Schizophrenia Research Forum - A Catalyst for Creative Thinking

Bone Marrow and Aspirin: New Ingredients in Biological Psychiatry

8 June 2010. Neurons control behavior. Antipsychotic drugs target neurotransmitter systems. These statements are among the most uncontroversial in neuroscience, but two surprising new studies suggest that even these seemingly settled issues may need to be reconsidered. What links them together is the unexpected roles of cells and drugs that we associate with the immune system.

In the May 28 issue of Cell, researchers at the University of Utah show that a pathological grooming behavior in mutant mice that has been proposed as a model of obsessive-compulsive disorder (OCD) is driven by a defect in bone-marrow derived glial cells, not neurons. In the May issue of the Journal of Clinical Psychiatry, Dutch researchers report that psychotic symptoms in schizophrenia patients were reduced when antipsychotic drug treatment was supplemented with aspirin, providing support for the idea that inflammatory processes may play a role in psychosis, a notion that has a large body of research behind it, though no major breakthroughs as yet.

A very well-groomed mouse
The Utah story begins in 2002, when Mario Capecchi and Joy Greer reported that mice with homozygous loss-of-function mutations in the Hoxb8 gene groomed themselves excessively, to the point that they developed large hairless patches and injured their skin; this phenotype had 100 percent penetrance in the mutants (Greer and Capecchi, 2002). (Interestingly, the Hoxb8 mutants also excessively groomed wild-type littermates.)

This was a remarkable finding, because the highly conserved Hox family of genes was primarily associated with establishment of the body plan during development (see, e.g., Lutz et al., 1996). The obvious parallels to trichotillomania, a form of OCD in which excessive pulling or twisting of the hair causes large bald patches on the scalp, led the Capecchi group to continue its analysis of Hoxb8 as a promising animal model of OCD.

In the new Hoxb8 study, a team in the Capecchi lab led by Shau-Kwaun Chen determined that the subset of cells showing Hoxb8 lineage in the adult brain of the mouse mutants were microglia, and that they are primarily found in the cerebral cortex, striatum, olfactory bulb, and brainstem. In newborn mutants, sparse cells of Hoxb8 lineage were seen primarily in the choroid plexus, meninges, and ventricular zone, suggesting that the microglia migrate to the forebrain postnatally. Adult mutant mice had at least 15 percent fewer of these microglia than did wild-type mice (the authors note that limitations in their labeling technique preclude a precise assessment, meaning that the reduction in microglia in Hoxb8 mutants could be higher).

“Although the origin of microglia is still debated,” the team writes, “there is general agreement that at least one subpopulation is of bone marrow origin (i.e., derived from circulating monocytes....)” To test this hypothesis, the group labeled white blood cells, and found that all hematopoietic lineages expressed Hoxb8, a finding that was replicated in bone marrow cells.

Transplanting behavior
However, the most dramatic evidence for the linkage between hematopoietic cells and glial cells of Hoxb8 lineage was seen in bone marrow transplant experiments. When Hoxb8 mutants received transplants of wild-type bone marrow, they gradually stopped their excessive grooming. Their skin injuries healed, and their hair regrew. Of six Hoxb8-mutant animals that received wild-type transplants, “[f]our of them fully recovered and were indistinguishable from wild-type mice.” Conversely, two wild-type mice that received transplants of Hoxb8-mutant bone marrow exhibited the excessive grooming phenotype, though they spent less time grooming than is usually seen in Hoxb8 mutants.

In some ways, these results are as striking as the original description of the Hoxb8 excessive grooming phenotype itself. Though there is increasing appreciation of the diverse roles played by glial cells, particularly astrocytes, in the nervous system (see, e.g., Allen and Barres, 2009), microglia are phagocytes, immune system cells that engulf and degrade debris and pathogens. Other than their pathological roles in neurodegenerative diseases such as Alzheimer’s disease, these are hardly the type of cell that one associates with the regulation of a behavior as specific as grooming.

