A comparison of three methods for detecting KRAS mutations in formalin-fixed colorectal cancer specimens.

BACKGROUND: KRAS mutation testing is required to select patients with metastatic colorectal cancer (CRC) to receive anti-epidermal growth factor receptor antibodies, but the optimal KRAS mutation test method is uncertain. METHODS: We conducted a two-site comparison of two commercial KRAS mutation kits - the cobas KRAS Mutation Test and the Qiagen therascreen KRAS Kit - and Sanger sequencing. A panel of 120 CRC specimens was tested with all three methods. The agreement between the cobas test and each of the other methods was assessed. Specimens with discordant results were subjected to quantitative massively parallel pyrosequencing (MPP). DNA blends were tested to determine detection rates at 5% mutant alleles. RESULTS: Reproducibility of the cobas test between sites was 98%. Six mutations were detected by cobas that were not detected by Sanger, and five were confirmed by MPP. The cobas test detected eight mutations which were not detected by the therascreen test, and seven were confirmed by MPP. Detection rates with 5% mutant DNA blends were 100% for the cobas and therascreen tests and 19% for Sanger. CONCLUSION: The cobas test was reproducible between sites, and detected several mutations that were not detected by the therascreen test or Sanger. Sanger sequencing had poor sensitivity for low levels of mutation.

Tamoxifen metabolite concentrations, CYP2D6 genotype, and breast cancer outcomes.

We explored whether breast cancer outcomes are associated with endoxifen and other metabolites of tamoxifen and examined potential correlates of endoxifen concentration levels in serum including cytochrome P450 2D6 (CYP2D6) metabolizer phenotype and body mass index (BMI). Concentration levels of tamoxifen, endoxifen, 4-hydroxytamoxifen (4OH-tamoxifen), and N-desmethyltamoxifen (ND-tamoxifen) were measured from samples taken from 1,370 patients with estrogen receptor (ER)-positive breast cancer who were participating in the Women's Healthy Eating and Living (WHEL) Study. We tested these concentration levels for possible associations with breast cancer outcomes and found that breast cancer outcomes were not associated with the concentration levels of tamoxifen, 4-hydroxytamoxifen, and ND-tamoxifen. For endoxifen, a threshold was identified, with women in the upper four quintiles of endoxifen concentration appearing to have a 26% lower recurrence rate than women in the bottom quintile (hazard ratio (HR) = 0.74; 95% confidence interval (CI), (0.55-1.00)). The predictors of this higher-risk bottom quintile were poor/intermediate metabolizer genotype, higher BMI, and lower tamoxifen concentrations as compared with the mean for the cohort as a whole. This study suggests that there is a minimal concentration threshold above which endoxifen is effective against the recurrence of breast cancer and that ~80% of tamoxifen takers attain this threshold.

Measuring stem cell frequency in epidermis: a quantitative in vivo functional assay for long-term repopulating cells.

Epidermal stem cells play a central role in tissue homeostasis, wound repair, tumor initiation, and gene therapy. A major impediment to the purification and molecular characterization of epidermal stem cells is the lack of a quantitative assay for cells capable of long-term repopulation in vivo, such as exists for hematopoietic cells. The tremendous strides made in the characterization and purification of hematopoietic stem cells have been critically dependent on the availability of competitive transplantation assays, because these assays permit the accurate quantitation of long-term repopulating cells in vivo. We have developed an analogous functional assay for epidermal stem cells, and have measured the frequency of functional epidermal stem cells in interfollicular epidermis. These studies indicate that cells capable of long-term reconstitution of a squamous epithelium reside in the interfollicular epidermis. We find that the frequency of these long-term repopulating cells is 1 in 35,000 total epidermal cells, or in the order of 1 in 104 basal epidermal cells, similar to that of hematopoietic stem cells in the bone marrow, and much lower than previously estimated in epidermis. Furthermore, these studies establish a novel functional assay that can be used to validate immunophenotypic markers and enrichment strategies for epidermal stem cells, and to quantify epidermal stem cells in various keratinocyte populations. Thus further studies using this type of assay for epidermis should aid in the progress of cutaneous stem cell-targeted gene therapy, and in more basic studies of epidermal stem cell regulation and differentiation.

The HOX homeodomain proteins block CBP histone acetyltransferase activity.

