The protein product encoded by the Oct-1 cDNA used throughout this thesis. The 5 ’ end of Oct-1 mRNA remains uncloned, and thus the endogenous Oct-1 polypeptide contains more amino-terminal amino acid residues than indicated here. In the figure, the DNA binding domain, referred to as the POU domain, and putative glutamine- and serine/threonine-rich activation domains are indicated.
7 4 3
'POU^
POUhd
Figure 3.3. The relative binding affinity of the complex containing endogenous Oct-1 to the three different binding sites in the IFN-6 promoter and to the perfect octamer motif.
A) An EMSA analysis using HeLa cell nuclear extracts on the three probes (NRD II, TATA and NRD I) derived from the IFN-6 promoter, as well as on the probe containing a perfect octamer motif. The photograph is from one autoradiographic exposure of a single gel.
< < H Q I- Û O (T < CC O Z I- Z
w
W Ww
B
A T G C A A A T
o c t a m e r
A T G T A A A T
NRD I I
A T A T A A A T
TATA
A T A G A G A G
NRD I
relative affinity of Octl;
octa > NRDII > TATA > NRDI
99
The POU domain is a 155-162 amino acid region composed of two distinct, independently folding, DNA-binding structures, both of which make specific contacts with DNA, and thus contribute to effective DNA binding (Sturm and Herr 1988, Aurora and Herr 1992, Botfield et al 1992). The more carboxy-terminal subdomain, the POU-homeodomain (60 amino acids), is similar to the DNA binding domain of the previously characterized members o f the homeodomain family of transcription factors, while the more amino- terminal POU-specific domain (74-82 amino acids) is characteristic of POU-proteins (reviewed in Verrijzer and Van der Vliet 1993). Homeodomains contain a helix-tum-helix DNA binding m otif (Laughon et al 1984), of the kind which were first identified in bacterial DNA binding repressor proteins (Sauer et al 1982). Although the amino acid sequence of the POU homeodomain is quite divergent from the homeodomains in non- POU proteins, both the overall structure and mode of DNA recognition of POU homeodomains are very similar to the previously characterized ones, such as the Engrailed and M A Ta2 homeodomains (Klemm et al 1994). Recent NMR studies have demonstrated that the POU-specific domain also contains a helix-tum-helix motif, topologically similar to the DNA-binding domain of the phage X repressor (Assa-Munt et al 1993, Dekker et al
1993), although in many POU proteins, including Oct-1, the turn region is longer. A POU- homeodomain and a POU-specific domain are covalently linked by a short (15-27 amino acids, 24 amino acids in Oct-1) and structurally disordered tether, which is nonconserved in sequence between different POU-family proteins. For example, Oct-1 and Oct-2 have nearly identical POU-specific and POU homeodomains - accordingly rendering their DNA binding specificity identical - however, their tether regions are very different (Herr et al 1988). Supporting the proposed flexible nature of the tether region, it is readily accessible to proteases (Botfield et al 1992). It is interesting to note that homeodomains are notoriously promiscuous in the sequence specificity of their DNA binding (figure 3.3.; reviewed by Laughon 1991); thus the POU-specific domain can be seen to function to supplement this limitation. A similar covalent association of two DNA binding folds has been utilized in Paired domain proteins to enhance target specificity. An interesting contrast to these is provided by a subclass of LIM proteins (Sânchez-Gârcia and Rabbitts 1994), which possess a homeodomain adjacent to the cysteine-rich zinc finger-like LIM domain. In these proteins, the DNA binding activity of the homeodomain appears to be inhibited by the neighbouring LIM domain (Sânchez-Gârcia et al 1993, Xue et al 1993).
The POU homeodomain has been shown to contact the 3' half (AAAT) of the consensus octamer target site ATGCAAAT, while the POU-specific domain associates with the 5' ATGC portion (Verrijzer et al 1990, Verrijzer et al 1992a, Klemm et al 1994) - the half sites lie on opposite sides of the DNA double helix. Biochemical studies have indicated that while the isolated Oct-1 homeodomain binds DNA reasonably well, the isolated POU-
100
specific domain binds only poorly. In addition, binding site selection and DNA binding competition studies indicate that Oct-1 prefers an arrangement in which the POU-specific binding half site and the POU homeodomain binding half site are juxtaposed (Verrizjer et al
1994). The POU-specific domain and POU homeodomain do not directly contact each other, despite that in the crystal structure analyses have revealed that they make contacts with overlapping phosphates (Klemm et al 1994). Even in the absence of direct protein- protein contacts between the domains, they may interact through subtle alterations in the structure or flexibility of DNA. This would be consistent with a biochemical study on chimaerie POU domains carrying corresponding protein segments from different POU proteins, which has suggested that the two subdomains influence each other's DNA binding specificity (Aurora and Herr 1992).
W e have investigated the DNA binding properties of the Oct-1 POU domain in isolation, separated from the amino- and carboxy-terminal regions of the Oct-1 protein. To allow the the production o f recom binant POU domains, subclones for cell-free translation (p T 7 0 c tlP 0 U H D + and p T 7 0 ctlP 0 U H D ", in which the production o f the POU mRNA v ariants is governed by bacteriophage T7 prom oter) and bacterial expression (pETO ctlPO U ) were created (see the legend to figure 3.4.). In these plasmid constructions we utihzed the Oct-1 cDNA insert in the parental plasmids pCGOctl and pCGOctlHD" (see figure 3.19.; a gift from W.Herr, Cold Spring Harbor).
The POU domains from T7 clones were synthesized in coupled in vitro transcription/ translation extracts from wheat germ (figure 3.4.A.). The POU HD+ clone encodes the wild type Oct-1 POU domain, whereas the product of the POU HD" clone contains a triple alanine substitution (W FC>AAA) created in the DNA recognition helix o f the POU homeodomain.
The pETO ctlPO U plasmid was designed so that the POU encoding region is in frame with a 3' vector sequence encoding a stretch of six histidine residues. The presence of the His- tag in the fusion protein allows the purification of the recombinant POU protein by virtue of the affinity o f the polyhistidine stretch for nickel-ions. The bacterial expression and subsequent purification were performed as described in the M aterials and Methods. A Coom assie-stained SDS-PAGE with the samples from the fractions collected during purification is shown in figure 3.4. The retrieval and elution o f the overexpressed POU domain was very efficient; furthermore, no contaminating bacterial proteins can be detected in the selected elution fraction.