Detection of protein folding defects caused by BRCA1-BRCT truncation and missense mutations

Most cancer-associated BRCA1 mutations identified to date result in the premature translational termination of the protein, highlighting a crucial role for the C-terminal, BRCT repeat region in mediating BRCA1 tumor suppressor function. However, the molecular and genetic effects of missense mutations that map to the BRCT region remain largely unknown. Using a protease-based assay, we directly assessed the sensitivity of the folding of the BRCT domain to an extensive set of truncation and single amino acid substitutions derived from breast cancer screening programs. The protein can tolerate truncations of up to 8 amino acids, but further deletion results in drastic BRCT folding defects. This molecular phenotype can be correlated with an increased susceptibility to disease. A cross-validated computational assessment of the BRCT mutation data base suggests that as much as half of all BRCT missense mutations contribute to BRCA1 loss of function and disease through protein-destabilizing effects. The coupled use of proteolytic methods and computational predictive methods to detect mutant BRCA1 conformations at the protein level will augment the efficacy of current BRCA1 screening protocols, especially in the absence of clinical data that can be used to discriminate deleterious BRCT missense mutations from benign polymorphisms.


INTRODUCTION
Germline mutations within the breast and ovarian cancer susceptibility gene BRCA1 predispose carriers to early-onset breast and breast-ovarian cancers (1). Accumulating evidence points to a role for the BRCA1 protein product in the regulation of multiple nuclear functions including transcription, recombination, DNA repair, and checkpoint control (2)(3)(4). Tumour-associated mutations occur throughout the BRCA1 coding sequence, but cluster to sequences encoding the N-terminal RING finger domain and the two carboxyterminal repeat BRCT domains (5)(6)(7).
The molecular details of how BRCA1 mutations contribute to the pathogenesis of cancer remain largely unknown. The functional significance of the BRCT region is highlighted by the high degree of sequence conservation within the BRCT regions of among mammalian, Xenopus, and avian BRCA1 homologues (8)(9)(10). Several lines of evidence reveal the BRCT is required for tumour suppressor function. A nonsense mutation, which removes 11 C-terminal residues of the second, BRCT (Tyr1853 → stop), is associated with early-onset breast cancer (11). Two cancer-linked BRCT missense mutations (12) that destabilize the BRCT fold (13)(14)(15), A1708E and M1775R, ablate the doublestrand break repair and transcription function of BRCA1 (16) and inhibit BRCT interactions with histone deacetylases (17), BACH1 (18) and the transcriptional co-repressor CtIP (19,20). Furthermore, mice with homozygous targeted mutations removing the C-terminal half of BRCA1 are viable but develop tumours, suggesting the missing BRCT and/or other domains are expendable for survival, but not for tumour suppression (21).
While all frameshift or nonsense mutations recorded in the Breast cancer Information Core (BIC) resulting in BRCA1 protein truncation are viewed as functionally deleterious (22,7), the physiological significance of the majority of missense variants has not been determined due to the absence of a distinctive functional assay for BRCA1. More than 70 missense substitutions have been recorded that alter the primary sequence of the tandem BRCT-repeats but pedigree analysis clarifying the disease linkage of these alleles is available for only eight of these variants (6,7,12,(23)(24)(25)(26)(27). Many of these amino acid substitutions may be linked with disease but remain as unclassified in the BIC because the presence of the allele has not been tested in the general population, or the segregation of the allele with disease within a family is unclear (6,7).
The recent determination of the x-ray crystal structures of the rat and human BRCA1 BRCT repeat domains were important first steps towards understanding tumourigenic BRCT mutations and provide a novel platform for the interpretation of the effects of these alterations in the absence of clinical data (15,28).
In the present study we directly evaluate the consequences missense mutation on the structure of the human BRCA1 BRCT repeats. Using a proteolysis-based assay to probe the BRCT for non-native conformations, we show that the majority of the tested missense and truncations alter the folding state of the BRCT. Cross-validated computational analyses using the BRCT structure and the sequences of proteins homologous to the human BRCT from EXPERIMENTAL PROCEDURES:

Computational analysis of risks associated with missense mutations
Method 1: Structure and sequence-based analysis. A probability of an effect on function for the missense mutations in BRCT was determined exactly as described using both feature set A and feature set B in reference (33). Briefly, the crystal structure (PDB ID: 1JNX), the multiple sequence alignment for proteins homologous to human BRCT (see Figure 3), and the chemical nature of the amino acid substitution are used to compute the values of features that are useful for predicting the effects of amino acid substitutions on protein function.
For example, the quantitative estimate of solvent accessibility for a residue in a structure or its normalized phylogenetic entropy from a multiple sequence alignment are both features that can be viewed as having a quantitative relationship to the probability of an effect on function for the introduction of a mutant amino acid (33). A probability of an effect on function for a test mutation is estimated by conditional probability as the fraction of training mutations derived from exhaustive mutagenesis of the Lac repressor (34) and T4 lysozyme (35) with an effect on function from among those with feature values are similar to the feature values of the test mutation. acids is evidence for its compatibility with biological function. Exclusion leads to a prediction of incompatibility.