Despite the identification of PBC proteins as cofactors that provide DNA affinity and binding specificity for the HOX homeodomain proteins, HOX proteins do not demonstrate robust activity in transient-transcription assays and few authentic downstream targets have been identified for these putative transcription factors. During a search for additional cofactors, we established that each of the 14 HOX proteins tested, from 11 separate paralog groups, binds to CBP or p300. All six isolated homeodomain fragments tested bind to CBP, suggesting that the homeodomain is a common site of interaction. Surprisingly, CBP-p300 does not form DNA binding complexes with the HOX proteins but instead prevents their binding to DNA. The HOX proteins are not substrates for CBP histone acetyltransferase (HAT) but instead inhibit the activity of CBP in both in vitro and in vivo systems. These mutually inhibitory interactions are reflected by the inability of CBP to potentiate the low levels of gene activation induced by HOX proteins in a range of reporter assays. We propose two models for HOX protein function: (i) HOX proteins may function without CBP HAT to regulate transcription as cooperative DNA binding molecules with PBX, MEIS, or other cofactors, and (ii) the HOX proteins may inhibit CBP HAT activity and thus function as repressors of gene transcription.

Positive and negative regulation of granulopoiesis by endogenous RARalpha.

Acute promyelocytic leukemia (APL) is always associated with chromosomal translocations that disrupt the retinoic acid receptor alpha (RARalpha) gene. Whether these translocations relate to a role for endogenous RARalpha in normal granulopoiesis remains uncertain because most studies addressing this question have used non-physiological overexpression systems. Granulocyte differentiation in cells derived from RARalpha-deficient (RARalpha(-/-)) mice was studied and evaluated in the context of agonist-bound and ligand-free RARalpha. Our results demonstrate that RARalpha is dispensable for granulopoiesis, as RARalpha(-/-) mice have a normal granulocyte population despite an impaired ability to respond to retinoids. However, although it is not absolutely required, RARalpha can bidirectionally modulate granulopoiesis. RARalpha stimulates differentiation in response to exogenous retinoic acid. Furthermore, endogenous retinoids control granulopoiesis in vivo, as either vitamin A-deficient mice or animals treated with an RAR antagonist accumulate more immature granulocytes in their bone marrow. Conversely, RARalpha acts to limit differentiation in the absence of ligand because granulocyte precursors from RARalpha(-/-) mice differentiate earlier in culture. Thus, the block in granulopoiesis exerted by RARalpha fusion proteins expressed in APL cells may correspond to an amplification of a normal function of unliganded RARalpha.

Frequent co-expression of the HOXA9 and MEIS1 homeobox genes in human myeloid leukemias.

There is increasing evidence that HOX homeobox genes play a role in leukemogenesis. Recent studies have demonstrated that enforced co-expression of HOXA9 and MEIS1 in murine marrow leads to rapid development of myeloid leukemia, and that these proteins exhibit cooperative DNA binding. However, it is unclear whether co-activation of HOXA9 and MEIS genes is a common occurrence in human leukemias. We surveyed expression of HOXA9 and MEIS1 in 24 leukemic cell lines and 80 patient samples, using RNase protection analyses and immunohistochemistry. We demonstrate that the expression of HOXA9 and MEIS1 in leukemia cells is uniquely myeloid, and that these genes are commonly co-expressed in myeloid cell lines and in samples of acute myelogenous leukemia (AML) of all subtypes except in promyelocytic leukemia. While HOXA9 is expressed in most cases of chronic myelogenous leukemia, MEIS1 is weakly expressed or not at all. Immunohistochemical staining of selected AML samples showed moderate to high levels of HOXA9 protein, primarily cytoplasmic, in leukemic myeloblasts, with weaker and primarily nuclear staining for MEIS1. These data support the concept that co-activation of HOXA9 and MEIS1 is a common event in AML, and may represent a common pathway of many different oncogenic mutations.

HOXA9 forms triple complexes with PBX2 and MEIS1 in myeloid cells.