However, the authors write, “[f]rom an evolutionary perspective it may make perfect sense to couple a behavior such as grooming, whose purpose is to reduce pathogen count[,] with the cellular machinery, the innate and adaptive immune systems, used to eliminate pathogens.”

In terms of mechanism, the authors point to previous research showing that microglia physically interact with synapses, as well as their well-studied release of neuronally active cytokines, and speculate that microglia thereby might regulate the activity of synaptic circuits.

Take two aspirins, lower your PANSS score
In another study that links immune function with behavior, a team at University Medical Center Utrecht in The Netherlands conducted a randomized, double-blind, placebo-controlled trial in which 33 out of 70 patients with schizophrenia spectrum disorder (schizophrenia, schizoaffective disorder, or schizophreniform disorder) were provided with 1,000 mg of aspirin daily in addition to antipsychotic medication for three months.

The study builds upon three previous small trials with another non-steroidal anti-inflammatory drug, the COX-2 inhibitor celecoxib, two of which found symptom responses to the drug.

At the final follow-up, the patients who had received aspirin had mean Positive and Negative Syndrome Scale (PANSS) scores that were lower than those of the patients who had received placebo: in the aspirin-treated group, the total PANSS score was 4.86 lower, and the positive PANSS score was 1.57 points lower. No significant differences were seen in other measures, including negative and general PANSS scores, or performance on a battery of cognitive tests.

The lowest total and positive PANSS scores were seen in patients with the shortest disease duration. However, more relevant to hypotheses that inflammation contributes to psychosis was a correlation between the lowest PANNS scores in these two categories and altered immune function, defined as a shift in the balance of T helper cell levels toward anti-inflammatory cytokines versus pro-inflammatory cytokines (in this case, levels of interleukin-4 [IL-4] vs. those of interferon-γ [IFN-γ]).

“The inflammatory hypothesis is gaining some ground, and there are some interesting genetic findings related to it, including our own early GWAS study implicating a cytokine receptor gene in risk for schizophrenia,” Anil Malhotra of the Zucker Hillside Hospital told SRF (see SRF related news story). Jeffrey Lieberman of Columbia University agrees, telling SRF that “the inflammation model has much relevance to various, and in many cases seemingly unlikely, diseases. In addition, this builds on the prior studies with COX-2 inhibitors [in schizophrenia].”

The authors concede that their sample size was smaller than they had originally intended to enroll, and that a three-month follow-up cannot address questions about long-term use of aspirin in high doses (to avoid aspirin-associated gastric distress in the treatment group, the Dutch researchers provided all subjects with omeprazole, a proton-pump inhibitor). Nonetheless, they write that the study had “the largest sample size and longest time of follow-up to date” in research on the use of anti-inflammatory drugs in schizophrenia, adding that the sample size “agrees with the recommended 40 to 100 patients for drug augmentation studies in schizophrenia.”

New York University’s Daniel Javitt is cautious about the findings, noting that there is “a much smaller difference at three months than at two, and an effect size in the total symptoms of about 0.3 for the LOCF [last-observation-carried forward].” Javitt also questions why data were provided only for IFN-γ and IL-4 when the paper mentions that measures were also made of the monocyte cytokines IL-12 and IL-6. “The most compelling finding is the result that patients with most severe inflammation markers showed best response,” Javitt told SRF. “This would seem to be worth following up and selecting a potentially responsive subgroup.”—Pete Farley.

Chen SK, Tvrdik P, Peden E, Cho S, Wu S, Spangrude G, Capecchi MR. Hematopoietic origin of pathological grooming in Hoxb8 mutant mice. Cell. 2010 May 28;141(5):775-85. Abstract

Laan W, Grobbee DE, Selten JP, Heijnen CJ, Kahn RS, Burger H. Adjuvant aspirin therapy reduces symptoms of schizophrenia spectrum disorders: results from a randomized, double-blind, placebo-controlled trial. J Clin Psychiatry. 2010 May;71(5):520-7. Abstract

Comments on News and Primary Papers
Comment by:  Christopher Pittenger
Submitted 18 June 2010
Posted 22 June 2010
  I recommend the Primary Papers

