The proposed plant connexin is a protein kinase-like protein.

Functional analogies appear to exist between gap junctions, the structures involved in maintaining conductivity between contacting animal cells, and plant plasmodesmata (see papers collected in Robards et al., 1990, for state-of-the-art discussion). These include similar behavior with regard to tracer dyes, similar electrical properties, and similar mode of down regulation by effectors of protein kinase C (Xu et al., 1990, and references therein). Accordingly, significant effort has been devoted to the search for plant homologs of connexins, the principal protein components of gap junctions. Xu et al. (1990) and Yahalom et al. (1991) reported the occurrence of several proteins in dicots and monocots that cross-reacted with antibodies against animal connexins. Such proteins were found mostly in the cell wall fraction and were localized in the plasmodesmatal area by immunogold labeling. These experiments culminated in the isolation by Meiners et al. (1991) of acDNA clone, CX32, from an Arabidopsis expression library that encodes a protein cross-reactive with anti-connexin antibodies. Tofurther confirm that thesequenced Arabidopsis clone codes for connexin, the authors attempted to align its deduced amino acid sequence to that of a rat connexin; the resulting alignment was not characterized in statistical terms, but it was concluded by the authors that the CX32 protein is a connexin homolog. We were surprised to notice that the alignment mentioned above contained an exceptional number of gaps (one gap per 10 residues on average) but did not reveal any well-defined amino acid motifs. Our attempts to align the CX32 amino acid sequence to animal connexins using the programs MACAW (Schuler et al, 1991) and OFTAL (Gorbalenya et al., 1989) failed to produce a statistically significant alignment. We then performed a comparison of the alleged Arabidopsis connexin sequence with amino acid sequence data bases using the BLAST program developed by Altschul et al. (1990), which is based on robust statistics for ungapped similar segments in protein sequences (Karlin and Altschul, 1993). The results were unexpected, showing that CX32 does not display appreciable similarity to any of the connexins but instead is clearly related to protein kinases, as shown in Figure 1. For example, the probability of the CX32 sequence matching the sequence of the Arabidopsis tyrosine/serine/threonine kinase by chance alone was computed to be below 10-l1. Analysis of the multiple alignment of the CX32 sequence with a selection of protein kinases (Figure l ) confirmed the highly significant conservation but showed that the CX32 protein contains an incomplete set of sequence motifs that are conserved in protein kinases and are thought to be required for their activity (Hanks et al., 1988; Taylor et al., 1992). The three N-terminal motifs characteristic of functional protein kinases, including a glycine-rich loop that is a principal structural element of the ATP binding site, are lacking in the published CX32 amino acid sequence. However, translation of the 5’ part of the CX32 cDNA sequence, which has been assumed to be unexpressed, revealed a putative equivalent to the motif 111 of protein kinases (Figure 1). It seems possible that the published sequence contains a frameshift error and that the actual gene may be longer than the sequenced cDNA clone, encoding also the N-terminal motifs. This is compatible with the observation that the CX32-specific mRNA isolated from Arabidopsis is somewhat longer than what is contained in the cDNA clone (Meiners et al., 1991). On the other hand, CX32 contains substitutions of several amino acid residues that are highly conserved in protein kinases and are directly implicated in catalysis (Taylor et al., 1992). Thus, it is not clear whether CX32 is a bona fide protein kinase or an inactive protein kinase homolog. At the same time, the plasmodesmatal localization of CX32 seems to be well established and is certainly provocative in view of the regulation of plasmodesmatal permeability. Our analysis thus shows that the alleged plant connexin is unrelated to animal connexins and is in fact a protein kinase-related protein. The reaction of anti-connexin antibodies with CX32 and other plant proteins (Meiners et al., 1991; Yahalom et al., 1991) may be due to the existence of structurally similar regions not reliably detectable by sequence comparison. It is uncertain whether this is relevant with respect to analogous subcellular localization of these proteins. Data base screening with the sequences of several animal connexins did not reveal related proteins from plants. It remains unclear whether or not plants have genes for real connexin homologs.