GATA-1 critique

Summary

    The gigantic strides which have been made in genome sequencing have ushered in a new era in molecular biology. However, if one were  to view the entire  molecular and biochemical machinery of a given organism as a beautiful painting inspired by nature, the genome sequence can be considered just an initial, albeit important, brush stroke.
    The human genome contains 3.2 trillion base pairs. Just within a single gene, the number of possible interactions involving DNA sequences of  short length with (for example) a proteinaceous transcription factor is very large. Thus, it is necessary to develop “mass production” techniques for surveying large regions of DNA for specific types of interactions.
    The authors of this paper seek to make an important contribution to biochemical methodology by demonstrating the utility of DNA microarray hybridization as they attempt to elucidate specific protein/DNA interactions over an entire 75 kb gene cluster, the K562-derived β-globin locus.
    The β-globin locus has been extensively studied as a paradigm of transcriptional regulation. The locus contains 5 genes (ε,γG,γA,δ,β), and one unexpressed pseudogene (ψβ) which code for the β-type subunits of hemoglobin. The regulatory mechanisms of the gene are interesting because the proper gene must be expressed at the proper time in development; in the normal adult, roughly 97% of β-type subunit in hemoglobin is β, while the developing fetus cycles through other products of the gene locus, and begins to express β only a few weeks before birth. Mutations in this locus can result in aberrant hemoglobin profiles which can cause anemic syndromes (thalassemias) or even death.
    The β-globin gene cluster contains a locus control region (LCR) which is located 6 - 25 kb to the 5′ side of the ε gene. Five sites on the LCR(HSs 1-5) are hypersensitive to DNase I, and thus are thought to be important in the regulation of the gene cluster by providing a “window” in the chromatin through which sequence and cell-type specific transcription factors can access DNA control sequences. Some controversy exists as to the exact role of the LCR in mediating transcriptional regulation and the mechanism by which this is accomplished. 
    GATA-1 is a transcription factor previously implicated in the regulation of the β-globin (as well as many other erythroid genes) locus in erythroid cells. GATA-1 is known to bind to the consensus sequence (A/T)GATA(A/G).  Protein footprinting studies using dimethyl sulfate (DMS) have indicated that certain GATA-1 consensus sequences within the β-globin locus are correlated with protein/DNA interactions, but such studies fail to positively identify what protein is involved.  Other research in K562 cells using chromatin immunoprecipitation (chIp) has identified HS2 and possibly HS3 of the LCR as sites of GATA-1 binding.
The authors conduct a comprehensive search of the β-globin locus in the erythroleukemic K562 cell line,  utilizing chromatin immunoprecipitation and DNA microarray hybridization techniques to identify portions of the β-globin gene locus which are involved in GATA-1 binding. The authors conclude that in addition to a previously known area of binding located in HS2, another previously unknown region of GATA-1 binding exists 5′ to the γG gene. More importantly,, they demonstrate that chromatin precipitation/DNA microarray hybridization (dubbed chIp-chip) can be a powerful method of surveying large regions of DNA for DNA/protein binding sites.

