Introduction
Cyanobacteria, or blue-green algae, are filamentous organisms with a remarkable trait. Most can perform both photosynthesis and nitrogen fixation, which are mutually exclusive properties, since oxygen inhibits nitrogenase. In order to do so when the organism is deprived of nitrogen, some of the cells within a filament differentiate to form thicker walls. These cells, which do not photosynthesize, are called heterocysts-they only fix nitrogen. The process of differentiation is the big fascinating picture…how does a filament decide which cells will differentiate? It has been observed that upon nitrogen deprivation, clumps of cells all begin to differentiate. At a certain time point, one of these cells releases a lateral inhibitor that prevents the cells around it from becoming heterocysts, and they go back to being vegetative. This process leads to a pattern of about one in every ten vegetative cells becoming a heterocyst.

But that still doesn't totally answer the question about why certain cells differentiate-why does one of the cells release the inhibitor and not others? One option is that heterocyst formation is linked to something that every cell has, but could be different in each cell. This brings up the possibility of cell cycle as linked to differentiation, which can explain why clumps of cells all start to change together…they are more closely related to each other than to other cells in the filament. But then, what controls cell cycle? One of the mechanisms to control a cell's cycle has been shown to be methylation of the DNA. Methyl groups added to individual base pairs in the gene OriC of E. coli control the initiation of DNA replication, as methyltransferase-negative strains are unable to coordinate initiations.

The majority of methyltransferases are present in order to protect the organism's DNA from digestion by restriction enzymes. The restriction enzymes cut DNA at certain sequences of base pairs. If there is a methyl group present, the enzyme will not cut the sequence. This is a mechanism by which cells protect themselves from foreign DNA, since foreign DNA will not have the same methylation as a cell's own. Anabaena, a strain of cyanobacterium, has been found to have 5 identifiable methyltransferases, which seems to be a rather large number, especially since most of these do not have matching restriction enzymes. These so-called orphan methylases are the ones that have been implicated in initiation control in E. coli (dam methylase). In another strain of bacteria, Streptococcus pneumonia, there are two methylases which recognize and place methyl groups on the same sequence of base pairs. However, the difference between the two is that one can methylate single-stranded, while the other methylates double-stranded DNA.

In Anabaena, there are also two methylases which both act on the sequence GATC, placing the methyl group on the adenine of that sequence. They are encoded by genes called dmnA and dmnD. The DmnA protein doesn't seem to fit structurally into any of the previously known groups or methylases, so it might be thought that it could be a defunct methylase that was recently incorporated into the organism's genome. However, knockout mutants of the gene were unable to survive, indicating that the gene was most likely vital. So the question remains; what exactly is one organism doing with two proteins methylating the same sequence?

The general goal of my research, then, was to figure out just what DmnD's function is in Anabaena. It will be interesting to see whether or not it is necessary for the survival of the organism, as that will indicate whether (or how much) the two genes interact. To accomplish that general goal, three different experiments were carried out with dmnD, each aimed at discerning a different aspect of its possible functions. The first, and most time-consuming, was to obtain a dmnD- strain of Anabaena, to see whether knocking the gene out would be lethal. Second, a strain of Anabaena was to be constructed that would over-express the gene product…to test the phenotype of such a mutant. Finally, the role of dmnD as a methylase in E. coli was tested by inserting the working gene into a dam- strain, or one that could not normally methylate the sequence GATC.

Knockout mutant
There are several methods of knocking out a gene. We chose to insert an antibiotic cassette because it makes the presence of the knockout gene very easy to select for, provided there isn't any other instance of resistance to that antibiotic in the organism. We chose to use the Nm^r cassette of CK3 to insert into the center of dmnD. Once the gene was cloned, it was inserted into the plasmid pBluescript at EcoRI sites for easy manipulation. It occurred in both orientations, creating the plasmids pUR164 and pUR165. This way we know exactly what we have-pUR164 is oriented opposite to lacZ, pUR165 is oriented parallel to lacZ-and can cover the bases if something doesn't turn out as expected. The orientation here really only matters for future manipulations in other plasmids, the orientation as compared to lacZ shouldn't make any difference.

