(see here for the breaking story)

In the letter, the FDA describes side effects that were discovered during Boyd Haley’s animal testing of OSR:

“soiling of the anogenital area, alopecia (hair loss) on the lower trunk, back and legs, a dark substance on lower trunk and anogenital area, abnormalities of the pancreas and l;ymphoid hyperplasia.”

Dr. Haley’s company website (as of 24 June 2010) describes his product as “”a toxicity free, lipid soluble antioxidant dietary supplement” .

Odd that they don’t mention any of the side effects they found in their (very limited) animal studies.

What this shows is that “Little Pharma” can be as duplicitous and profit-driven and disregarding of patient safety as “Big Pharma”; the only sustantive difference being that “Big Pharma” gets much more regulatory scrutiny.

If anyone out there is giving OSR to their children (or taking it themselves), read the FDA warning letter very carefully before you give (or take) another dose.  I would suggest putting any remaining OSR in a safe place, out of reach of children and pets.

Consider seeing a real doctor as soon as possible for a thorough physical examination. Be sure to tell the doctor that you or your child have been taking a potentially toxic chemical.

 Finally, whether or not you or your family members have been or are taking OSR, consider this a warning about the hazards of poorly tested (or untested) and minimally-regulated “supplements”.

Prometheus

On November 17th and November 24th, researchers from the MIND Institute published two papers in the journal Neurotoxicity Research. These two papers were:

Tian Y, Green PG, Stamova B, et al. Correlations of gene expression with blood lead levels in children with autism compared to typically developing controls. (pub online 17 Nov 2009; rec’d 15 Sept 2009; accepted 12 Oct 2009)

and

Stamova B, Green PG, Tian Y, et al. Correlations between gene expression and mercury levels in blood of boys with and without autism. (pub online 24 Nov 2009; rec’d 15 Sept 2009; revised 15 Oct 2009; accepted 10 Nov 2009)

[Note: both author lists are almost perfectly identical. Alphabetically, the authors are: Paul Ashwood, Peter G. Green, Jeffrey P. Gregg, Robin Hansen, Irva Hertz-Picciotto, Isaac N. Pessah, Frank R. Sharp, Boryana Stamova, Yingfang Tian, Judy Van de Water and Xiaowei Yang. The Tian et al paper includes Glen Jickling as an author; the Stamova et al paper includes Jennifer Teng. Otherwise, both articles are written by the same group of people.]

Both studies were done as part of the Childhood Autism Risks from Genetics and Environment (CHARGE) study at the University of California at Davis.

In both studies, blood samples were analysed for lead or mercury and the RNA (in the white blood cells) was isolated and copied using reverse transcriptase to make cDNA copies of the RNA present. This cDNA was hybridized with an Affymetrix Human U133 Plus 2.0 GeneChip microarray to determine the gene expression levels. (If none of this sounds like English to you, hang on – I will try to explain it all in a moment)

The lead study (Tian et al) subjects were 37 children with autism (32 male, 5 female) and 15 typically developing children (11 male, 4 female). Their ages were similar, with mean ages of 44.2 and 41.2 months, respectively.

The mercury study (Stomova et al) had 33 autistic subjects and 51 typically developing control subjects – all male. Their mean ages were 45.3 and 43.3 months, respectively.

Autism diagnoses were confirmed through the use of the ADI-R and ADOS tests (both well-validated diagnostic tests for autism) and the typically developing control subjects were examined to exclude overt behavioral, developmental or autism spectrum disorders.

The Stomova et al study (mercury) makes reference to the fact that the blood tests for mercury and the gene expression data had been collected prior to the study as part of the CHARGE study. The Tian et al study (lead) does not explicitly state this, leaving it an open question whether their data was “fresh” or “canned”. However, that makes little difference.

First, let me discuss the blood lead and mercury levels, since I know some of you have been holding your breath waiting for the answer.

Lead: there was NO significant difference between the two groups (breathe). The mean (+/- SD) blood lead levels were 1.30 +/- 1.01 mcg/dL in the autistic subjects and 1.30 +/- 0.58 mcg/dL in the typically developing subjects.

Mercury: there was NO significant difference between the two groups (breathe). The mean (+/- SD) blood mercury levels were 0.46 +/- 0.73 mcg/L in the autistic subjects and 0.60 +/- 0.82 mcg/L in the typically developing subjects. Among the autistic subjects (n=33), four had blood mercury levels below the detection limit (0.01 mcg/L) and the highest level was 3.0 mcg/L; the typically developing controls (n=51) has one subject with blood mercury below the detection limit and the highest blood mercury level was 4.3 mcg/L.

So much for the “poor excretor” hypothesis, eh?

Those of you who just wanted to see the blood lead and mercury levels can leave now - please gather your coats and books and leave quietly.

Gene expression and microarrays for the Layperson:

[Note to anyone who has experience with microarrays. This is a very simplified description of what microarrays are and how they work.]