Evaluation

    1. Chromatin Immunoprecipitation and PCR Assay
    Human erythroleukemic K562 cells were treated with formaldehyde to effect crosslinking between binding proteins and chromatin. The cells were then lysed and the extracted chromatin was fragmented by sonication to roughly 500 bp segments. The chromatin fragments were then treated with antibodies to GATA-1 and immunoprecipitated using protein A/G agarose gel beads. The crosslinks were reversed and the DNA purified. Three different antibodies to GATA-1 were used to create three separate immunipurified DNA samples.
    The immunoprecipitated DNA was then amplified using PCR. Oligonucleotide primers used in PCR amplification  for HS1, HS2, and HS3 are listed in Table 2.
    HS1, HS2, and HS3 immunopurified PCR product were assayed on PAGE using mock-immunoprecipitant product as a negative control and untreated “input” DNA from lysated cells as a positive control. Level of DNA enrichment was estimated relative to that seen in the mock-immunoprecipitated products.
    The authors conclude from their assay(fig. 1b) that HS2 is the only region of immunopurified DNA tested which demonstrates significant binding to GATA-1. However, an examination of their data would seem to indicate a possibility that this conclusion is erroneous. Immunopurified HS3 DNA, while definitely exhibiting weaker enrichment than HS2 elements, has strong enough enrichment  to all three antibodies to GATA-1 used to cause us to question the conclusion. In addition to this, the arbitrary nature of their estimate of enrichment introduces another possible error factor. HS2 is assigned a fold enrichment factor of 2.8, while HS3 is given a 1.8.  How are we to know that 1.8 is truly insignificant? Is this because their microarray hybridization (see below) failed to demonstrate binding in the HS3 region?      Other authors have presented evidence of GATA-1 binding in HS3. The authors of this paper point out that there is no apparent correlation between number of consensus sequences in a given  fragment and fragment GATA-1 binding as indicated by microarray. It is clear that unknown and possibly hierarchical guiding principles are at work.  Perhaps the HS3 GATA-1 interactions (if they exist) require another mechanism which is being inhibited by the experimental process;. the sonication procedure may have sheared the DNA in the HS3 region such that GATA-1 binding was disrupted prior to immunoprecipitation.    Or, more simply, the cells were not grown in an environment which would render HS3 maximally binding to GATA-1.  A site not binding under one set of experimental conditions doesn’t mean it never does, especially when we are dealing with a gene locus which expresses and varies its products in a complex manner.  The authors use PCR assay data for the 10 ng input control of HS1, HS2, and HS3 as a justification for their conclusions. According to this logic, HS2 must be the only region binding to GATA-1 because the immunoprecipitated HS2 shows greater enrichment than the 10 ng input DNA sample, whereas areas HS1 and HS3 show less enrichment than their corresponding 10 ng input controls. While this appears to be true, one has to wonder why the 10 ng HS2 input assay shows such weak enrichment compared to HS3 and HS1 10 ng input DNAs. Why would HS2 input DNA amplify so poorly relative to HS3 and HS1 input? Also, why is it acceptable to use the 10 ng HS2 input assay result as a positive control when similar-appearing enrichment for immunoprecipitated HS3 derived from 1 ng of DNA is discounted out of hand as insignificant?

2. Use of chIp/chip to survey β-globin locus for GATA-1 binding sites

    The authors construct several DNA microarrays of the entire 75 kb β-globin locus, dividing DNA derived from bacterial artificial chromosomes (BAC) into 74 separate fragments of approximately 1kb size and amplifying the fragments using PCR. The DNA fragments were then spotted on each individual microarray a minimum of four times. Immunoprecipitated DNA  was labeled with Cy3, a red fluorescent dye; in addition to this, background DNA (“control DNA”) was labeled with Cy5, a green fluorescent dye. The labeled DNA was then hybridized to the microarrays, and the intensity of fluorescence was measured as a median ratio Cy3/Cy5.
    Microarrays were constructed for exposure to all three forms of GATA-1 immunoprecipitated DNA, as well as to various controls. Consistent results were obtained throughout. Two regions of the β-globin locus were found to exhibit intense Cy3 fluorescence. These were the previously suspected HS2 region (fragment 009BG) and a previously unknown region of GATA-1 binding located 5′ to the γG gene (fragment 032BG). Notably, fragments containing HS3 failed to exhibit intense Cy3 fluorescence (fig. 2a and 2b). The authors note that microarrays hybridized with DNA immunoprecipitated with antibodies to other known β-globin transcription factors (Nrf1, CBP) demonstrated different binding profiles, and that microarrays hybridized with mock-immunoprecipitated DNA fail to display increased Cy3/Cy5 signal intensity in any locus region.
    The authors admit that several microarrays demonstrated anomalous and unreproducible Cy3/Cy5 signal foci in various regions of the locus. It would appear from error estimates on figure 2 (a and b) that in some cases these aberrations were quite large.They postulate that this is due to GATA-1 possibly binding to an area of repetitive DNA, and the homology between these sequences and those of the array create these interactions.
    In this particular experiment, the most one can reasonably conclude is that a GATA-1 binding site probably exists 5′ to the γG gene. This result does not exclude GATA-1 binding in other portions of the locus (other than the HS2 site). It is possible if not likely that DNA immunoprecipitated from K562 cells grown under different environmental conditions and at different stages in development would demonstrate different binding patterns on chIp/chip.
    The lack of Cy3/Cy5 signal intensity in HS3 fragments on the microarrays is problematic considering the criticisms mentioned above. It may be, as the authors claim, that HS3 truly has no GATA-1 binding activity. Perhaps the methods used to amplify and hybridize immunoprecipitated DNA to the microarray contain a step which somehow alters HS3 fragments and precludes them from hybridizing properly, or Cy5-labeled background DNA is interfering with the hybridization process. Both of these explanations, however, seem to be very unlikely. 

