Main

We could identify no sequence similarity between FLASH and the Apaf-1/CED-4 or DED domains by searching the non-redundant protein sequence database at the NCBI using the gapped BLAST or PSI-BLAST programs3,4, and over 1,000 sequences in the database were found to be more similar to the ‘CED-4 homology’1 and ‘DED homology’ regions of FLASH than were CED-4 or Apaf-1.

Searching databases, however, may only detect less than half of all similarities between sequences in proteins that are considered to be homologues on the basis of structural comparisons5,6. Further analysis is needed, for example by direct comparison of functionally analogous proteins. We compared FLASH with Apaf-1/CED-4 and with DED-containing proteins by using the MACAW program7, but failed to detect any blocks of statistically significant sequence similarity (data not shown). We also used the PHI-BLAST program to assess the importance of the ATP-binding (P-loop) signature in FLASH (this program screens for similarity only those sequences that contain the specified signature), but found no similarity to the apoptotic ATPases even in this reduced search space. The other four motifs typical of the apoptotic ATPases2 are not conserved in the published alignment of FLASH with these ATPases.

To confirm the presence of DED-related domains in the FLASH sequence, we searched it with the DED profile by using the SMART system8 and an independent method based on the PSI-BLAST program that detects all known DED domains9, but we were unable to find any similarity to DED.

These tests cannot rule out a subtle similarity, although we believe that structure predictions for FLASH and phylogenetic analysis may effectively do so, at least with regard to the purported ATPase domain. Compositional complexity analysis using the SEG program10 indicates that FLASH is largely a non-globular protein (Fig. 1). The entire ‘CED-4 homology’ region of FLASH is predicted to be non-globular, which is incompatible with the compact structure based on a parallel β-sheet with inserted α-helices that is typical of ATPase domains11. The P-loop in ATPases and GTPases is preceded by a hydrophobic β-strand, but this feature is lacking in FLASH.

Figure 1: Diagram of the predicted domain organization of FLASH (roughly to scale).
figure 1

The numbered bar indicates amino-acid residue positions. Boxes, predicted globular regions; lines, predicted non-globular domains; CC, coiled-coil. Regions of alleged similarity1 to the apoptotic ATPases (CED-4) and DED domains (DRD, or DED-related domains) are indicated by broken lines. Predicted non-globular domains were detected by using the SEG program10, with the following parameters optimized for partitioning protein sequences into globular and non-globular domains: window length, 45; trigger complexity, 3.4; extension complexity, 3.7. Coiled-coil domains were predicted using the COILS2 program12; boundaries of the strongly predicted coiled-coil domain are indicated.

Full size image

The argument against structural similarity is supported by phylogenetic evidence. We have cloned and partly sequenced a human homologue of FLASH which has 67% amino-acid identity with FLASH in an alignment of 1,250 residues; the P-loop signature, however, is not conserved (data not shown; GenBank accession no. AF165161).

The structural and evolutionary evidence thus indicates that FLASH contains no globular domains with predictable functions and is not homologous to its functional analogues. FLASH does contain a predicted coiled-coil domain (Fig. 1) which may mediate functionally important protein–protein interactions12 and so is probably the best available lead we have from the sequence for further experiments.