The recent study from the Capecchi laboratory, in which the excessive grooming phenotype observed in HoxB8 knockout mice (Greer and Capecchi, 2002) was found to be mediated by the absence of HoxB8 in hematopoietically derived cells rather than in neurons, represents a startling and important advance. It comes as a surprise to many in the community—certainly to me—that a phenotype as specific and ethologically relevant as syntactic grooming would be modifiable by a specific alteration in microglia. And yet this is precisely what the new paper shows—and shows with very elegantly designed and performed experiments, which leave little doubt as to the striking conclusion. This study will increase interest in the interaction between immune or inflammatory processes and specific behaviors in a variety of basic and pathological contexts, and this is a salubrious advance in the field.

More vexing is the question of whether or not these mice in general, and the new finding in particular, advance our understanding of any specific neuropsychiatric condition. The mice have been described, in the title of the original paper and in numerous contexts since, as a potential mouse model of obsessive-compulsive disorder (OCD). This is a provocative assertion, and it requires careful consideration.

Certainly the idea of a deep connection between an aberrant immune response in the brain and the symptoms of OCD is not a new one. P.A.N.D.A.S. (Pediatric Autoimmune Neuropsychiatric Disorder Associated with Streptococcus) is a pediatric syndrome (still somewhat controversial) in which symptoms of OCD begin rather suddenly after a streptococcal infection in a susceptible child and then follow an episodic course, with exacerbations triggered by subsequent infections (Swedo et al., 1998). Localized brain inflammation triggered by autoantibodies, analogous to rheumatic fever and Sydenham’s chorea, represents the hypothesized causal link. Some authors have suggested that such an autoimmune pathogenesis may also be at play in adult OCD and not just in a small subset of pediatric-onset disease (e.g., Bhattacharyya et al., 2009). The new HoxB8 study, which produces a compulsive behavior through unclear actions (or lack of action) by HoxB8-deficient microglia, is quite different from such an autoantibody-mediated pathogenesis, but it resonates with the hypothesized connection between dysregulation of the immune system and OCD symptomatology.

However, the connection to OCD is based on an intuitive resemblance of the observed excessive grooming to repetitive and inflexible behaviors seen in OCD—that is, to the “face validity” of the model. Face validity can be a fickle guide in models of psychiatric disease. For example, excessive grooming has been described as OCD-like in other contexts (e.g., Welch et al., 2007), but it could as easily be interpreted as a model of trichotillomania, autistic stereotypy, Tourette syndrome, amphetamine-induced stereotypy, drug habit, or something ethologically unique to mice. Grooming is a particularly difficult phenotype to interpret in a cross-species comparison as it is quite variable among species—rodent grooming is quite different from primate grooming—and presumably subject to substantial selective pressure, due both to ecological and social factors (not to mention the presence or absence of fur). Therefore, the mere fact that the animals groom excessively is a slim basis for describing them as a model of OCD.

Recent studies have tried to extend the face validity of similar proposed models in other genetically modified mice by examining anxiety. For example, anxiety along with excessive grooming is seen in mice with a mutation in SAPAP3 (Welch et al., 2007) or with SliTrk5 (Shmelkov et al., 2010). Since OCD is categorized in DSM-IV as an anxiety disorder and often presents with significant anxiety (although not always, in my experience), this additional phenotype strengthens the face validity of the model. Even so, such face-validity comparisons are best considered as analogies to human conditions. It is far from clear that observable measures of anxiety in a mouse adequately parallel the psychic and cognitive type of anxiety experienced by patients with OCD. Furthermore, anxiety is an extraordinarily non-specific psychiatric symptom—I often ask residents, as an exercise, to name psychiatric disorders that are not frequently characterized by anxiety, and they are hard pressed to come up with more than a handful.