Aberrant activation of the HOX, MEIS, and PBX homeodomain protein families is associated with leukemias, and retrovirally driven coexpression of HOXA9 and MEIS1 is sufficient to induce myeloid leukemia in mice. Previous studies have demonstrated that HOX-9 and HOX-10 paralog proteins are unique among HOX homeodomain proteins in their capacity to form in vitro cooperative DNA binding complexes with either the PBX or MEIS protein. Furthermore, PBX and MEIS proteins have been shown to form in vivo heterodimeric DNA binding complexes with each other. We now show that in vitro DNA site selection for MEIS1 in the presence of HOXA9 and PBX yields a consensus PBX-HOXA9 site. MEIS1 enhances in vitro HOXA9-PBX protein complex formation in the absence of DNA and forms a trimeric electrophoretic mobility shift assay (EMSA) complex with these proteins on an oligonucleotide containing a PBX-HOXA9 site. Myeloid cell nuclear extracts produce EMSA complexes which appear to contain HOXA9, PBX2, and MEIS1, while immunoprecipitation of HOXA9 from these extracts results in coprecipitation of PBX2 and MEIS1. In myeloid cells, HOXA9, MEIS1, and PBX2 are all strongly expressed in the nucleus, where a portion of their signals are colocalized within nuclear speckles. However, cotransfection of HOXA9 and PBX2 with or without MEIS1 minimally influences transcription of a reporter gene containing multiple PBX-HOXA9 binding sites. Taken together, these data suggest that in myeloid leukemia cells MEIS1 forms trimeric complexes with PBX and HOXA9, which in turn can bind to consensus PBX-HOXA9 DNA targets.

Loss of function of the homeobox gene Hoxa-9 perturbs early T-cell development and induces apoptosis in primitive thymocytes.

Hox homeobox genes play a crucial role in specifying the embryonic body pattern. However, a role for Hox genes in T-cell development has not been explored. The Hoxa-9 gene is expressed in normal adult and fetal thymuses. Fetal thymuses of mice homozygous for an interruption of the Hoxa-9 gene are one eighth normal size and have a 25-fold decrease in the number of primitive thymocytes expressing the interleukin-2 receptor (IL-2R, CD25). Progression to the double positive (CD4+CD8+) stage is dramatically retarded in fetal thymic organ cultures. This aberrant development is associated with decreased amounts of intracellular CD3 and T-cell receptor beta (TCRbeta) and reduced surface expression of IL-7R and E-cadherin. Mutant thymocytes show a significant increase in apoptotic cell death and premature downregulation of bcl-2 expression. A similar phenotype is seen in primitive thymocytes from adult Hoxa-9-/- mice and from mice transplanted with Hoxa-9-/- marrow. Hoxa-9 appears to play a previously unsuspected role in T-cell ontogeny by modulating cell survival of early thymocytes and by regulating their subsequent differentiation.

HOX homeobox genes exhibit spatial and temporal changes in expression during human skin development.

The spatial and temporal deployment of HOX homeobox genes along the spinal axis and in limb buds during fetal development is a key program in embryonic pattern formation. Although we have previously reported that several of the HOX homeobox genes are expressed during murine skin development, there is no information about developmental expression of HOX genes in human skin. We have now used reverse transcriptase polymerase chain reaction, in conjunction with a set of degenerate oligonucleotide primers, to identify a subset of HOX genes that are expressed during human fetal skin development. In situ hybridization analyses demonstrated that there were temporal and spatial shifts in expression of these genes. Strong HOXA4 expression was detected in the basal cell layers of 10 wk fetal epidermis and throughout the epidermis and dermis of 17 wk skin, whereas weak signal was present in the granular layer of newborn and adult skin. The expression patterns of HOXA5 and HOXA7 were similar, but their expression was weaker. In situ hybridization analysis also revealed strong HOXC4 and weaker HOXB7 expression throughout fetal development, whereas HOXB4 was expressed at barely detectable levels. Differential HOX gene expression was also observed in developing hair follicles, and sebaceous and sweat glands. None of the HOX genes examined were detected in the adult dermis.

Effects of HOX homeobox genes in blood cell differentiation.

