size, of the best guess for effective dose, comparing against only one other option (or nothing).
Trying to microdose and test many things at once is going to make this signal vanish to nothing.
The logical way to do it is to have doses starting really tiny, then doubling every ~week.
Every week, patients are given a health survey, and if any effect is obvious in the results data (ie. patients given drug 43 have, on average, a 2 C increased core body temperature), then the dosage of that specific drug is either held fixed (since it is clear we have reached the 'has an effect' dosage), or removed entirely from the trial (if whatever the effect is is deemed likely to hide more valuable yet smaller data from other drugs)
Then your get dropout problems and by the time there is an effect you don't have the statistical power to detect it. You need 3-6 months on an AD, psychiatic, metabolic, or autoimmune drug to see a Real effect so that is how long you must wait to ramp the dose. It could take years to go from a conservative guaranteed nontoxic dose to an effective one. But in that time most patientss will have gotten tired of waiting and switched drugs, moved to a new city and dropped out of the trial, or die. Sure it could work for some things like antimicrobials or erectile dysfunction drugs that have a rapid response, but that isn't where most of the drug development is. It is in those categories I listed.
In addition to what the person has already said: An "obvious" effect is harder to spot than you think. Asking for statistical power to spot an effect in a week when you're trying a dozen compounds at once-- you'd need millions enrolled in this study, and they'd need to be very patient.
It does some anomalous that some states are moving toward assisted
dying which (I don't personally agree with for complex reasons), which
is essentially a very humane stance, and yet elderly people or those
who are terminally ill cannot, it seems, waive their "safety right"
under law and consent to be trial subjects for experimental drugs with
high risks.
Although it does seem like many companies are still not allowing this. I don't know much on the topic and I'm not sure why (financial reasons perhaps?) they'd refuse, but I've always found it absurd that we can poison ourselves to death (literally) with all sorts of substances legally, but experimentation with life saving methods are restricted. I've seen some very strange arguments against the experimentation like "well they could die sooner!" - yeah, but that's up to the person to decide, just like all other decisions they've made preceding it. A patient has the right not to choose any treatment at all, that could definitely make the die sooner. Why is this where the line is drawn for some people? It seems arbitrary.
The research teams want to get their drug approved so it can help many people. This means proving it is safe and effective.
Unfortunately, using it in a non-study context risks a whole lot of confusion-- if the marginal, near-death case that wouldn't qualify for your trial dies in a weird way, what do you do?
Not to mention that prior to approval, there are no economies of scale in production.
I wasn't aware of this progress. Thanks for the link. It's definitely
an ethical hot potato. My concern would always be around pressure, and
being sure that a choice really is a choice.
In Canada there's assisted dying for people with mental illness: perfectly physically healthy people, who might even decide to live a fair bit longer if they felt they were contributing to saving lives by allowing testing.
I'm not sure what you're even suggesting here. First of all, according to the canadian government[1], euthanasia is not currently permitted in cases only involving mental illness. It's been delayed twice already, currently until 2027, and may be delayed further or scrapped entirely in the future for all we know.
Second, what would they be testing? Drugs for their condition? That wouldn't apply to the type of situation necessary for euthanasia under Canadian law[2]:
>To be considered as having a grievous and irremediable medical condition, you must meet all of the following criteria. You must:
1. have a serious illness, disease or disability
2. be in an advanced state of decline that cannot be reversed
3. experience unbearable physical or mental suffering from your illness, disease, disability or state of decline that cannot be relieved under conditions that you consider acceptable
If there was an experimental treatment that might actually help them, these criteria would not apply.
If you're suggesting safety testing of drugs for other illnesses, that's extremely unethical and I seriously doubt the medical community would be willing to do it, even with consent. And I seriously doubt a number of people large enough to give meaningful results would consent in the first place. It's also of dubious usefulness because any treatment intended for disease X might have its safety profile and risk/benefit analysis altered by the presence or non-presence of X.
If it's a big enough effect size you wouldn't need many people taking it, at least for something like Alzheimers. If it stops/reverses it in just a few people out of 10 you'd know to scale up.
I'd take whatever the hell you'd give me even for a .00001% chance of it working at a 25% chance of it screwing me up further if I had Alzheimers as well.
Stochastic effects in small samples could lead us to scale up treatments that only work for those specific 10 people, potentially wasting resources and raising false hopes for the broader Alzheimer's population...
Mice are excellent models for human disease due to their genetic similarity, with 98% of human genes having mouse counterparts. Their comparable reproductive and nervous systems, along with susceptibility to similar diseases, make them valuable research subjects.
DNA manipulation in mice, through breeding or gene editing techniques like CRISPR-Cas, allows researchers to study specific genes and diseases. Using mice is faster than conducting initial studies on humans, allowing for rapid testing of hypotheses and treatments before moving to human trials. Mice also have much shorter lifespans and faster reproductive cycles, enabling researchers to study diseases and genetic effects across multiple generations in a fraction of the time required for human studies.
Maybe if OP would like to sacrifice their children and grand children... sounds a bit insane doesn't it?
