I will be leaving for China this weekend. Travel exigencies may prevent me from posting in the near future. In this regard I have opted for posting today. Hopefully I will be able to summarize my travel experiences in regards to the Chinese government’s autism related initiatives when I come back. For now I would like to talk about a very recent article by the Courchesne group (New England Journal of Medicine, March 2014) that has received a lot of attention in the public media. The research involved quantitating the amount of RNA expression for different markers in the cerebral cortex of 11 children with autism (aged 2 to 15 years) and an equal number of controls. Sampling was extremely limited with only small tissue blocks taken from a couple of sites in 3 different lobes. The results showed columnar-like abnormalities affecting most layers of the cortex, with the most severe abnormalities being noted in layers 4 and 5. According to the authors, as well reviewers commenting on the study, the results suggest some type of defect during brain development.
The small numbers in the series is normal for postmortem studies. However, the extremely limited sampling may prevent generalizations. Although the authors tried to establish whether tissue degradation had occurred, some experienced neuropathologists remain skeptical and believe the same may be artifacts of tissue degradation. In an interview for Wired Magazine, Rovert Hevner, one of the word’s leading neuropathologists whose expertise is in brain development, interpreted the findings as possible artifacts (see: http://www.wired.com/wiredscience/2014/03/disordered-cortex-autism/). Hevner said that patches with missing molecular markers may simply correspond to areas where RNA degraded more quickly than in the surrounding tissue.
Viewers who are familiar with our blog site should already know about the problems faced when handling autopsied tissue. The tissue collection used in the present study should be used cautiously. The following is a note by a distinguished autism researcher voicing her concern when using frozen tissue as in the Courchesne study:
“…it was very disappointing to discover that the majority of the brain samples showed extensive degradation and that no meaningful conclusions could be drawn from the experiments. If we had not decided to perform the autoradiography and the hemalum staining after the Western blot experiments, we would have not been aware that we were working with degraded tissue samples. Several research groups received the same brain samples that we got and because they did not perform brain sections, they did not realize the problem with the tissue quality and went on to publish their findings.” Catalina Betancur and Salah El Mestikawy, Universite Pierre et Marie Curie, Paris France, ATP Report, 2010.
It appears that the Courchesne article may join the group of published studies that Catalina made reference to. It is not only tissue degradation that may offer a confound to the results of this study but also patient comorbidities (e.g., whether they suffered seizures), and the conditions surrounding death of the patient and tissue collection (e.g., the time from death to when the tissue were collected). Although neuropathologists deal with all of these variables when discussing their experiments, these were never considered in the results section of the present study.
An experienced investigator would have never approached the project by using frozen tissue. They would have preferred to err on the side of caution. The present results only offer uncertainties. According to Hevner, histological techniques should have been used in order to identify and characterize any possible cerebral cortex malformation. In effect many people have already done so. Previous studies by Bauman, Bailey, Hutsler and Wiegle have described abnormalities suggestive of migrational defects during brain development. For those interested, description of these developmental abnormalities can be found in previous blogs by the author (e.g., http://bit.ly/1mAyzK3, http://bit.ly/1dXOASh).
Previous studies have found cortical patches of abnormalities in the brains of autistic individuals (Casanova et al., 2010; 2013). These patches were described as focal dysplasias (i.e., areas of malformation that happen during brain development). These studies were done by screening whole brains that had been serially sectioned and using microscopy to analyze attendant cellular changes. In comparison to the Courchesne study previous efforts had much better sampling and resolution. The use of histology techniques also prevented some of the concerns in regards to tissue degradation that are present in the Courchesne study.
Serial coronal sections through the brain of an autistic individual. Outlined patches describe malformed tissue. Microscopic examination of these areas revealed a probable desynchronization in the migration of excitatory and inhibitory cells as they form the cerebral cortex (Casanova et al., 2013)
Maybe the results of the Courchesne study are real. Maybe they are not the result of tissue degradation or other variables that influenced the death of the patients or how their tissues were collected. I am happy that the new study reproduced my own findings (Casanova et al., 2010; 2013). However, the study would have certainly gained from a proper review of the literature. Other authors, including ourselves, have already interpreted similar findings from a neurodevelopmental perspective and offered useful correlations to the clinical symptomatology of autistic patients, e.g. seizures, sensory problems.
References
Casanova MF, El-Baz A, Vanbogaert E, Narahari P, Switala A. Minicolumnar core width by lamina comparisons between autistic subjects and controls. Brain Pathology 20(2): 451-458, 2010.
Casanova MF, El-Baz AS, Kamat SS, Dombroski BA, Khalifa F, Elnakib A, Soliman A, Allison-McNutt A, Switala AE. Focal cortical dysplasias in autism spectrum disorder. Acta Neuropathologica Communications 1:67, 2013. doi:10.1186/2051-5960-1-67
Addendum 6/26/14 : The following was added as a criticism to Courchesne’s article in a SFARI web site called cross talk (for original comments see: http://bit.ly/1pKMj39). The site included comments by 3 distinguished researchers criticizing the study. The attached few paragraphs were posted by a biostatistician in the comments section:
I just read Chow et al. The evidence for their claims are either weak or obfuscated. Many of the reported findings in that paper are not corrected for multiple comparisons (most are at p < 0.05). For example, in Table S2, if one takes 20,000 probes and asks for differential expression at p < 0.05, one expects 1,000 false positives. The authors found 106 probes differentially expressed at that threshold – they are likely all false positives.
For selected analyses, corrections for multiple comparisons are reported using the false discovery rate – the FDR. However, they are reported with varying unconventional thresholds (FDR < 0.1, FDR < 0.27) and sometimes not reported at all. This is a hallmark of reporting only what is necessary to fit preconceived notions. One sees this a lot in the neuroimaging literature.
These issues make the claims in Chow et al. difficult to interpret from a statistical perspective. Additionally, Chow et al. is also based on a very small sample size. Until replication of their data is shown (ideally by an independent research group) and stronger data and analyses are provided, I would caution over interpreting their findings. You may be reading signal from noise.