MerCASBA: an updated and refined database of caspase substrates
- First Online:
- Cite this article as:
- Fridman, A., Pak, I., Butts, B.D. et al. Apoptosis (2013) 18: 369. doi:10.1007/s10495-012-0789-z
To the Editors
In this correspondence, we describe a significant update and expansion of the original CASBAH (CAspase Substrate dataBAse Homepage) created by Luthi and Martin . CASBAH contained both putative (not formally identified in the cited reference(s)) and experimentally confirmed caspase substrates. In MerCASBA (MERck CAspase Substrate dataBAse), we discarded putative entrants and reviewed the literature sources of all other entrants to retain only experimentally confirmed sites. Additionally, minor errors inadvertently introduced in CASBAH, including mistyped substrate sequences, amino acid position numbers, and UNIPROT IDs were corrected. We also added caspase recognition sequences identified in recent papers using unbiased proteomic approaches [2, 3]. Finally, we undertook a comprehensive analysis of scientific literature published since CASBAH to identify novel substrates. The focus on experimentally validated cleavage sites distinguishes this study from other recent work .
We compared the amino acid frequency at every position between P6–P4′ in MerCASBA versus their distribution in the human proteome (Fig. 1b). One striking observation is the paucity of either Glu or Pro at position P1′—each is under-represented more than tenfold (p < 0.001). Our analysis of all 724 caspase cleavage sites within intact proteins confirmed an earlier study , which also found that Pro and Glu were the least preferred amino acids at P1′. The inefficient processing of potential caspase substrates containing Pro at P1′ is supported by experimental data: caspase 3 does not cleave in vitro between Asp and Pro of DMMDP in syntaxin-5, despite its location in an exposed loop region (Fig. 4 of ). Our observation that Pro is not favored at the P1′ position similarly argues against it being a recognition site for caspase cleavage. Salvesen et al., adopting a peptide/purified caspase model , reported that small amino acids are preferred at the P1′ position of caspase substrates. Analysis of the MerCASBA database confirms this; for example Gly at position P1′ is fivefold more enriched than its natural frequency of occurrence. Similarly, we observed a higher abundance of Ala and Ser at P1′ than their natural frequency. In another example of alignment with earlier experimental studies , our current findings are consistent with executioner caspases containing a narrow S4 pocket that strongly favors Asp in the P4 position. Conversely, Lys and Arg, positively charged amino acids, are clearly not favored in position P4, as they are 10- and 20-fold (respectively) less abundant than their natural frequency of occurrence. MerCASBA does not agree with or confirm all previous experimental findings in caspase enzymology: in contrast to earlier reports [8, 9], we see no obvious preferences for hydrophobic amino acids at P5. In fact, negatively charged Glu appears to be marginally preferred at this site.
In CASBAH some selection bias was unavoidable since early entrants were identified from studies evaluating predicted caspase cleavage site(s) in defined proteins of interest; often that particular protein was examined because it contained a DxxD motif and, as a result, Asp at position P4 was found in 45 % of the cleavage sites. In comparison, using unbiased proteomic approaches [2, 3], only 21 % of the substrates contained the amino acid Asp at P4. Similarly, MerCASBA moves away from the original bias as only 33 % of the cleavage sites contain the DxxD motif.
We acknowledge and appreciate the foundation of the initial CASBAH as a valuable resource for the caspase field and hope that our updated and validated MerCASBA database will be equally useful to the apoptosis research community.
Conflict of interest
The authors are employees of Merck and Co. and declare they have no conflict of interest.