Genes and what they do: The human genome has (depending on how you count them) somewhere between 25,000 and 35,000 protein-coding genes. When one of your cells “decides” that it needs more of the protein coded for by a certain gene, a set of signaling molecules (which I’m going to gloss over, since it would take thousands of pages to describe what we know about how this happens) triggers the transcription of that gene into a messenger RNA (mRNA) transcript. This mRNA transcript is like a blueprint for the protein. The mRNA binds to a ribosome and that ribosome “reads” the mRNA and constructs the encoded protein. Simple, right?

Except, of course, that it is never that simple. Depending on the “strength” of the signal to make the mRNA transcript, anywhere from a single copy to thousands of copies of the mRNA might be made. That mRNA has a limited “lifetime” in the cell (otherwise, how would the cell ever stop making that protein) and that “lifetime” can be altered by a variety of cell processes.

On top of that, the protein made from the mRNA transcript has its own “lifetime” – some are long; some are short – and that “lifetime” can be shortened by a number of factors: heat, heavy metals (in some cases), oxidation, etc.

One way to see the impact of various conditions on the transcription of genes – the “gene expression” – is to measure how many copies of the mRNA transcript are in the cell. More copies (generally) mean more gene expression (usually interpreted as a response to signaling pathways in the cell). The gene microarrays – like the Affymetrix Human U133 Plus 2.0 GeneChip microarray – contain short segments of DNA (probes) that are designed to pair up with and bind to specific parts of the cDNA copy of the mRNA transcript from a specific gene (with certain important exceptions, noted later).

Why the cumbersome process of making a cDNA copy of the mRNA? Well, for starters, RNA is notoriously unstable, due (in large part) to the vast number of RNA-digesting enzymes in every cell on the planet (bacteria, archaeal, fungal, human, etc.). The easiest way to overcome this problem is to use reverse transcriptase to make a DNA copy (cDNA) of the mRNA. The cDNA is much more stable and easier to work with as a result. Also, in the process of copying the mRNA into cDNA, you can label the cDNA with fluorescent dyes, radioactive elements or a variety of other handy tags. This allows you to see how much has stuck to the probes on the microarray.

The power of the microarray is the amazing number of genes that can be measured at one go. The Affymetrix GeneChip used in these studies has 54,120 probes corresponding to 38,572 human genes (this includes RNA-only genes and common variants, which is why the number is higher than the total number of human genes). That’s a lot of data to get in one experiment!

Now, those of you who are still awake at this point may have noticed that the GeneChip has more probes than genes. This is for a variety of reasons, including the need to cover common variants in gene sequences. There are also probes that stick to different parts of a single gene (see: Alternative splicing), which provide an internal control.

One of the greatest advantages of microarrays is also their biggest problems – the massive amount of data. We’ve learned from doing microarray studies on bacteria that most of the genes that show a change in expression are not involved in a specific response to the environmental change and are seen in many types of cellular stress. The problem of finding which changes are physiologically relevant can be daunting.

Some limitations of gene expression:

While the ability to measure gene expression (mRNA production) is a marvelous advance in biology, there are some significant limitations that we need to recognise. To begin with, an increased production of a specific mRNA can mean many things: it could mean that the cell needs a higher concentration of the protein; it could mean that the protein is being destroyed, and so more is needed as a replacement. It could also mean that the protein is defective - improperly folded, for instance - or that the mRNA itself is defective and the “blueprint” is making defective proteins that don’t work.

To make things even more complicated, there is another system that regulates the translation of mRNA through small interfering pieces of RNA that bind to and block the translation of the mRNA to protein. When this is happening, the production of mRNA is often unchanged or increased, but the protein production is decreased.

With all of these factors interposed between the production of mRNA and the cellular concentration of the protein it codes for, it can be difficult to say for certain that increased gene expression - as measured by higher levels of mRNA - means that higher amounts of the gene product (a protein, usually) are being produced.

Now, back to the studies.

Back to the studies:

With the blood lead or mercury levels and gene expression data in hand, the authors looked for correlations between them. In both studies, they “transformed” the lead levels to the log(2) (base two logarithm) in order to get a better correlation with gene expression levels (which were probably also log(2) transformed – the papers are a bit unclear on that, but it is pretty much standard).

In the lead study (Tian et al), they looked at the correlation between log(2) lead levels and probe set expression (not the same as gene expression) and found the following:

Probes whose expression correlated with lead levels in autistic subjects: 1829
Probes whose expression correlated with lead levels in typical subjects: 1712
Probes whose expression correlated with lead levels when groups combined: 2172

Unfortunately, with 54,120 probe sets on the microarray, we would expect to see (on average) 2706 probe sets showing correlation as the result of random chance, so these results are no statistically significant.

The same lack of statistical significance dogged the mercury study (Stamova et al), where they looked at genes (instead of probe sets) and found:

Genes whose expression correlated with mercury levels in autistic subjects: 1276
Genes whose expression correlated with mercury levels in typical subjects: 1293
Genes whose expression correlated with mercury levels when groups combined: 1113

With 38,572 “genes” (some are open reading frames), we would expect to see (on average) 1928 genes showing correlation as the result of random chance, so these results are also not statistically significant.