Also you can cut them open when you are done. Can't do that to people. Sometimes you need biochemical tests and imaging very quickly post mortem to avoid changes in protein postranslational modifications. This is particularly important in AD studies. We needed to kill the mice and have their hippocampi in a grinder with phosphatase inhitor in under 10 minutes.
I mean, I don't think many people with Alzheimers are having kids. And while mice are genetically similar to us the 'Alzheimers' they're made to have isn't the same disease so we might have been chasing our tail for the last few decades.
Yes but many people that have had kids get Alzheimer's. AD isnt aggressively selected against because it is only deleterious past reproductive age.
It's true but AD is likely not one disease but is many different mutations causing a family of shared symptoms which makes it hard to study. It's often better to gain a mechanistic understanding in a reductive model because you'd not have the statistical power in a mixed population to see anything at all.
Lost my mum to that in the middle of the pandemic. Long mental decline including hemispatial neglect*, but physically (mostly) fine right up until the last few days in the care home, when she forgot about drinking — though we had to work that out much later and from the autopsy as nobody around her could reasonably have watched her 24/7 to count glasses of water not consumed.
* such a bizarre thing to witness: her sitting at the dinner table, the food eaten on only one side of her plate while the other side was full, and her insisting the plate was empty until one of us rotated it 180°.
Alzheon's ALZ-801 pill should be approved within a year and should be basically an Alzheimer cure for 15% (current phase 3 trial) and later 2 thirds of people.
I've recently learned about this and was surprised it's somewhat under the radar. Am I missing something?
oops. shouldn't try 'research' on my phone. thanks.
Incidentally, I'm sure it used to be possible to google search using InChI. Even using the InChI-Key now doesn't work super well it seems. I have to quote it to get this PubMed link:
Shouldn't be too bad. My gut would be toa naphthalene acetic acid ester, treat with LDA, add (4-pyridinyl)acetaldehyde to get a diketone by claisen-type condensation, then treat with hydroxylamine then reflux with hydrazine to form the pyrazole not sure how you would get the n-oxide but there is likely a selective oxidant available.
The n oxide kind of scares me a bit. Seems like it could have some interactions with liver enzymes. Usually you avoid them if you can.
>Auditory, visual or transcranial magnetic stimulation at a frequency of 40 Hz – similar to the frequency of a cat’s purr – worked to dissolve plaques in the brain but again did not show notable cognitive enhancements, Mody said.
my fascination with cats just got even bigger :) One can wonder whether having a cat would prevent/slow the plagues accumulation to start with.
"Gamma oscillations and application of 40‐Hz audiovisual stimulation to improve brain function"
"Currently, 40‐Hz audiovisual stimulation was demonstrated to affect synaptic plasticity and modify the connection status of related brain networks in animal experiments and clinical trials."
I love that “in mice” is in the title (both on HN and in the original article). It seems like a promising result and I really appreciate them not over hyping it.
"Wei et al., have studied the subunit composition of γ-aminobutyric acid type A receptors responsible for the tonic inhibition of parvalbumin positive interneurons and identified a small molecule (DDL-920) as a potent, efficacious, and selective negative allosteric modulator of these receptors."
I the term small molecule has a specific meaning in biopharma and the biochemical formula is almost certainly not going to reveal much information. Might as well just use the convention of the compound ID.
Related, in the supplemental data the compound's reference is Ref XX. Must have slipped by the reviewers.
I'm not entirely sure I understand your comment, but "small molecule" is a drug discovery term of art, and includes things with relatively high molecular weights and complex formulae. Anything sub-1000 daltons is generally considered a "small molecule".
With a PhD in Chemistry, I understand what a "small molecule" is. But by looking at the link for a few seconds, I found no further information about the compound. I am not a big fan of "secret sauce". Maybe I missed it in the paper.
In this particular PNAS study you have to be hard-core reader and rummage around in the Methods section to uncover this awkward detail:
“Mouse strains used were C57BL/6J (Black 6, Jackson Laboratory; JAX stock #000664), JAX stock #003725 (Gabrd−/− mice on a C57BL/6J background…”
No wonder translation fails so often. It is not the mice, it is scientists cutting corners!
Some of us working with mice use many different genotypes of mice to ensure somewhat more robust results that have a better chance to generalize across mouse genomes and perhaps even human populations.
This is not new news. Here is a classic on this topic in Alzheimer’s research published in Neuron in 2019:
https://pubmed.ncbi.nlm.nih.gov/30595332/
One more point: There is a great deal of variation in expression of key genes/proteins mentions in this recent PNAS study, in particular parvalbumin and the GABA receptors.
For example parvalbumin protein levels in the hippocampus of different strains of mice at different ages vary over a 20-fold range. Here are the hard data from GeneNetwork.org;
https://genenetwork.org/show_trait?trait_id=61839_VFHILDKDKS...
And GABA receptors also have high levels of variation in hippocampus and other brain regions. Here is a good paper that compares mouse and human GABA receptor variation.
https://pubmed.ncbi.nlm.nih.gov/22506031/