3. PCR Assay of GATA-1 binding sites 009BG and 032BG
    After identifying GATA-1 consensus sequences within the entire locus and determining that there is no connection between consensus site density and GATA-1 binding as determined by microarray, the authors turn to closely examining binding sites 009BG and 032BG by comparing enrichment of 250-300bp fragments of anti-GATA-1 immunoprecipitated DNA with mock-immunoprecipitated fragments.
    The PAGE assay (fig.3) for region 009BG shows significant enrichment as compared to controls (value 2.9) in the middle part of the fragment, which does correspond to the one GATA-1 consensus sequence  which is found within the fragment. Flanking regions of this consensus sequence are not significantly enriched, as demonstrated on the assay results.
    A similar assay is conducted for region 032BG. The authors conclude that PCR assay of 5 fragments displayed significant enrichment in areas -1557 to -1289 (value 2.8) and -1055 to -822(value 4.8) from the γG start codon (fig.4). The -1557 to -1289 region has 3 GATA-1 consensus sequences; the highly enriched area from -1055 to -822 contains two.
    The data for this assay is much clearer than for the earlier PCR assay; the criticisms, however, are similar. Is an enrichment value of 1.8 for one of the 032BG fragments really insignificant? The authors do provide a logical explanation for the segment of 032BG which has an enrichment value of 2.0 (fragment overlap with an adjacent binding site).
    It is noteworthy that, while also assaying several other anti-GATA-1 immunoprecipitates representing different parts of the locus  with this focused technique, the authors conspicuously excluded HS3 from analysis. It would have been interesting to see how  immunoprecipitates were enriched using HS3 primer given the results of the earlier PCR assay, and to compare this result to the specific DNA sequences of the HS3 fragments.

Discussion

  The methodology performed in this experiment is more important than the result; clearly, a new and previously unsuspected part of the  β-globin locus which binds to GATA-1 has been identified, and this identification was made possible by the use of the microarray technique. Before the development of such microarrays, the vast size of the genome dictated that scientists limit their investigations to gene sequences which were suspected of interacting with proteins, so as to guarantee that such expensive and time-consuming experiments would produce meaningful results.  One wonders how long it would have taken to identify fragment 032BG as an area of GATA-1 binding using more traditional methods.
The use of the microarray allows a researcher to sample broad areas of genes for protein/DNA interactions; the limitation of such a technique is that, while providing tempting clues and insights into previously unsuspected interactions, definite answers are not provided.
    While the authors present good evidence with respect to GATA-1 binding in regions 009BG and 032BG, other conclusions may well be suspect. The use of hybridized erythroleukemic cells and the complexity of β-globin regulation mandate that we must be careful in interpreting results and dogmatically applying them to β-globin function in the living vertebrate. Other cell lines used in GATA-1 research have demonstrated different GATA-1 binding profiles than what is shown here.
    The identification of region 032BG as GATA-1 binding may be of assistance in helping to decipher how the LCR functions to regulate the gene. Could region 032BG be involved in this process? A mutant could be developed which has the LCR and region 032BG deleted and chromatin sensitivity measured as compared to the known behavior of mutants possessing the LCR deletion alone or in combination with the Hispanic deletion which has been shown to have a strong negative effect on chromatin sensitivity.
    A closer look at region HS3 may be of benefit. In particular, chIp-PCR assay of 200-300 bp fragments from region HS3 using antibodies to GATA-1 would help in the ascertainment of whether any consensus sequence binding sites are active. This experiment should be run on not just DNA from the K562 line, but other cell lines as well, as the expression patterns of different cell lines is most likely different.  An interesting variation of this “focused” experiment would be to examine the entire β-globin locus using incrementally larger fragments of immunoprecipitated DNA, for instance1000 bp and 2000 bp segments as opposed to 500 bp and 300 bp. The authors of this paper point to larger fragments of DNA as being a possible source of error in experiments which claimed to show HS3/ GATA-1 binding, but their use of smaller fragments may itself be creating a false impression. While information derived from such experiments would be less specific, it may help identify whether GATA-1 binding in some instances requires more complex direction from multiple, more distant sites, such as the type of interactions hypothesized as part of the looping or linking models of LCR regulation of the gene cluster and which could involve GATA-1.  The use of larger fragments of immunoprecipitated DNA could be applied to a microarray experiment; indeed one can envision a whole series of PCR assay/  microarray experiments in which the variable is the fragment length of immunoprecipitated DNA. Adding another variable (mutants which have various portions of the β-globin locus deleted) in combination with this could yield even more information. This would be very expensive and time intensive, but may shed light on not just GATA-1 binding in the β-globin locus, but the binding of transcription factors to DNA in general.