However, the authors were not that easily dissuaded. They looked at genes (or probe sets) that were shared across two or more categories and came up with four lists:

Genes/probes that show correlation only in the autistic subjects
Genes/probes that show correlation only in the typical subjects
Genes/probes that show similar correlation in both groups
Genes/probes that show different correlation in each group

Here, at last, they found statistical significance, even if the physiological significance is still in question. Picking through the lists of genes/probes, the authors made lists of affected metabolic and signaling pathways and proposed possible ways that mercury and lead could contribute to the constellation of signs and symptoms that we call “autism”.

However, there is a bit of a problem.

Actually, more than a bit.

Given that these two papers were written by essentially the same people and that they were submitted to the same publication on the same day (see above), it seems odd that they didn’t combine them to show the similarities between the effects of lead and mercury on gene expression.

Odd, indeed.

Of course, some of the reason might have been what my PhD advisor referred to as the “LPU phenomenon”. LPU (least publishable unit) refers to the minimal amount of data necessary to generate a publishable paper. Some researchers, realizing that tenure committees and granting agencies look at the number of publications rather than the amount of information, will split up a single research project into multiple LPU’s, thereby maximizing the publication impact of a single project.

On the other hand, it could be that the gene lists didn’t correlate too well between the studies and putting them in two separate publications would make that less obvious.

I took a look at the gene lists (published in the supplemental material) and made an interesting finding. Of the 1031 probe sets found “highly significant” in the lead study and the 531 probe sets (the supplemental material gave a list of probe sets) found “highly significant” in the mercury study, the two studies shared only 20.

That seemed like a significant finding to me.

Part of the explanation is that lead and mercury have some different effects on cellular physiology. For example, lead competes with calcium and zinc in certain cellular function, while mercury does not. Although both elements react with sulfhydryl (-SH) groups, mercury is more reactive. And while both elements can lead to oxidative stress, mercury is a more powerful oxidizer than lead.

Still, given their similarities, there should be more of an overlap. In addition, the probe sets that did overlap were primarily in the group of probe sets where correlation to the lead/mercury level was seen only in the typical controls. None of the probe sets that showed significant but different correlation with lead/mercury level were shared (Figure 1).

Figure 1: Probe sets shared between the Tian et al (lead) and Stamova et al (mercury) studies. Click on image to see full size.

In figure 1 (above), the probe sets shared between the Tian et al (lead) and Stamova et al (mercury) studies are listed in their correlation groups. The red lines connect probe sets that were in different groups in the two studies. There were several probe sets that were not associated with known proteins or functional RNA. These are abbreviated in the figure as “ORF’s” - open reading frames - meaning that they have start and stop codons and other structural features of protein-coding genes but aren’t known to produce actual proteins. Seven of the twenty shared probe sets fit this description. In addition, two of the shared probes sets are known by the manufacturer to bind to two or more different genes.

Finally, the shared probe sets that are associated with known genes show no apparent specific connection to the effects that lead and mercury have in common (e.g. reaction with sulfhydryl groups and oxidation). Some of them may be expressed in response to cellular stress, but none of them appear to be specific to dealing with sulfhydryl-reactive metals or oxidation (e.g. metallothionein genes, glutathione reductase, glutathione synthetase, etc.). [People who want to check this themselves should look the genes up at Online Mendelian Inheritance in Man or the National Library of Medicine. Note: some of the gene labels are peculiar to Affymetrix - check their website for further information.]

Having had a lot of experience with microarrays, I can sympathize with the authors’ dilemma. They produce so much data that there simply has to be something significant in all of it. The problem is that their question – such as it is – is too broad. They are looking at data collected from the environment, with no ability to control for numerous variables (or to even know what those variables might be). It may be that there are “hidden variables” somehow related to lead or mercury level that are driving their data. Or it could simply be that they were the victim of a highly unlikely coincidence.

At any rate, their conclusion that the autistic subjects showed a different pattern of gene expression compared to typical controls may be correct, but it only applies to the subjects of these studies and only at the time their blood was drawn. What correlation there might be between lead or mercury levels and gene expression remains to be seen.

It would have been useful if the authors had looked at something that is clearly not thought to be associated with autism - such as red blood cell count or serum sodium concentration - and seen if a pattern of gene expression was associated with that, as well. Since the data are available, it would be relatively easy to do.

The reason for looking at the correlation with what is most likely an unrelated and random fluctuation would be to show how much noise there is in the system they are using. If sodium or red blood cell count showed the same level of correlation between blood level and gene expression (different genes, most likely) as lead and mercury did, we would know that the results of Tian et al and Stamova et al were due to “noisy” data. If, on the other hand, there were far fewer genes whose expression correlated to serum sodium concentration or red blood cell count, then we could be more confident that these studies weren’t simply finding a spurious pattern in “noisy” data.