Structural effects of BRCT truncation mutations
We previously demonstrated that the tandem BRCT repeat region of human BRCA1 forms a proteolytically resistant globular domain and that a cancer-linked mutation, Y1853ter, which removes the 11 C-terminal residues of the protein, reduces this proteolytic stability (13).
To determine to what extent the BRCT fold could tolerate truncation mutations, we subjected a series BRCT deletion mutants to a proteolytic sensitivity assay ( The deletion experiment also demonstrates the protein can tolerate removal of up to 8 residues, but further deletion from the C-terminus greatly impairs the native folding of the domain, rendering it highly sensitive to proteolysis ( Fig. 1, Fig. 2d). Consistent with this finding, the transcriptional activation activity of the BRCT domains was abolished by C-terminal deletions that truncate beyond a hydrophobic pair of residues, L1854 and I1855 (24).
These hydrophobes mark the C-terminal boundary for conservation of mammalian, avian and Xenopus BRCA1 homologues (Fig. 3), and make critical aliphatic contacts to the β-sheet of the C-terminal BRCT in the structures of the human and rat BRCA1-BRCT repeats ( Fig. 2d) (13,28). Hence, the transcriptional activation defects observed for BRCT deletion mutants likely result from destabilization of the protein.

Missense substitutions destabilize the BRCT
Similar to the truncation mutants, two cancer predisposing missense mutations, A1708E and M1775R, are destabilizing and exhibit altered BRCT protease susceptibility (13)(14)(15). To gain insights into the effects of other patientderived mutations recorded in the BIC, we generated 23 additional missense variants and tested these proteins for proteolytic sensitivity (see Experimental Procedures, Fig. 4). 20/25 of the missense mutations tested showed varying degrees of enhanced sensitivity to tryptic digestion at 20 °C (Fig. 4a). Five of six of the mutations that substitute an arginine into the protein (C1697R, S1715R, G1738R, P1749R and W1837R) also show increased sensitivity to chymotryptic cleavage at 20 °C (Fig. 4b) suggesting that destabilizing effects, rather than the introduction of a new trypsin cleavage site, are responsible for the protease sensitivity. Mutant M1775R is also clearly destabilizing (14) and shows sensitivity to chymotrypsin at elevated temperatures (15).
The expression levels of the BRCT variants in the reticulocyte lysates typically range between 0.3-1.2 fold of wild type levels. Since the expressed variants constitute less than 5-10% of the total protein digested in the lysates and we are using logarithmic increases in trypsin concentrations, we can quantify the percent protein remaining following digestion at each level of protease, and directly compare these values to establish a proteolysis-based hierarchy for the severity of the destabilizing effects (Fig. 5 (27). Conversely, R1699Q has little to no effect on BRCT structure and appears to have little effect on transcription activation (25).
iii. BRCA1-fold mutations: We have designated a third class of mutations as BRCA1-fold mutants. These substitutions include residues that participate in folding of the BRCT linker region, and residues that do not fall at BRCT-fold special positions, but are buried and conserved amongst BRCA1 homologues ( Fig. 5c). All of the tested BRCA1-fold class mutants tested alter the folding of the domain. The majority of these mutants (C1697R, S1715R, G1738R, G1738E, and P1749R) introduce charged residues into the protein core and are highly destabilizing. The A1752P mutant likely disrupts the linker helix and is highly destabilizing. The position of R1751 in the crystal structure of human BRCA1 is unclear, but the equivalent residue in the rat structure indicates this residue is involved in salt bridging interactions and the packing of BRCT linker helix (28).
This arginine is conserved amongst all known BRCA1 homologues and the R1751Q mutation may disrupt similar electrostatic stabilization in the human protein.

iv. BRCT-fold mutations:
A conserved hydrophobic clustering signature for the BRCT fold superfamily of proteins was originally identified using sequence based methods (39,40). Two of the three tested BRCT-fold class mutants that target residues that contribute to intra β-sheet packing, I1766S and V1809F, are destabilizing. The third β-sheet mutant, M1652I, has been classified as a benign polymorphism (41,42) and does not increase the protease sensitivity of the domain.

Mutations that destabilize the BRCT predispose carriers to disease
Pedigree analysis clarifying the disease predisposition of BRCA1 alleles is currently available for 8 of the 79 reported BRCA1-BRCT single amino acid substitution variants (  (Fig. 2).
Taken together, these results indicate that protease based detection of altered BRCT stability provides a novel and powerful predictive tool that can be used to assess disease linkage of BRCT mutations in instances where pedigree data is not available. Thus, we suggest that the 20 destabilizing missense mutants and truncations greater than eight amino acids are cancer-predisposing.