The burgeoning number of articles concerning the role of HOX genes and hematopoiesis ensures that this will continue to be an area of very active research. It seems clear that HOX genes are expressed in stage- and lineage-specific patterns during early stages of hematopoietic development and differentiation. Several lines of evidence suggest that multiple genes of the HOXB (B2, B4, B6-B9), HOXC (C6, C8), and HOXA (A5) are involved in erythropoiesis. Similarly, a number of genes of the HOXA, HOXB, and HOXC appear to play a role in lymphoid cells. Furthermore, several genes, such as A9, A10, B3, B7, and B8, may control myelomonocytic differentiation. The question arises as to whether such a multiplicity of HOX genes reflects redundancy or indicates subtlety of the regulatory machinary. A similar complexity has been observed for hematopoietic cytokines, and the current view is that, although multiple molecules may have similar or overlapping effects, each factor has a specific function and regulatory combinations appear to play a critical role in controlling hematopoietic cell processes (99). One challenge for the future is to delineate in more detail the precise expression patterns of these genes in the many distinct subpopulations of blood cells and during fetal development. Overexpression of HOX genes in hematopoietic cells can dramatically perturb the differentiation of various cell lineages and can contribute to leukemogenesis. Future studies may involve the overexpression of alternatively spliced versions of different HOX genes or of truncated versions of HOX genes to ascertain the functional domains of the proteins that mediate the biologic effects. The findings in HOX knockout mice confirm a role for these genes in normal blood cell development. Further work in this area will require careful examination of fetal hematopoiesis and of animals bearing multiple HOX gene knockouts. Involvement of HOX genes in leukemia is just beginning to be appreciated. Establishing the true extent of HOX gene mutations in human disease will require strategies such as comparative genomic hybridization (100) and analysis of high density oligonucleotide arrays (101). The holy grail of homeobox work is to discover the physiologic processes and specific target genes regulated by HOX proteins. Given the broad range of tissues in which HOX genes are expressed, they would appear to be involved in very basic cellular processes, e.g., cell proliferation and death, adhesion, and migration, etc., rather than the direct regulation of tissue-specific genes. The search for target genes may be made easier by the further characterization of cooperative DNA binding between HOX proteins and other transcription factors. We speculate that HOX proteins do not behave as conventional transcriptional activators or inhibitors but rather may mark genes for potential future activation, i.e., they may establish competency to execute specific differentiation programs, with the actual activation being accomplished by transcriptional pathways triggered by exogenous signals. This proposed function may be an architectural one, involving changes in the conformation of DNA and/or altering interactions between DNA and histones, thus making areas of the genome more or less accessible to other protein factors (102). If this is the case, we may need to develop new assays to discern the molecular action of HOX proteins. The ease of manipulating the hematopoietic systems would appear to make it a very attractive model for explicating the general functions of this remarkable family of genes.

AbdB-like Hox proteins stabilize DNA binding by the Meis1 homeodomain proteins.

Recent studies show that Hox homeodomain proteins from paralog groups 1 to 10 gain DNA binding specificity and affinity through cooperative binding with the divergent homeodomain protein Pbx1. However, the AbdB-like Hox proteins from paralogs 11, 12, and 13 do not interact with Pbx1a, raising the possibility of different protein partners. The Meis1 homeobox gene has 44% identity to Pbx within the homeodomain and was identified as a common site of viral integration in myeloid leukemias arising in BXH-2 mice. These integrations result in constitutive activation of Meis1. Furthermore, the Hoxa-9 gene is frequently activated by viral integration in the same BXH-2 leukemias, suggesting a biological synergy between these two distinct classes of homeodomain proteins in causing malignant transformation. We now show that the Hoxa-9 protein physically interacts with Meis1 proteins by forming heterodimeric binding complexes on a DNA target containing a Meis1 site (TGACAG) and an AbdB-like Hox site (TTTTACGAC). Hox proteins from the other AbdB-like paralogs, Hoxa-10, Hoxa-11, Hoxd-12, and Hoxb-13, also form DNA binding complexes with Meis1b, while Hox proteins from other paralogs do not appear to interact with Meis1 proteins. DNA binding complexes formed by Meis1 with Hox proteins dissociate much more slowly than DNA complexes with Meis1 alone, suggesting that Hox proteins stabilize the interactions of Meis1 proteins with their DNA targets.

Mice bearing a targeted interruption of the homeobox gene HOXA9 have defects in myeloid, erythroid, and lymphoid hematopoiesis.

Several homeobox genes of the HOXA and HOXB clusters are expressed in primitive blood cells, suggesting a role for HOX genes in normal hematopoiesis. The HOXA9 gene is expressed in CD34+ marrow cells and in developing lymphocytes. We examined blood-forming organs of mice homozygous for an interrupted HOXA9 allele to determine if loss of HOX gene function is deleterious to hematopoiesis. HOXA9-/- mice have approximately 30% to 40% reductions in total leukocytes and lymphocytes (P < .001) and a blunted granulocytic response to granulocyte colony-stimulating factor (G-CSF). Homozygous mice have significantly smaller spleens and thymuses. Myeloid/erythroid and pre-B progenitors in the marrow are significantly reduced, but no significant decreases are noted in mixed colonies, day 12 colony-forming units-spleen (CFU-S), or long-term culture-initiating cells (LTC-IC), suggesting little or no perturbation in earlier progenitors. Heterozygous animals display no hematopoietic defects. The abnormalities in leukocyte production are transplantable, indicating that the defect resides in the hematopoietic cells. These studies demonstrate a physiologic role for a HOX gene in blood cell differentiation, with the greatest apparent influence of HOXA9 at the level of the committed progenitor.

