The PSD contains molecular machineries for NMDA, AMPA and mGlu receptors, all of which have been implicated in the pathophysiology of various neuropsychiatric illnesses 
. Thus, examining the PSD of patients' brains will greatly aid pathophysiologic investigations, yet such study paradigms have not been developed for human postmortem brains.
The purpose of this study was to assess whether the PSD can be isolated as a biochemical fraction from postmortem brains and to test the integrity of these fractions. Our results show that methods employing density gradient based purification (Method 1) 
and pH based differential extraction (Method 2, 3)
, can produce fractions that are highly enriched in the PSD proteins with a reasonable degree of purity and in a reproducible fashion.
All three methods were effective in producing PSD fractions that are mostly devoid of presynaptic proteins. A few differences between these methods, however, may determine the utility of each depending on the purpose of projects. Methods 2 and 3 are superior to Method 1 in the yield of PSD fractions and therefore are more appropriate when the availability of the tissues is limited, while Method 1 is recommended when the issue of purity is critical.
EM examination of subcellular fractions showed relatively intact ultrastructures in subcellular fractions from postmortem brains. In SPM fractions, synaptosomes were well formed, containing intact synaptic vesicles. In the PPF and much less frequently in the PSD, scaffolds of opposing presynaptic and postsynaptic membranes were observed, as shown in rodent counterparts 
. In addition, ultrastructural characteristics of PSD fractions appear comparable to those previously reported in rodents 
. These data support the notion that the PSD, as a microdomain, is relatively well-maintained in postmortem brains, and can be isolated to a reasonable degree of purity. Various confounding factors associated with postmortem brain tissue may affect subcellular ultrastructures and the integrity of proteins 
. These factors could contribute to variability in the yield and protein composition of PSD fractions, which would make comparisons between individuals or groups unreliable. A few parameters that we have tested in the present study, however, suggest that there is a reasonable degree of consistency in PSD samples of human subjects. First, the rate of enrichment of PSD proteins in the PSD fractions, tested in 12 subjects, appears to be relatively consistent between samples (), while the PMI of the subjects ranged between 3.5 and 22 h and the freezer storage time varied from 5 to 17 years. Second, protein composition of the PSD fractions, as assessed by the ratios of PSD-95 relative to NR1 for example, was grossly similar in the PSD fractions among all subjects. The rate of enrichment and protein composition are relatively consistent, most likely because the PSD is a resilient microdomain that can endure harshness of the fractionation procedures.
PSD fractions of postmortem brain tissue may permit the study of protein-protein interactions with some confidence. The results of the NR1 IP showed that the association of NR1 with PSD-95, PLCγ or NR2A in the PSD was almost as consistent among individuals as those in the post-nuclear fractions ( as an example). This may suggest that the fractionation procedure of the PSD does not disturb protein-protein interactions. It is of note, however, that the association of NR1 with NR2A and PLCγ was lower in the PSD than in post-nuclear fractions (). This could be due to different protein associations between the two fractions. Alternatively, it may be the case that only more robust protein-protein associations survive during the fractionation procedure of the PSD, although the selective process is still consistent among samples.
2D LC-MS/MS was employed to further characterize the protein composition of PSD fractions. PSD enrichments from two biological samples were combined in a statistical analysis, which integrates data from consensual and non-consensual peptide identifications in a Bayesian fashion (48). Our list of identified proteins appears highly inclusive, yet provides an extensive coverage of known PSD proteins, as evidenced by the proteins cataloged in . Combined with Western blot results confirming enrichment of PSD proteins, our 2D LC-MSMS data provide further evidence that PSD proteins can be enriched by a biochemical fractionation of postmortem brain tissue. The number of proteins in the PSD is presently unknown 
. Using immuno-EM, EM tomography, and scanning transmission EM, Chen et al (2005) and Petersen et al (2003) have estimated that an average PSD with a 360 nm diameter and a total molecular mass of 1.10±0.36 gigadaltons, could be composed of 10,000 proteins of 100kD 
Proteomic analyses are prone to detecting contaminants that are either highly abundant in target tissues or co-enriched during biochemical fractionation 
. While our list of identified proteins is highly inclusive, it may serve as a comprehensive list of proteins in the PSD of human postmortem brains. The next step will be to determine which of these proteins are specific to the PSD or highly enriched in this microdomain. Once confirmed, PSD proteins on the list can be quantitatively evaluated in conjunction with immunoprecipitation and heavy label internal standards.
The ultimate goal of this study was to test a study paradigm in which to examine the PSD of human postmortem brains with respect to its integrity, protein composition, and protein associations. Our results showed surprisingly well-maintained ultrastructures in biochemical fractions, relatively consistent yield of PSD fractions and the stability of protein compositions during the fractionation procedure. Considering various postmortem confounding variables and their effects on a wide array of protein properties, it will still be important to be selective for parameters that are stable enough for comparisons. When such cautions are exercised, however, this approach can provide insights into protein-protein interactions that are critical for glutamatergic and other signaling mechanisms in post-synaptic neurons.