The plasmid pUR164 was used to begin all manipulations for the knockout mutant. The CK3 cassette for neomycin/kanamycin resistance was cut out of pRL442 using HinCII. We tried to make an underdigest of the restriction reaction because there is a HinCII site just within the promoter region of the cassette, but we were unsuccessful in the first attempt. We continued on, however, with the hope that the lack of a few base pairs from the promoter would not affect the function of the cassette. Meanwhile, we managed to create an underdigest in case function was affected. So, after cutting pUR164 with StyI (it has one restriction site in the plasmid, directly in the center of dmnD), the ends were built up to blunt with T4 Polymerase so that they would match with the blunt ends of the cassette. The cut and blunted pUR164 was then ligated with CK3, and the result was two different plasmids, based on the two possible orientations of the cassette. pUR167 placed the cassette antiparallel to the gene, while in pUR168, they are parallel. Both plasmids must be used for the rest of the manipulations, since it is uncertain whether the orientation of the cassette will affect the outcome.

Once the gene was successfully interrupted, it had to be readied for conjugation. To do so, it was ligated into pRL271, a plasmid that may be used in conjugation with Anabaena. pRL271, pUR167, and pUR168 were all cut with SpeI and XhoI, and the interrupted dmnD fragment from pUR167 and pRL168 were ligated into pRL271, creating pUR169 (from pRL167) and pUR170 (from pRL168). The use of the SpeI and XhoI sites created only one possible orientation of the fragment with the interrupted gene and included the entire gene. The E. coli strain DH5a, which was used in all of the previous manipulations, cannot perform conjugation with Anabaena, unfortunately, so the plasmid DNA was electroporated into HB101 along with helper plasmids pRL443 and pRL623 to aid in conjugation.

Conjugation was performed first with pUR169, as it would be difficult to balance two conjugations at the same time in the lab. The same procedure was also used with pUR170, however. Two dilutions of the Anabaena suspensions were used so that a proper amount of cells would end up surviving on the membrane-not so few that not enough colonies could be observed, yet not so many that one colony would be indistinguishable from the next. The first was a 1:1 ratio of Anabaena to E. coli, the second a 1:10 ratio (total was .6 ml in both cases). After both had been allowed to grow on BG11 media for two days, the membranes were transferred to BG11 plates with Nm, to select for cells which had incorporated the cassette (inside the gene) into their genomes. In other words, we were selecting for single recombinants. Once the membranes were transferred to the Nm plates, massive deaths occurred, presumably to those colonies of Anabaena which had not incorporated the interrupted dmnD into their own genomes. Ten of the surviving colonies (single recombinants) were then harvested and grown up.

Once we had ten colonies of single recombinants grown up to an adequate volume in flasks, we looked for double recombinants, or colonies of cells that had completely replaced the dmnD in their genome with the interrupted one from the plasmid. In order to do this, we must look back to pRL271, the plasmid we used for conjugation. In addition to being useful for conjugation, it contains a gene called sacB, which codes for a protein that breaks sucrose into its component parts. When a cell has this protein, it will take each molecule of sucrose and cleave it into two molecules of monosaccharide. This changes the solute concentration within the cell, causing water to diffuse in, lysing the cell. This means that any cells containing the sacB gene will not survive in media containing sucrose. Single recombinants will still contain all of pRL271 in their genomes (including sacB), and so will not survive on sucrose, whereas double recombinants have gotten sacB out of their genomes and will not care about the presence of sucrose in the media. There is only one problem with this approach: sacB is obviously an extremely unfavorable gene to have, and so cells with mutated sacB will be naturally selected for at a relatively high rate. This simply means that we could not simply assume that anything that grows on sucrose is a double recombinant.

