New label-free methods for protein relative quantification applied to the investigation of an animal model of Huntington Disease

Abstract

Spectral Counts approaches (SpCs) are largely employed for the comparison of protein expression profiles in label-free (LF) differential proteomics applications. Similarly, to other comparative methods, also SpCs based approaches require a normalization procedure before Fold Changes (FC) calculation. Here, we propose new Complexity Based Normalization (CBN) methods that introduced a variable adjustment factor (f), related to the complexity of the sample, both in terms of total number of identified proteins (CBN(P)) and as total number of spectral counts (CBN(S)). Both these new methods were compared with the Normalized Spectral Abundance Factor (NSAF) and the Spectral Counts log Ratio (Rsc), by using standard protein mixtures. Finally, to test the robustness and the effectiveness of the CBNs methods, they were employed for the comparative analysis of cortical protein extract from zQ175 mouse brains, model of Huntington Disease (HD), and control animals (raw data available via ProteomeXchange with identifier PXD017471). LF data were also validated by western blot and MRM based experiments. On standard mixtures, both CBN methods showed an excellent behavior in terms of reproducibility and coefficients of variation (CVs) in comparison to the other SpCs approaches. Overall, the CBN(P) method was demonstrated to be the most reliable and sensitive in detecting small differences in protein amounts when applied to biological samples.

Fig 1. Scatter plots of the quantitative measurements of all the E. Coli proteins.

Scatter plots of the quantitative measurements of all the E. Coli proteins identified in the two replicates of mixture A by the four different normalization methods used, i.e. NSAF (panel A), R_SC (panel B), CBN(P) with f = 1/P (panel C) and CBN(S) with f = 1/t (panel D). On the X-axis is reported the normalized Spectral Counts of replicate 1, and on the Y-axis the normalized Spectral Counts of replicate 2.

Fig 2. Comparative analysis of all spectral counts normalization methods applied to E. Coli proteome.

(A) Median values for the best fitting slopes calculated for each pair of technical replicates for all samples analyzed (except mixture B) with each normalization method. (B) Coefficient of variation (CV) for the evaluation of data dispersion for all the normalization methods used in the analysis of E. Coli proteome. (C) Representation of the logarithmic Fold Change distribution around the theoretical value indicated by the point line for all normalization methods. (D) Box plots of the CV for the fold change in the mix E / mix D pair, calculated on all the four possible pairs of the technical replicates. In panels B and D the box and whisker extremes represent 25–75%.

Fig 3. Fold change analysis comparing all spectral counts normalization methods.

(A) The mean Fold Change values for unchangeable proteins (BSA, HBA, HBB, and PYG) within the ten different pairs of mixtures calculated by the four normalization methods. The theoretical values are 0 and it is indicated by a point line. (B) Comparison of the experimental and theoretical Fold Change values for the relative quantification of ENO and ADH obtained by using the SpC-based methods and reported in the logarithmic scale. (C) Table reporting linear regression best-fit values for all ENO and ADH linear regressions.

Fig 4. Visualization of identified proteins with all methods in the WT and HD mouse model.

(A) The histogram shows for each method the number of proteins that appeared to be both differentially expressed and statistically significant. (B) Venn diagram referred to all proteins. The central area represents the proteins common to all methods.

Fig 5. Validation of a selected group of proteins differentially expressed in WT and HD mouse model.

(A) Western blot assays performed on total protein extracts from three mutant (zQ175(1), zQ175(2), zQ175(3)), and three wild-type (wt(1), wt(2), wt(3)) mice with antibodies against the selected proteins. β-actin was used for normalization. (B) MRM superimposed traces of transitions of one of CAMK2A proteotypic peptide reported for zQ175 (left panel) and WT (right panel) together with one ACTIN peptide transition. (C) Densitometric analysis of data from the western blot of panel A. The indicated values in the graph represent the percentage of arbitrary units compared to WT to which 100% was assigned. Results are represented as the as mean ± SD (standard deviation). The statistical significance was evaluated by parametric (Welch’s) or non-parametric (Mann-Whitney) tests when data failed the Shapiro–Wilk normality test. * p < 0.05, ** p < 0.01 *** p < 0.001, **** p < 0.0001. (D) Fold Change measured by MRM analysis of pooled HD and WT samples, respectively.