The Abd-B-like Hox homeodomain proteins can be subdivided by the ability to form complexes with Pbx1a on a novel DNA target.

Previous studies showed that the Hox homeodomain proteins from paralog groups 1-8 display cooperative DNA binding with the non-Hox homeodomain protein Pbx, mediated by a canonical YPWM. Although the Abd-B-like Hox proteins in paralogs 9-13 lack this sequence, Hoxb-9 and Hoxa-10 were reported to bind with Pbx1a to DNA. We show that these interactions require a tryptophan 6 amino acids N-terminal to the homeodomain. Binding site selection for Hoxb-9 with Pbx1a yielded ATGATTTACGAC, containing a novel TTAC Hox-binding site adjacent to a Pbx site. In the presence of Pbx1a, Hoxb-9 and Hoxa-10 bound to targets containing either TTAC or TTAT. These data extend previous findings that interactions with Pbx define a Hox protein binding code for different DNA sequences across paralog groups 1 through 10. Members of the 11, 12, and 13 paralogs do not cooperatively bind DNA with Pbx1a, despite the presence of tryptophan residues N-terminal to the homeodomain in Hoxd-12 and Hoxd-13. Hoxa-11, Hoxd-12, or Hoxd-13, in the presence of Pbx1a, selected a TTAC Hox site but lacking a Pbx1a site. These data suggest that Abd-B-like Hox proteins bind to a novel TTAC site and can be divided by their cooperative binding to DNA with Pbx1a.

Overexpression of HOXA10 in murine hematopoietic cells perturbs both myeloid and lymphoid differentiation and leads to acute myeloid leukemia.

Multiple members of the A, B, and C clusters of Hox genes are expressed in hematopoietic cells. Several of these Hox genes have been found to display distinctive expression patterns, with genes located at the 3' side of the clusters being expressed at their highest levels in the most primitive subpopulation of human CD34+ bone marrow cells and genes located at the 5' end having a broader range of expression, with downregulation at later stages of hematopoietic differentiation. To explore if these patterns reflect different functional activities, we have retrovirally engineered the overexpression of a 5'-located gene, HOXA10, in murine bone marrow cells and demonstrate effects strikingly different from those induced by overexpression of a 3'-located gene, HOXB4. In contrast to HOXB4, which causes selective expansion of primitive hematopoietic cells without altering their differentiation, overexpression of HOXA10 profoundly perturbed myeloid and B-lymphoid differentiation. The bone marrow of mice reconstituted with HOXA10-transduced bone marrow cells contained in high frequency a unique progenitor cell with megakaryocytic colony-forming ability and was virtually devoid of unilineage macrophage and pre-B-lymphoid progenitor cells derived from the transduced cells. Moreover, and again in contrast to HOXB4, a significant proportion of HOXA10 mice developed a transplantable acute myeloid leukemia with a latency of 19 to 50 weeks. These results thus add to recognition of Hox genes as important regulators of hematopoiesis and provide important new evidence of Hox gene-specific functions that may correlate with their normal expression pattern.

Overexpression of HOXB3 in hematopoietic cells causes defective lymphoid development and progressive myeloproliferation.

HOXB3 mRNA levels are high in the earliest CD34+ lineage- bone marrow cells and low to undetectable in later CD34+/CD34- cells. To gain some insight into the role this gene may play in hematopoiesis, HOXB3 was overexpressed in murine bone marrow cells using retroviral gene transfer. Thymi of HOXB3 marrow recipients were reduced in size compared with control transplant recipients, with a 24-fold decrease in the absolute number of CD4+ CD8+ cells and a 3-fold increase in the number of CD4- CD8- thymocytes that contained a high proportion of gammadelta TCR+ cells. B cell differentiation was also perturbed in these mice, as indicated by the virtual absence of transduced IL-7-responsive pre-B clonogenic progenitors. Recipients of HOXB3-transduced cells also had elevated numbers of mature granulocyte macrophage colony-forming cells in their bone marrow and spleen. Together these results suggest roles for HOXB3 in proliferation and differentiation processes of both early myeloid and lymphoid developmental pathways.

The role of HOX homeobox genes in normal and leukemic hematopoiesis.