Let’s look for a moment at what the authors didn’t see or, as I like to call it, “the dog that didn’t bark” (see: Adventure of the Silver Blaze, by Sir Arthur Conan Doyle ).  There are three genes involved in heme synthesis that are very sensitive to lead: aminolevulinic acid synthetase (ALAS), aminolevulinic acid dehydratase (ALAD) and ferrochelatase (FECH). Of these three genes, whose expression would be expected to rise in correlation with blood lead levels, only one (ALAD) showed a correlation, and that only in the typical subjects. And, to make matters even worse, the expression levels showed a negative correlation with blood lead levels (i.e. the expression of ALAD went down as the blood lead levels went up). This, as you might guess, is the opposite of what would be expected.

At this point we also have to address what I think of as the “Elephant in the Drawing Room” problem with both of these studies. The RNA for both studies came from white blood cells (human red blood cells don’t have a nucleus and so don’t have significant levels of RNA). White blood cells - especially those in circulation - are pretty much terminally differentiated. That means - among other things - that large sections of their genome are “switched off” and don’t get transcribed. After all, there is little need for white blood cells to make dendrites or produce collagen and would probably be disastrous if they did. For that reason, terminally differentiated cells shut off large parts of their genome in a way that is - for all normal purposes - permanent.

As a result, the gene expression in white blood cells is quite different from the gene expression in - for instance - neurons.

There’s the elephant.

Of course, much of the metabolic and structural machinery in cells is shared across tissue types. Liver cells, neurons and white blood cells all share the same Kreb’s cycle and make lipids with the same enzymes. However, autism hasn’t been distinguished as a disorder of white blood cells - it is more generally thought to be a disorder of the brain.

Granted, it would be far more difficult to get parents to consent to a brain biopsy than to a blood draw and the IRB would probably have a thing or two to say, as well. However, it seems a bit premature to draw firm conclusions about what it going on in the brain from a small sample of white blood cells.

Conclusion:

These studies are a good trial run at correlating gene expression in autism with environmental exposures. The information gathered should be helpful in designing future, more definitive experiments.

These studies also show the peril of using gene expression microarrays, especially in a community (ecological) setting where there are so many uncontrolled variables - especially with small numbers of subjects. They also show the need to correlate not only with gene expression but also with physiology and tissue type.

It would be interesting to see if a repeat study, using different subjects, would find the same gene expression pattern. If so, that would be a huge step toward validating the findings of these studies. However, without independent confirmation of their findings, the conclusions of these studies remain highly tentative.

Prometheus

I think we’ve all heard that before.

There were many problems with their math, not the least of which was being too lazy to look up the actual numbers, but I realised that this was the tip of a much larger iceberg of innumeracy, especially as it pertains to understanding prevalence and risk.

So, to begin at the beginning, let us start with fractions.

Most of the time, risk or prevalence is expressed as a fraction, although it may not always look like a fraction (e.g. 1 in 100 is the same as the fraction 1/100). And - harkening back to our elementary school days - the two components of a fraction are the numerator (the top number) and the denominator (bottom number).

I bring up these apparently irrelevant mathematical issues because, in the world of risk and prevalence, there are two major types of errors: numerator errors and denominator errors (although, sometimes, there are errors of both).

In the example I mentioned above (the ‘blog comment), the risk of contracting a certain vaccine-preventable disease was calculated (wrongly, I must add) by dividing the number of people in the US who contracted that disease in a year by the population of the US. This was then stated - indirectly - to be the risk of an unvaccinated person contracting the disease.

Perhaps you’ve already noticed the error - it’s a denominator error. Since most people in the US are already immune to this disease (mostly by vaccination), the proper denominator would have been the number of unvaccinated people in the US. Let’s see how this changes the numbers:

If we use measles as an example, there were 140 cases of measles reported in 2008 (still compiling and verifying 2009 reports). At the end of 2008, according to the US Census Bureau, there were 300,459,786 people in the US. If we use the incorrect method from the example, that would give a “risk” of contracting measles of 0.47 per million per year (1 in 2,146,141 per year).

That seems a pretty low risk, doesn’t it? It’s a bit higher than your “risk” of winning the Powerball lottery, but still quite low.

But is it accurate? [Hint: No]

The correct way to calculate your risk of contracting measles is to divide the number of reported cases in the US by the number of vulnerable people in the US. By “vulnerable”, I mean those people who haven’t been vaccinated and haven’t had measles. That is a bit harder number to find.

The CDC’s NIS shows that, in 2008, 92.1% of children ages 19-35 months had received at least one MMR vaccination. Going back as far as 1994, that number seems fairly steady - about 90 - 92%. By the age of school entry, that percentage (in the 2007 - 2008 school year) was up to 94.9%. Even if we assume that this percentage doesn’t change, it would mean that - at most - 5% of the population is vulnerable to measles.

But even that isn’t an accurate number, because people born before the measles vaccine was available (1963) - and even the years immediately after the vaccine was introduced - would have gotten the disease if they weren’t vaccinated (it is highly contagious).

By 1968, the incidence of measles had dropped low enough to assume that anyone born after 1968 who was not vaccinated is not immune. So, that means that 5% of the US population age 41 or less is vulnerable to measles. This estimate compares with the value found by Hutchins et al (2004) for measles immunity in 1999, which supports the estimate.