An additional evaluation uses predictive validity—that is, the ability of medications used to treat OCD to ameliorate the observed phenotypes. This was done in the studies of both the SAPAP3 and SliTrk5 mice, referenced above, in which SSRIs were found to ameliorate both excessive grooming and anxiety phenotypes. Such a predictive validity test also has significant weaknesses, however, because of the imprecise nature of both psychiatric diagnosis and psychiatric pharmacotherapy. SSRIs are of benefit in only 50-60 percent of cases of OCD, even with optimal dosing (see, e.g., Bloch et al., 2009). And SSRIs are also used to treat depression, generalized anxiety disorder, social anxiety disorder, premenstrual dysphoric disorder, post-traumatic stress disorder, bulimia, anorexia, and a host of other conditions. Therefore, response to an SSRI does not validate a mouse model as recapitulating core aspects of the neurobiology of OCD, and lack of response should not be interpreted as particularly undermining of a model’s validity.

Ultimately, the most valid models of OCD, or of any other neuropsychiatric condition, will be those based on confirmed aspects of the neurobiology of the disorder, such as well-validated genes of large effect size, specific molecular or cellular changes strongly and specifically associated with the disorder, or environmental stressors or insults with a similarly specific association. Such toe-holds into the pathophysiology of the disorder can be leveraged by carrying them over to animal models in which their downstream consequences, and potential ameliorative therapies, can be examined. Obviously, such validated hints of the pathophysiology of disease are few and far between in neuropsychiatric conditions, and virtually absent in obsessive-compulsive disorder, a relatively under-studied condition. In my view, until we have such biological grounding on which to base the construct validity of our models, it is best not to describe these or any other mice as representing “an animal model of OCD.” It is better to speak of an animal that exhibits excessive grooming, plain and simple. The relationship to OCD, or any other human neuropsychiatric condition, remains an empiric question that will be challenging to answer in a satisfying way.

However, this comment does not detract in any scientifically important way from the importance of the recent paper from the Capecchi group. They have certainly provided a valuable examination of an animal that exhibits maladaptively excessive and inflexible grooming. In addition, their striking finding of the critical role of microglia in producing this phenotype is of profound importance and will push the field towards a deeper appreciation of the importance of immune-brain interactions, not only for general brain health, but for the development and modulation of specific phenotypes. This is an exciting finding indeed.


Greer JM, Capecchi MR. Hoxb8 is required for normal grooming behavior in mice. Neuron . 2002 Jan 3 ; 33(1):23-34. Abstract

Swedo SE, Leonard HL, Garvey M, Mittleman B, Allen AJ, Perlmutter S, Lougee L, Dow S, Zamkoff J, Dubbert BK. Pediatric autoimmune neuropsychiatric disorders associated with streptococcal infections: clinical description of the first 50 cases. Am J Psychiatry . 1998 Feb 1 ; 155(2):264-71. Abstract

Bhattacharyya S, Khanna S, Chakrabarty K, Mahadevan A, Christopher R, Shankar SK. Anti-brain autoantibodies and altered excitatory neurotransmitters in obsessive-compulsive disorder. Neuropsychopharmacology . 2009 Nov 1 ; 34(12):2489-96. Abstract

Welch JM, Lu J, Rodriguiz RM, Trotta NC, Peca J, Ding JD, Feliciano C, Chen M, Adams JP, Luo J, Dudek SM, Weinberg RJ, Calakos N, Wetsel WC, Feng G. Cortico-striatal synaptic defects and OCD-like behaviours in Sapap3-mutant mice. Nature . 2007 Aug 23 ; 448(7156):894-900. Abstract

Shmelkov SV, Hormigo A, Jing D, Proenca CC, Bath KG, Milde T, Shmelkov E, Kushner JS, Baljevic M, Dincheva I, Murphy AJ, Valenzuela DM, Gale NW, Yancopoulos GD, Ninan I, Lee FS, Rafii S. Slitrk5 deficiency impairs corticostriatal circuitry and leads to obsessive-compulsive-like behaviors in mice. Nat Med . 2010 May 1 ; 16(5):598-602, 1p following 602. Abstract

Bloch MH, McGuire J, Landeros-Weisenberger A, Leckman JF, Pittenger C. Meta-analysis of the dose-response relationship of SSRI in obsessive-compulsive disorder. Mol Psychiatry . 2009 May 26. Abstract

View all comments by Christopher Pittenger