The analysis of proteins secreted by cells represents an important tool for the understanding of cellular communication, the behavior of cells in tissues and - more specifically - the composition of the TEM. While workflows to analyze the secretome in serum-containing media have been described, they are hampered by the presence of highly abundant serum proteins, which limit the observable dynamic range and thus the analytical depth of the secretome [27]. To overcome these issues associated with serum-containing media, serum-free culture conditions have been used for secretome analysis and the identification of low abundant secreted proteins [28]. However, cells grown under these conditions often display limited cell growth, show in many cases reduced viability and will adapt their intracellular pathways to these conditions [20], this potentially skewing the composition of the secretome. To overcome these issues, bioorthogonal metabolic labeling approaches represent an attractive alternative as they allow to grow cells in the presence of serum and to specifically isolate newly synthesized [29–31] and secreted proteins using click chemistry [32] out of complex protein mixtures, e.g. present in the serum containing culture supernatant. While the analysis of the secretome based on azide-containing amino acid metabolic labeling in combination with click chemistry pioneered by Eichelbaum et al. [18] has distinct advantages, the use of amino acid analogs may induce yet undescribed effects in the investigated cells. Along these lines, it was reported that azide-containing amino acids are incorporated at reduced rates into proteins. This has been illustrated for the Meth analog AHA, showing an around 400-times reduced translational activity [22].
These previous observations prompted us to investigate in detail the consequences of AHA labeling on the gene and protein expression profiles and correlate changes with the presence of proteins in the secretome. We observed distinct effects on cell viability (Fig. 1a) and proliferation (Fig. 1b) after 18–24 h of AHA labeling using MC38 cells. Furthermore, AHA labeling for 20 h resulted in the induction of apoptosis ranging from delicate, as observed for MC38 cells, to significant effects, as observed for Jurkat cells or primary T cells from OT-I mice (Fig. 1c). This clearly indicates that AHA labeling can affect cell viability with different severities and thereby can be expected to have an impact on the secretome. In line with this, secretome analysis of Jurkat and OT-I T cells after 20 h of AHA labeling revealed, unlike what was observed for MC38 cells (Supplementary Figure S1a), no significant enrichment of AHA-containing proteins (Supplementary Figure S1b, c). This could indicate that these cells cannot efficiently take up and incorporate AHA and thus experience methionine starvation. For these cell types alternative secretome workflows using either click chemistry-based enrichment of secreted glycoproteins [33, 34] or click-selective tRNA synthetases [35, 36] might be viable alternatives. However, these approaches either focus on a subset of the secretome or require transfection of the cells of interest.
As MC38 cells were only marginally affected in viability, showed no significant induction of apoptosis (Fig. 1) and allowed for a significant enrichment of AHA-containing proteins (Supplementary Figure S1a), they appeared to be well suited for AHA-based secretome analysis. We therefore decided to investigate potential AHA effects in detail using the workflow depicted in Fig. 2. Our results indicate that AHA labeling resulted in the up- or down-regulation of more than 30% of the genes identified in transcriptome analysis (Fig. 3a, b) and more than 20% of the proteins identified in proteome analysis (Fig. 4a, b). GO analysis revealed that AHA labeling resulted in the induction of an unfolded protein response, induction of cellular stress as well as inhibition of gene transcription, DNA replication and the induction of apoptosis (Fig. 5). These findings are well in line with the reduced translational activity of AHA previously described [22] and point towards effects of AHA on protein folding and/or stability.
Aligning the regulated expression of genes and proteins with proteins secreted by MC38 cells, we found that more than 30% of the proteins identified in the secretome were up- or down-regulated with regard to their gene and protein expression profile (Fig. 7a). This can be expected to have a significant quantitative and qualitative impact on the composition of the secretome, resulting in an under- or overrepresentation of proteins in the secretome, thus inducing an unwanted and unpredictable bias in the data. We next extended this type of analysis to 3 human lung carcinoma and to 3 human melanoma cell lines. Similar to our observations for MC38 cells, AHA labeling caused an up- or downregulation of genes ranging from 13.2–34.5% resulting in a clear clustering and separation of cells grown in Meth or AHA conditions after PCA analysis (Fig. 3c, d). The observed regulation of gene expression affected also the composition of the proteome and again, we found a clustering of cells grown in Meth or AHA conditions after PCA analysis and an up- or downregulation of protein expression varying from 6.5–18.6% (Fig. 4c, d). Not surprisingly, the level of changes in the gene expression profile correlated with changes in the proteome composition, as nicely exemplified by SK-MEL-37, showing the highest, and NCI-H1650 cells, showing the lowest level of alterations (Fig. 3d and Fig. 4d). As observed for MC38 cells, a substantial number of proteins identified in the secretome of the six human tumor cell lines displayed an AHA-induced up- or downregulated expression profile, ranging from 15.0–39.1% (Fig. 7b). This strongly suggests that the detected secretome composition is modulated by the AHA labeling. In line with this, the GO analysis of the regulated genes and proteins (Figs. 5, 6) indicated that AHA might induce protein misfolding and reduced protein translation.
The low number of proteins detected in the secretome of the lung carcinoma cell lines compared to what was identified in the melanoma cell lines might be a consequence of the observed slow growth rate of these cells. As a consequence, these cells might display a lower rate of protein translation and therefore, lower numbers of proteins carrying an AHA label, allowing their purification by click chemistry.
To alleviate the effects on the qualitative and quantitative protein composition of the secretome induced by AHA labeling, shorter incubation times may be used, but those will also reduce the amounts of labeled proteins and thus limit the sensitivity of this approach, as it will reduce the amounts of proteins available for click chemistry-based enrichment from serum containing culture medium.
Concludingly, our data describe a profound effect of metabolic labeling using AHA on both transcriptome and proteome level. In addition, not all cell types seem to be able to either take up or incorporate AHA. This became evident during the analysis of proteins secreted by the Jurkat T cell line or primary T cells derived from TCR-transgenic OT-I mice. These cells displayed high rates of cell death (Fig. 1c) and in contrast to e.g. MC38 cells, showed no significant enrichment of click chemistry-captured proteins (Supplement Fig. 1).
Our results indicate that bioorthogonal metabolic labeling-based analyses targeting newly synthesized[37] and/or secreted proteins should be preceded by detailed analyses regarding cellular viability, the induction of apoptosis and, if possible, alterations in the gene expression profiles.