According to the US Census Bureau, there are about 172 million people in the US age 41 years or less, so that gives us - at most - 8.6 million vulnerable people. Now the risk of contracting measles is 140 divided by 8.6 million or 16.3 per million per year (1 in 61,428).

After calculating the risk of contracting measles, we need to calculate the risk of death or serious complications. Measles has a case-fatality rate of 2 per thousand, so the risk of contracting measles and dying of it is about 0.03 per million per year - in the current situation, where 95% of the population is immune.

Other serious complications of measles include pneumonia (about 6% of cases) and encephalitis (1 per 1000 cases). Adding these to the risk of dying brings the total risk of serious complications to 0.8 per million per year. If we exclude pneumonia as a “serious” complication, the combined risk of contracting measles and having a permanent, life-altering (or life-ending) complication is 0.05 per million per year.

The risk of serious complications (i.e. death or permanent disability)  from the MMR vaccine (discounting the as-yet-undemonstrated “autism connection”) is less than 1 per ten million doses (1 per million allergic reaction, less than 10%  of which are “life threatening” = less than 1 per ten million), which (because the recommendation is two doses) works out to less than 0.2 per million per lifetime. With an average lifespan of 75 years, that works out to less than 0.003 per million per year, so the risk from the disease is over ten times greater than the risk of the vaccine even with 95% of the population immune.

Oh, and by the way - the MMR vaccine protects against three diseases, not just measles. We’ll just ignore that for right now.

And even this approximation doesn’t show the true risk of forgoing  just the measles vaccine (let alone the MMR) because we haven’t considered how having a large immune population prevents spreading and how that has limited the number of measles cases reported.

Measles is transmitted from person-to-person, a single infection provides life-long immunity and it has no non-human reservoir and no known long-term carrier or dormant state. In this respect, it is similar to smallpox, polio, mumps, rubella, and many other vaccine-preventable diseases. If it is not transmitted, the measles virus “dies out”. It doesn’t “hang out” in the environment. That is why measles could be eradicated, just as smallpox was.

Currently (since 2000), measles is not endemic in the US, largely because there aren’t enough susceptible (non-immune) people in close enough contact to keep the virus going. Measles in the US is an imported disease that, until 2008, was rarely transmitted beyond the person importing it and any under-age (i.e. less than 2 years old) or immune-compromised people they came in contact with.

Starting in about 2008, the percentage of immune people in the US had slipped far enough that imported cases were able to spread locally in pockets of non-immune people. The August 22, 2008 edition of Morbidity and Mortality Weekly Report (MMWR) details two outbreaks of measles that occured in the US that year. In both cases, the outbreaks occured within groups that did not vaccinate for religious or philosophical reasons and were home-schooled.

This latter point is worth noting - even though these children did not attend a public or private school, they still contracted measles from one another.

Here is a telling statement from the MMWR report:

The number of measles cases reported during January 1–July 31, 2008, is the highest year-to-date since 1996. This increase was not the result of a greater number of imported cases, but was the result of greater viral transmission after importation into the United States, leading to a greater number of importation-associated cases. These importation-associated cases have occurred largely among school-aged children who were eligible for vaccination but whose parents chose not to have them vaccinated.  [emphasis added]

As the percentage of non-immune people in the country rises, imported measles cases will spread to more people, further raising the risk of infection to non-immune people and increasing the already large benefit to risk ratio of vaccines.

What this shows is that those people who choose to not vaccinate should - at the least - take precautions against associating with other people who don’t vaccinate. This would help reduce their risk of infection to the levels I calculated above.

Perhaps they should wear some sort of lapel pin, similar to what many fraternal organisations (e.g. Masons, Rotarians, etc.) have. Except, of course, that instead of stepping forward and embracing when they see a fellow member (with or without secret handshake), they should immediately turn about and walk briskly in opposite directions, to avoid transmitting vaccine-preventable diseases.

As the events of 2008 showed us, there will not be a gradual increase in measles spread as vaccine coverage declines - there will most likely be an abrupt increase as the percentage of non-immune people (and their proximity to one another) crosses a critical threshold.

And it is important to note that non-immune people are not just the children of parents who choose not to have them vaccinated. They include children too young to be vaccinated and people who are immune-suppressed due to disease, cancer or genetic disorders. They include the elderly, whose immune systems are weaker, and those people who - for one reason or another - did not develop an adequate immune response to vaccination.

Those who choose to not vaccinate and think they are letting others take the risks for them are fooling themselves; they are taking the greater risk - even now.

Prometheus

Non-hyperlinked References:

Hutchins SS, Bellini WJ, Coronado V, et al. Population immunity to measles in the United States, 1999. J. Infec. Dis.. 2004; 189(Suppl 1):S91–7

This morning, when I went to find the Hewitson et al (2009) article in the journal Neurotoxicology - the article that had been the inspiration for my post “A ‘Made for Court’ Study?” - I found that it had been withdrawn (see here).