To check those colonies that did grow on sucrose for double recombination, a simple comparison test should work. Double recombinants should grow on neomycin, due to the presence of the cassette, but they should not grow on erythromycin, which has a resistance gene in pRL271. Single recombinants in which the sacB gene is not doing its job will grow on both antibiotics. So twenty colonies were picked from each of the sucrose plates and placed on both Nm and Em plates. Unfortunately, most of the colonies (all but three) grew better on erythromycin than on neomycin, which points to a hypothesis that double recombinants are not possible and that all of the growth on sucrose was due to mutation of sacB. We decided to check the single recombinants and see how the plasmid was incorporated-whether it could go in both orientations or not, because that could give us a clue as to why double recombination did not appear to happen at first glance. A Southern blot was then performed for all ten original single recombinants. The results appeared to show two different sized fragments, although they were not terribly clear due to an error made during the process. This is consistent with two orientations of the inserted plasmid.

Overexpression
To overexpress a gene, it is placed into a multi-copy plasmid, which is one that transcribes the gene more often than most vectors. We took the intact dmnD gene from pUR164 using the enzymes PstI and BspDI, which flank it within the multiple cloning site. This fragment was ligated into pRL11(the overexpression vector) and the resulting plasmid was called pUR179. The plasmid was then purified and electroporated into the HB101 strain of E. coli along with pRL1124 and pDS4101 for helper plasmids. The same procedure for conjugation was then followed as in the previous experiment, and after two days on plain BG11, the membranes were transferred to erythromycin-5.0 plates. The erythromycin plates were used because pRL11 is Em resistant, where wild-type Anabaena is not; so wild-type will die and those cells which have incorporated the plasmid will not. Also, the gene was placed just downstream of the gene for the Em resistance. So when that gene is induced by the presence of Em, transcription is begun by the very strong promoter which will also end up helping to transcribe dmnD. The promoter it is under in the plasmid may not be strong enough to transcribe the gene a lot, so the Em promoter just gives it an extra kick.
Unfortunately, the membranes were not transferred onto the Em plates at the appropriate time, so there was some other growth on the plates. It was not identified, but it was likely that it was parasitic. An attempt was made to salvage one membrane by placing it on Em, but that was unsuccessful. So the conjugation was performed over again, and no results had been obtained before I left.

Methylation in E. coli
One more thing that would be interesting to test would be whether or not dmnD could replace the methylation function in E. coli that couldn't methylate GATC sequences. The dam gene usually does this, so the dam- strain to be used in this experiment was GM4715. This experiment had not yet been performed with dmnA, so I tested both, so I could compare the results. I had both genes intact in the pRL11 plasmid, which would work very well as an expression vector. So each plasmid (pUR154 for dmnA; pUR179 for dmnD) was electroporated into the E. coli, strain GM4715 and 20ml of the cultures were plated on Cm plates with Em. The Em in this case was simply to induce the transcription of the respective genes within the E. coli (both have the gene placed just downstream of the Em resistance gene, and are thus under the control of its promoter). Three days later, the plates were checked, and neither plate appeared to have any growth, so samples were replated with more cells. The next day, the first plates were checked again and found to have growth, although it could only be seen under a microscope on the plate with dmnD.

Each strain of bacteria was then grown for several days in liquid media, and the DNA was isolated from each. There appeared to be slightly less DNA in the sample with dmnD, but that can explained by the apparent slower growth of that strain. The big test was next: whether or not the genomic DNA was methylated at the adenine in GATC. If it was, then the gene inserted on pRL11 is responsible, and this could mean that it performs the same function in Anabaena…even that methylation is its main function. To test this, DNA from each strain was cut with both DpnI and DpnII. DpnI cuts at GmATC sequences, and DpnII cuts at GATC sequences. So the only difference in cutting is whether or no the sequence is methylated. DmnA showed a few very large segments under DpnI, but more, smaller, distinct segments after cutting with DpnII. This follows with the gene not working to methylate the DNA, for whatever reason. The plasmid with dmnD showed lots of segments under both enzymes, although those under DpnII were a bit more distinct. So dmnD appears to lead to methylation of at least some of the DNA, even if it isn't as much as the natural dam might.

