We analyzed CSF in 37 twins and 65 controls using the SEM technique to identify potential genetic and environmental effects influencing accumulation of particles in CSF. The main findings from this report are as follows.
First, the structures in CSF were strongly associated with schizophrenia and bipolar disorder. These results confirm previous findings in schizophrenia 
and bipolar disorder 
Second, the structures were rare in the healthy controls, implying that structures detected with this method are exceptional in the normal population. The CSF of the twins and controls were sampled following identical procedures, which were also used in a previous study investigating bipolar disorder 
, which rules out the possibility that the findings were due to different sampling techniques.
Third, we found an association between structures in the CSF and co-twin status, which indicates a genetic or environmental effect. Control for potential confounding factors in the logistic regression models suggests that the CSF structures do not seem to be explained by medication, alcohol abuse, anxiety disorder, somatic disorder, BMI or type of needle used during lumbar puncture.
Finally, we studied the distribution of CSF structures in the monozygotic and dizygotic co-twins and observed a clear trend towards a higher level of positive CSF findings in the monozygotic co-twins. These results did not reach the statistical threshold, probably because of a lack of power due to small sample size. Taken together the findings are consistent with a genetic influence on the CSF findings.
CSF/serum albumin ratio has been suggested as a marker of blood-CSF barrier dysfunction 
. Earlier studies have reported on increased ratio in some patients with schizophrenia 
. However, none of the participants with schizophrenia or bipolar disorder in our sample had an increased albumin ratio. Thus, when included in the model blood-CSF barrier dysfunction, did not have any effect on the main results. We also included markers for acute inflammatory activity, current medication and information about somatic disorders, (e.g., inflammatory-related disorders and neurological disorders) in the analysis. The results support the assumption that the presence of CSF structures is not sensitive to ongoing infection or inflammation. Investigating the presence of previous infections, we decided not to include reports of unspecified virus infections in that they are very common and unspecific. The effect of psychotropic medication (antipsychotics, lithium, anticonvulsants and antidepressants), examined in the proband group only did not show any effect on the presence of CSF particles when tested separately or combined. This finding contradicts the view that CSF structures are related to psychotropic medication.
When analyzing the probands with schizophrenia and schizoaffective disorder and their co-twins separately from the probands with bipolar disorder and their co-twins we found a stronger association between schizophrenia and schizoaffective disorder and particles in CSF compared to bipolar disorder. There was no significant association between the co-twins discordant for bipolar disorder and particles. However when we analyzed without the patients with bipolar disorder type II and their co-twins the association between group-status and particles increased compared to our main results indicating that the particles in CSF may be associated with psychotic symptoms. This finding is in line with the tendency reported in the study by Båve et al. in which patients with bipolar disorder type I were more likely to display CSF structures than patients with bipolar II disorder 
In 13 of the 20 individuals with a positive CSF-finding, structures were identified in the first 0.6 ml of CSF but not the second fraction. A consensus guideline for neurological research from 2009 prescribes that the very first mL of CSF should be discarded 
. Interestingly following this recommendation would only have identified individuals with CSF structures in the second fraction. One hypothesis is that a low level of neuroinflammation or neurodegeneration in schizophrenia and bipolar disorder gives rise to a smaller amount of apoptotic products in the CSF. The individuals who also display structures in the second CSF fraction may have numerous particles in their CSF, indicating that they have a less effective cleansing function of CSF or a higher level of neurodegeneration in the brain 
. The possibility that the structures are unspecific artifacts does not accord with the negative findings in the controls.
Only 3 controls presented structures in their CSF. One female singleton control presented spherical particles in her CSF of a similar type found in patients with schizophrenia and bipolar disorder (). Particles of similar size and form were found in the second CSF fraction of an unaffected, un-medicated male co-twin of a brother with schizoaffective disorder (). Further, a monozygotic twin pair that served as controls presented particles in their CSF (). The structures observed in the first twin are similar to protein-like aggregates, possibly related to the increased albumin ratio in CSF (). The structures in the second twin were similar to the spherical particles observed in the proband sample (). The findings in both twins may be related to a blood-CSF barrier dysfunction. They may also be latent disease carriers for bipolar disorder or schizophrenia.
Furthermore, spherical CSF-particles may also occur in other disorders involving the central nervous system as recently reported in amyotrophic lateral sclerosis by Zachau et al. 
. In this study the CSF of an ALS patient displayed 100 times more phosphatidylserine-positive microparticles and 400 times more cell-derived microparticles of leukocyte origin compared to healthy controls although the leukocyte count in CSF was normal. Similar studies of microparticles in CSF in schizophrenia and bipolar disorders are called for.
Strengths and limitations
Strengths of the present study include the careful lifetime assessment of the psychiatric diagnosis, controlling for potential confounding factors in the analysis and an identical collection procedure of the CSF in all participants included in the study. Moreover, the control group is large and population based. However, several limitations of the study also need to be addressed. The possible confounding effects of perinatal and childhood infections were not clarified because of lack of information. The small sample size of disease discordant twin-pairs prevented further analysis of the genetic contribution of the CSF findings. In addition, psychotic symptoms of the probands were not investigated using a structured instrument such as SANS and SAPS 
adjacent to the collection of CSF.
Further research in larger twin samples of cases diagnosed with schizophrenia and bipolar disorder as well as high-risk individuals (e.g., first-degree relatives of patients with psychosis) may help to answer the question of whether structures of the CSF precede a manifest disease state and may serve as a trait marker. Further work is needed to explore the origin and the composition of the anatomical structures in both the first and second CSF fractions.
The main conclusion from this study is the strong statistical support for structures in CSF identified with the SEM technique in patients with bipolar disorder or schizophrenia. The results indicate that the CSF structures likely originate from trait-dependent factors.