Given that there has been a recent cry of “Show me the monkeys!” from “the usual suspects” in anticipation of the next episode of Hewitson et al, I find the withdrawal of their first paper intriguing. I can’t help but wonder if this is somehow related to the results they intended to publish in their second paper - the one that “the usual suspects” have been crowing about.

The Pollyanna part of me (a very small part, I assure you) wants to believe that the authors have withdrawn the paper in order to correct its many serious flaws. Of course, this would require not just a simple re-write but a complete redesign of the study and starting again from scratch - rather like “remodeling” a house by tearing it out, foundation and all, filling in the hole and starting over. This may be the case.

Another possibility is that they want to re-do their statistical analyses and conclusion, since their data show not only that they cannot distinguish between thimerosal and the hepatitis B vaccine as a cause of the “neurodevelopmental abnormalities” but also that they can’t actually say that there was any significant difference between the treated and control groups. After all, negative results are results, too.

However, I suspect that the real reason may be that the editors of Neurotoxicology took a long, hard look at the paper and decided that it wasn’t worth publishing, after all. Some small (or large) part of the impetus behind that decision - if that is, indeed, what happened - might be the recent conclusions of the GMC regarding the “anchor” author, Andrew J. Wakefield. That’s not the best reason to withdraw the paper, but it’s better to do the right thing for the wrong reasons than not at all.

Whichever way it turns out, the “Show me the monkeys!” cry is going to sound a little more hollow now that the first article of the series has been withdrawn.

Doubtless, the first act of “the usual suspects” will be to paint this as part of the “massive conspiracy to supress the Truth about autism”. However, Neurotoxicology has been very sympathetic to the “something-in-vaccines-causes-autism” movement, publishing several low-quality studies by people (not necessarily even researchers - see this one) who feel that vaccines somehow cause autism, so it’s a bit of a stretch to start screaming that they are “censoring” autism research now.

We (or, at least, I) don’t know why the article was withdrawn, and it may be for reasons that I’ve not contemplated. But having an article withdrawn after being accepted is never a good thing. Again, I hope that it was withdrawn by the authors because they have read the criticisms about their study and want to re-write it to correct their mistakes. Of course, even though I hope that is the reason, I realise that isn’t the most likely reason. Only time will tell.

 Meanwhile, where are the monkeys?!?

Prometheus

In growing numbers, people are taking their autistic children to “clinics” - in Costa Rica, in Germany, in Russia - to get “stem cell” injections. I put “stem cell” inside inverted commas because it is not entirely clear that what these children are receiving are actual stem cells.

And that might be the “good news” in this post - more about that later.

Stem cells have been in the news a lot, especially the past year, since President Obama cleared the way for embryonic stem cell research. So, today, almost everybody above the age of three has heard of them - but how many people really know what they are and what they can (and can’t) do? Not so many, I think (based on what I’ve heard people say about stem cells).

What are stem cells?

Judging by the many and varied things that the lay press have said about stem cells, you might be forgiven for thinking that they are magical little beings that swim to the site of whatever medical problem exists and fix it - sort of like the “nano-machines” that periodically crop up in science fiction stories. However, sad to say, that isn’t the case.

Stem cells are nothing more than a type of cell that can differentiate (develop into) a different type of cell - sometimes many different types of cells (and can proliferate - divide - indefinitely). Far from being magical semi-sentient beings, they are quite prosaic and exist in your bone marrow, under your skin, in your brain - pretty much everywhere in your body. They range from the humble karatinocyte stem cell of your skin - which can only produce karatinocytes (the outer layer of your skin) - to the omnipotent stem cells present in the first few cell divisions after fertilization, which can each develop into a complete organism (see: “identical twins”).

What has some scientists excited about stem cells is the potential to use certain types of them to treat illnesses and injuries that are currently beyond our abilities. In a few cases, we have already seen these therapies work - in most cases, they remain tanalizingly out of reach.

There are a number of different types and degrees of stem cells, which complicates the discussion considerably. The cells of a zygote (fertilized egg) that is still in its first few cell divisions can each become a complete organism (as mentioned above), but before long (a few hours, in most cases), those cells have differentiated to the point where they can’t make an entire organism, but they can still produce cells of any tissue or organ of the body. Once they have “committed” to going down a particular developmental path, they cannot (usually) go back (without our “help”). Eventually, the differentiation process progresses to the point where the cell is terminally differentiated - it has become a liver cell or a neuron and it will not (again, usually) become anything else.

There are two general features of a terminally differentiated cell: it can only divide a few times - at most - and it cannot generate or develop into a different type of cell (again, in biology, there are always the rare exceptions to this and every other rule).

Unipotent and multipotent stem cells:

In order to deal with cell death due to injury or senescence (”wearing out”), most tissues and organs have a pool of unipotent and multipotent stem cells. Unipotent stem cells - as the name implies - can generate one type of cell (e.g. the keratinocyte stem cell can only make keratinocytes); multipotent stem cells can generate a range of related cell types. A good example of multipotent stem cells are the marrow stem cells, which can generate any of the blood cells - red cells, white cells (all types) and platelets - but cannot make, for instance, neurons or skin cells.