Conclusions
Not too many conclusions could be drawn from the results observed. The experiments will have to be carried through their ends and more experiments conducted before anything is concluded. But from the minimal results, it was possible to make some tentative conclusions. With the knockout mutant experiments, it was possible to conclude that a single recombinant was possible, but that it is a little more difficult to obtain a double, if it isn't impossible. With the single recombinants, the Southern blot showed two different sized fragments containing the probe. This indicates with high probability that the plasmid has inserted itself into the genome in both orientations. So even if a double recombinant cannot be made, both orientations of single recombination can be tested for various functions.

The experiment with E. coli of strain GM4715 ended up with some fairly interesting results, however. If the strain with dmnA was cut more frequently by DpnII than DpnI, it indicates that the DNA was not methylated at GATC sequences. This means that, even though dmnA is supposed to be a methylase, this is probably not its main function in Anabaena. The strain containing dmnD was cut only a little bit more frequently by DpnII, and the fact that DpnI cut much at all shows that some methylation of GATC sequences occurred due to the presence of dmnD. With the above pieces of information, it is possible to propose that the reason for the presence of the two enzymes that seem (from looking at sequences) to perform the same function is that they don't actually both methylate GATC. DmnD methylates, while DmnA performs some other function yet to be ascertained. Perhaps at one time DmnA did methylate and eventually lost this function, but the point is that today it appears that it doesn't. The puzzler, then, is why the strain with dmnD grew so much slower than that with dmnA. Perhaps methylating the DNA is energetically more expensive than GM4715 is equipped to handle.

Further Studies
The list of possible other ways to look at and investigate this gene is endless, but there are a few that I came across while doing the experiments I have described.

Many directions could be taken with the experiments with the knockout mutant of dmnD. First would be to investigate the single recombinants further, see if there is any difference in phenotype when a second copy of the gene (an interrupted one) in introduced into the genome. The heterocyst formation upon nitrogen deprivation would be the first primary target. With other knocked-out genes, the single recombinants have shown dramatic differences from both wild-type and double recombinant strains. The obvious direction to go from single recombinants would be to continuing to try to find a double recombinant. The first place to look would be with those three colonies that grew on Nm but not Em. They are as likely due to human error in transferring between plates as to double recombination, but there is the possibility which should not be thrown out. The single recombinants can also again be plated on sucrose and then the antibiotics to see if it was merely the fact that double recombination was too rare an event to show up in the 200 colonies plated.

For the overexpression experiment, there is one obvious thing: the experiment must be finished. To do so, the single recombinants with pRL11 must be obtained and analyzed. They can be deprived of nitrogen to observe the heterocyst phenotype brought on by an excess of the protein made by dmnD.

There is one more thing I would like to see with the experiment that inserted dmnA and dmnD into dam- E. coli. The growth rates of each experimental strain of E. coli should be compared to a control of GM4715 with pRL11 inserted without any extra genes, to see exactly how each of the other genes really affected growth rate. Previously, they were only compared to each other, so it wasn't seen how they compared to a natural growth rate (presumably, the addition of only pRL11 shouldn't affect it). The amount of cutting by DpnI and DpnII could also then be compared to a control, to see whether there is any cutting at all without a methylase being added. The growth rate could also be compared to a strain with a gene added on pRL11 that is known, definitely, to methylate GATC.

Acknowledgments
Many thanks to Jeff Elhai for inviting me to use his lab this summer. Also, the greatest appreciation for Andrey Matveyev for all of his help and guidance, along with the occasional diversion with amusing stories. Thanks as well to everyone else in the lab for the wake-up calls, much help, and some great times!