Predicting the structural consequences of mutation on the BRCT
The recent development of computational methods that incorporate detailed structural and sequence information to predict the effects of single amino acid substitutions on protein structure/function provides us with alternative tools to study the BRCT mutations (33, Chasman and Lau -submitted). We have applied two independent methods to predict the potential effects of the 25 missense mutations studied here (Table 1), and all known BRCT single amino acid substitutions recorded in the breast cancer information core (www.nhgri.nih.gov/Intramural_research/Lab_transfer/Bic/) (Supplementary Table 1).
In the first computational method, a set of quantitative and qualitative

Protein destabilization ablates BRCT mediated transcriptional activation
When tethered to a GAL4 DNA binding domain, the BRCT domains can activate transcription in yeast and mammalian systems (44)(45)(46). Significantly, potential targets of BRCA1 transcriptional regulation include the p53-responsive genes encoding p21 as well as GADD45 (47,48) suggesting that BRCA1 has a role in regulating DNA repair and checkpoint controls. The BRCT may modulate these functions through direct recruitment of the RNA polymerase holoenzyme (49,50), however, the physiological significance of these effects and the precise biochemical mechanism by which the BRCT activates transcription remains unclear (reviewed in 2). Nevertheless, this intrinsic activity forms the basis for a BRCA1 functional assay that has been used to probe for defects caused by several BRCT missense mutations (24,25,27).
Comparison of the transcription and protease based assay data reveal a striking correlation between destabilizing phenotypes and transcriptional defects (Table 1). That is, less stable BRCT variants including C1697R, R1699W, A1708E, S1715R, G1738E, and M1775R, as well as the truncation mutants, disrupt transactivation function, whereas mutations with no effect on structure (M1652I and R1699Q) are fully active in these assays. However, it has yet to be determined whether BRCT protein misfolding causes BRCA1 tumour suppressor inactivation via BRCA1 transcription function, DNA repair function, or both.

The Protease based assay for ranking BRCT destabilizing effects
Proteolytic degradation proceeds via an unfolded state for small globular proteins (51, 52) indicating that a correlation between proteolytic resistance and the thermodynamic stability of a protein may exist. This principle forms the basis for phage-based proteolytic selection methods where the evolution of proteins with increased thermodynamic stability closely follows the selection of polypeptides with enhanced resistance to degradation by increasing concentration of protease (53, 54). Thus, the application of a protease-based assay to assess the structural consequences of missense mutations on the BRCT provides a quick, effective, complimentary method to categorize and rank the extent of destabilization of the mutant BRCT proteins.
A recent biophysical assessment of the effects of 8 missense substitutions and the truncation Y1853ter on the thermodynamic stability of the BRCT revealed that four of these missense mutations and the truncation were highly destabilizing and could not be produced as soluble protein in E. coli (14).
All four of these missense mutants, A1708E, G1738E, G1788V and W1837R and the truncation show extreme sensitivity to tryptic digestion (Table 1 (Table 1). Altogether, these data indicate a three-tiered hierarchy of destabilizing effects inferred from the proteolytic data is consistent with results obtained from solubility analysis and direct thermodynamic measurements of BRCT protein stability. Highly destabilizing mutations show sensitivity to low levels of trypsin and tend to be degraded or insoluble when expressed in E. coli. Intermediate thermodynamically destabilizing mutations are sensitive to moderate levels of protease and can be produced in soluble form in The remaining set of mutations may not affect the folding detected by the proteolysis assay and yet still affect the functional properties of human BRCT.
The computational methods we have explored represent a first attempt to identify alternative correlates within this class of disease predisposing substitutions. The computational methods were largely consistent with the proteolysis data, whether or not there was an effect on protein stability. The purely sequence-based computational methodology was more consistent with the experimental evidence than the structure-and sequence-based approach. Whether the discrepancy can be interpreted or not remains to be seen through further studies. Methods for predicting the biological consequences of amino acid substitutions is an area of active research, especially since the genome initiatives are discovering too large a number of amino acid altering genetic variants with potential effects on biological function for experimental analysis (for example, see 55, 56).

Detection of BRCT mutations
The An adaptation of the PTT, where a protease digestion step is added could be appropriate for the detection of the large majority of cancer-associated BRCT mutations (Fig. 6a). Here, oligonucleotides would be specifically designed to amplify BRCT coding sequence (aa 1646-1863) from patient samples, and the translation step would be followed by a trypsinolysis series. This test would have the distinct advantage of sensing the protein destabilizing effects of both missense and truncation mutations. Conservative estimates indicate it could detect as much as 80% of the cancer-associated mutations that fall within the BRCT coding region. Alternatively, for cases where a BRCT missense mutation has already been detected by sequencing, the mutant BRCT coding sequence could be produced by PCR (Fig. 6b). Direct transcription/translation from the PCR product, followed by protein digestion, would provide a quick, relatively inexpensive test for mutant BRCT conformations. * To whom correspondence should be addressed. Research.

ABBREVIATIONS:
The abbreviations used are: BRCT, BRCA1 carboxy-terminal domain; BIC, Breast Cancer Information Core Database;