Pluripotent stem cells:

This is the type of stem cell that most of the media “hype” is all about. These stem cells can develop into any cell type from any of the three germ cell layers. These are not found in significant numbers beyond infancy, although there have been a number of studies showing that they do persist (in small numbers) into adulthood.

One of the major breakthroughs in stem cell research - and one that might not have happened this soon without the politically-motivated ban on embryonic stem cell research - has been the ability to take adult cells [Note: in stem cell research, cells become "adults" shortly after birth of the organism.] and “reprogramme” them into pluripotent stem cells. This not only gets us around some rather sticky moral and political controversies, it also gets us around the problem of the immune system. More about that in the next section.

Embryonic vs Adult stem cells: 

The next classification of stem cells refers to their origin. Thus we have embryonic stem cells (ESC) that come from the inner cell mass of an embryo, adult stem cells (generally multipotent stem cells) and induced stem cells (iPSC, iMSC) that are made from either adult stem cells or somatic (terminally differentiated) cells.

How do stem cell therapies work?

Adult stem cells - generally marrow stem cells, since they are easiest to “harvest” - have been used for some time in the treatment of leukemia and lymphoma. They have even been used - with significantly less success - in the treatment of breast cancer and brain cancer. The reason that bone marrow stem cells are so useful is not because they have some magical anti-cancer activity; they simply allow the oncologists to use much higher doses of chemotherapeutic drugs. One of the limiting factors in chemotherapy for cancer is the bone marrow - higher doses run the risk of killing off too much (or all) of the bone marrow stem cells, killing the patient (usually due to infection from low white blood cell count - red cells and platelets can be transfused).

By taking out some of the patient’s own bone marrow stem cells and saving them, they can be re-infused after the chemotherapy has been completed - in essence, they “re-seed” the marrow. This allows them to use much higher chemotherapy doses, which (in some situations) can make the difference between a relapse and a remission.

A similar process is used - experimentally, for now - in the treatment of multiple sclerosis [1]. Multiple sclerosis is an auto-immune disease, where a group of immune cells are reacting to the patient’s own tissues (the myelin covering of their nerves, in this case). Recent advances in cell identification and sorting have allowed researchers to isolate only stem cells from the marrow (and none of the terminally differentiated cells that are causing the problem). After the stem cells are removed, the patient receives a course of chemotherapy (and occasionally radiation) to kill off the immune system, after which the stem cells are re-infused to “re-seed” the marrow with (hopefully) healthy cells. This appears to be somewhat promising in limited trials to date, but it is far from established therapy.

A bit more experimental is the use of stem cells to repair damaged tissues, such as heart muscle, nerves (spinal cord) and brain. To do this you need pluripotent stem cells (or you need to extract the stem cells from the tissue/organ - a technique that hasn’t been developed yet). You can use either embryonic stem cells (ESC) or induced pluripotent stem cells (iPSC). So far, the few clinical trials using stem cells for cardiac disease are either in the early stages or not yet started.

Early on in stem cell research - before the discovery of techniques to induce terminally-differentiated adult cells to become pluripotent stem cells - it was thought that only embryonic stem cells were pluripotent. But studies (and a few clinical trials) using embryonic stem cells ran into problems with the immune system. Embryonic stem cells (unless they were harvested from the patient’s umbilical cord blood or a genetically identical donor) are foreign to the recipient, so there is the problem of rejection so familiar in organ transplants, where the recipient’s immune system attacks the stem cells. If the stem cells are (or differentiate into) immune cells, they can even turn about and attack the recipient’s cells, a phenomenon known as graft vs host disease. Either situation calls for immune suppression, which limits the usefullness of embryonic stem cells.

The advantage of using iPSC’s is that they are (usually) the patient’s own cells, so there is essentially zero chance of rejection or immune reaction. Of course, if the problem is a genetic one, there is probably little point in using the patient’s own cells, since they will carry the same mutation.

Unfortunately, iPSC’s carry some “baggage”, as well - literally. In order to “reprogramme” adult cells to become iPSC’s, certain genes - that have been permanently inactivated in terminally differentiated cells (and even in multipotent stem cells) - need to be “turned on”.  Initially, this was done using lentiviral vectors - retroviruses that had been “engineered” to carry non-inactivated versions of the four critical genes (Oct3/4, Sox2, c-Myc and Klf4) into the cells and insert them into the DNA [2]. This worked very well, but the problem is that lentiviruses are rather….indiscriminate about where they insert themselves, so there is a chance that they will do so in a place that inactivates a critical gene. This is why so many of the lentiviruses are known as oncoviruses (cancer-causing viruses). As you might imagine, this limited the use of iPSC’s to experimental animals.

More recently (2008), a research team has managed to convert embryonic fibroblasts to iPSC’s without using a viral vector, using plasmids [3] and even more recently, another team managed to do it with proteins alone [4]. Both of these techniques are - needless to say - still being refined and are not ready for clinical trials.