A sizable amount of new data points to a role for the HOX family of homeobox genes in hematopoiesis. Recent studies have demonstrated that HOXA and HOXB genes are expressed in human CD34+ cells, and are downregulated as cells leave the CD34+ compartment. In addition, expression of certain genes, including HOXB3 and HOXB4, is largely restricted to the long-term culture-initiating cell enriched pool, containing the putative stem cell population. Studies have also shown that HOX genes appear to be important for normal T lymphocyte and activated natural killer cell function. Overexpression of Hox-b4 in transplanted murine marrow cell results in a dramatic expansion of stem cells, while maintaining normal peripheral blood counts. In contrast, overexpression of Hox-a10 resulted in expansion of progenitor pools, accompanied by unique changes in the differentiation patterns of committed progenitors. Overexpression of Hox-a10 or Hox-b8 led to the development of myeloid leukemias, while animals transfected with marrow cells overexpressing Hox-b4 do not appear to develop malignancies. Blockade of HOX gene function using antisense oligonucleotides has revealed that several HOX genes appear to influence either myeloid or erythroid colony formation. Mice homozygous for a targeted disruption of the HOX-a9 gene show reduced numbers of granulocytes and lymphocytes, smaller spleens and thymuses, and reduced numbers of committed progenitors. These studies demonstrate that HOX homeobox genes play a role in both the early stem cell function as well as in later stages of hematopoietic differentiation, and that perturbations of HOX gene expression can be leukemogenic.

Overexpression of HOXB4 enhances the hematopoietic potential of embryonic stem cells differentiated in vitro.

Little is known about the molecular mechanisms controlling primitive hematopoietic stem cells, especially during embryogenesis. Homeobox genes encode a family of transcription factors that have gained increasing attention as master regulators of developmental processes and recently have been implicated in the differentiation and proliferation of hematopoietic cells. Several Hox homeobox genes are now known to be differentially expressed in various subpopulations of human hematopoietic cells and one such gene, HOXB4, has recently been shown to positively determine the proliferative potential of primitive murine bone marrow cells, including cells with long-term repopulating ability. To determine if this gene might influence hematopoiesis at the earliest stages of development, embryonic stem (ES) cells were genetically modified by retroviral gene transfer to overexpress HOXB4 and the effect on their in vitro differentiation was examined. HOXB4 overexpression significantly increased the number of progenitors of mixed erythroid/myeloid colonies and definitive, but not primitive, erythroid colonies derived from embryoid bodies (EBs) at various stages after induction of differentiation. There appeared to be no significant effect on the generation of granulocytic or monocytic progenitors, nor on the efficiency of EB formation or growth rate. Analysis of mRNA from EBs derived from HOXB4-transduced ES cells on different days of primary differentiation showed a significant increase in adult beta-globin expression, with no detectable effect on GATA-1 or embryonic globin (beta H-1). Thus, HOXB4 enhances the erythropoietic, and possibly more primitive, hematopoietic differentiative potential of ES cells. These results provide new evidence implicating Hox genes in the control of very early stages in the development of the hematopoietic system and highlight the utility of the ES model for gaining insights into the molecular genetic regulation of differentiation and proliferation events.

Hox homeodomain proteins exhibit selective complex stabilities with Pbx and DNA.

Eight of the nine homeobox genes of the Hoxb locus encode proteins which contain a conserved hexapeptide motif upstream from the homeodomain. All eight proteins (Hoxb-1-Hoxb-8) bind to a target oligonucleotide in the presence of Pbx1a under conditions where minimal or no binding is detected for the Hox or Pbx1a proteins alone. The stabilities of the Hox-Pbx1a-DNA complexes vary >100-fold, with the proteins from the middle of the locus (Hoxb-5 and Hoxb-6) forming very stable complexes, while Hoxb-4, Hoxb-7 and Hoxb-8 form complexes of intermediate stability and proteins at the 3'-side of the locus (Hoxb-1-Hoxb-3) form complexes which are very unstable. Although Hox-b proteins containing longer linker sequences between the hexapeptide and homeodomains formed unstable complexes, shortening the linker did not confer complex stability. Homeodomain swapping experiments revealed that this motif does not independently determine complex stability. Naturally occurring variations within the hexapeptides of specific Hox proteins also do not explain complex stability differences. However, two core amino acids (tryptophan and methionine) which are absolutely conserved within the hexapeptide domains appear to be required for complex formation. Removal of N- and C-terminal flanking regions did not influence complex stability and the members of paralog group 4 (Hoxa-4, b-4, c-4 and d-4), which share highly conserved hexapeptides, linkers and homeodomains but different flanking regions, form complexes of similar stability. These data suggest that the structural features of Hox proteins which determine Hox-Pbx1a-DNA complex stability reside within the precise structural relationships between the homeodomain, hexapeptide and linker regions.