So, if anybody is getting “stem cell therapy” today, it is either from their own bone marrow (and will produce only blood cells) or it is from embryonic stem cells (and carries the risk of rejection and/or graft vs host disease).

Risks of stem cell therapy:

The risks of stem cell therapy are hard to quantify because it is difficult to separate the risks of the other parts of the therapy from the risks of the stem cells. This is because most of the patients who have undergone stem cell therapy to date have received bone marrow stem cells (their own or someone else’s) and have also received large doses of chemotherapy and/or radiation, which muddies the water as far as following the risks of stem cell infusions goes. However, there are some “brave maverick doctors” in places like Russia who are injecting embryonic stem cells into the spinal fluid of children with ataxia-telangiectasia (and, apparently, other genetic neurological disorders). The outcome of one of these children was reported in PLoS Medicine:

“In May 2001 at the age of 9 y, in March 2002 at the age of 10y, and in July 2004 at the age of 12 y, he was taken by his parents to be treated in Moscow with repeated transplantation of fetal stem cells.”

Approximately one year after his last stem cell treatment, he was seen in hospital:

“…[he] presented to the Sheba Medical Center in February 2005 with recurrent headaches. On examination he had severe neurological deficits characteristic of AT, affecting mainly his motor functions and making him wheelchair bound.”

Although not explicitly stated in the case report, the stem cell treatments were apparently not working, based on their description of his condition. What they found, however, was worse than “not working”:

“MRI performed in February 2005 to investigate the headaches revealed a right infratentorial lesion slightly compressing the brain stem and another lesion at the cauda equina (Figure 1A and 1B). The lesions grew slowly as evidenced by repeat MRIs in June and July 2006. In September 2006 at the age of 14 y, surgery was performed and a tumor localized at L3–4 level attached to the cauda equina nerve roots was removed. Additional ’satellite’ lesions were identified attached to nerve roots rostral to the main lesions (Figure 2A and 2B).”

In short, this lad had two separate brain and spinal cord tumours. Under the microscope, these tumours were not cancerous, but looked like disorganized neural tissue. When they were tested genetically, the tumours did not match the patient’s genetic markers. They were, in fact, from two separate donors.

 Although this child received embryonic stem cells from two different donors, there is no reason why the same problem couldn’t happen with either autologous embryonic stem cells (i.e. from stored cord blood) or iPSC’s. In fact, one of the defining characteristics of pluripotent stem cells (embryonic, adult or induced) is there ability to form teratomas - tumors containing tissues from all three embronic layers (ectoderm, endoderma and mesoderm).

What about using stem cells in autism?

Part of the problem with using stem cells to treat autism is that we don’t know what we are treating. Despite the enthusiastic promotion of various “theories” about what causes autism, there is no generally agreed upon pathology or “lesion” to treat. Even genetic studies fail to show one single genetic cause of autism, suggesting that what we call “autism” is a number of different disorders with a similar (or not so similar) appearance. Injecting stem cells in the vague hope that they will find the problem and fix it is foolish. Stem cells have no more idea of how to “fix” autism than we do - which is to say, “none”.

The “good news” I referred to above is that, based on the descriptions of what they are doing, the clinics where parents are taking their autistic children for “stem cell therapy” are using - at best - multipotent blood stem cells. The descriptions are more promotional than informational, so it is entirely possible that their “techniques” are yielding no stem cells whatsoever. This is “good” because infusing real pluripotent stem cells into the blood or (worse yet) into the spinal fluid carries the risk of creating tumors without any known (or even suspected) potential for benefit.

In the event that someone invokes the concept of “neuroinflammation” as a reason to try stem cell therapy, I’d like to point out that, to date, effective stem cell treatments for “neuroinflammation” and autoimmunity have involved also giving large doses of cytotoxic drugs to kill the errant immune system cells prior to re-infusing the patient’s stem cells. Explain how that would work without the cytotoxic drugs (or radiation) and you’ll get a Nobel Prize in Medicine.

I fervently hope that none of these clinics are using viral vectors to create or “activate” pluripotent stem cells, as this carries a known risk of carcinogenic transformation. I know that they aren’t using any of the non-viral techniques because they are too new and too complicated. I suspect that they are simply re-infusing the patient’s own blood. And I hope that they are using good sterile technique when they do so.

At best, “stem cell therapy” for autism is offering false hope; at worst….. who knows?

Prometheus

UPDATE: See this article in the Milwaukee Journal Sentinal about the sales techniques used by a stem cell therapy center.

References:

[1] Capello E, Vuolo L, et al. Autologous haematopoietic stem-cell transplantation in multiple sclerosis: benefits and risks. Neurol Sci. (2009); 30(Suppl 2):S175-S177

[2] Takahashi K,Yamanaka S. Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell. (2006);126:663–676

[3] Okita K, Nakagawa M, et al. Generation of mouse induced pluripotent stem cells without viral vectors. Science. (2008); 322:949-953

[4] Zhou H, Wu S, et al. Generation of induced pluripotent stem cells using recombinant proteins. Cell Stem Cell. (2009); 4(5):381-384