First of all, were the two data sets you used of a same taxon-sampling? were the relationships of your interest strongly supported? Haplotype network analysis is sensitive to missing data, and an ITS phylogeny (better generated with likelihood methods as Aaron suggested) may be more reliable in this case to infer species phylogeny from gene phylogenies.

Different genes will often have different histories.

Depending on the strength of the signal different analyses can often produce different topologies. Neither analysis you have done is more reliable than the other in my opinion - perhaps go to a likelihood/Bayesian based approach.

You could run an analysis where you weighted the character states differently but this would need some justification. Differences are often the most interesting aspect of phylogenetic analyses - perhaps there is a biological explanation for the diescrepency

Thank you very much for your suggestions.

I'll try the other approaches. Effectively the current results are both interesting. I am working on a very isolated population, exaploid and growing at high elevation, very different from any other known taxa. It shows an own single type of haplotype that results more related to another species geographically fully separated. Does this uniformity of haplotype allow me to consider this population an autopolyploid? Chromosomes are too small for karyotyping.

On the contrary, ITS analysis puts this population close to a diploid species which grows in the same territory but at low elevation (some tetraploid counts are also known). Morphologically and ecologically they are totally different. Can I hypothesyze a speciation process from this taxa by reproduction from unreduced gametes (2n=18 + 2n=36)? I will appreciate any other your suggestions.

It is not uncommon to get different results. You might check whether the genes underly selectino preassure using for example Tajimas D or the McDonal Kreitman test (implemented in DNASP: test for neutral evolutino). For the ITS i would expect neutral evolution but perhaps the other gene might show some signs of selection preassure going on. Probably the results are different as the two markers are evolvind at different "speed". Hence resulting in different pictures in the phylogenetic analysis.

I think you can hypothesyze a speciation via autopolyploidy, however, you many need to device a better demostration to persuade your audiences/readers, e.g. multiple nuclear loci. LEAFY and NEEDLY was frequently used as additional phylogenetic nulcear markers, you may want to try them after CAREFUL considerations.

To construct a network-like reticulate phylogenetic tree from multiple nuclear loci, see references elsewhere, e.g. Mol Biol Rep (2011) 38:1171–1185, DOI 10.1007/s11033-010-0215-y; Plant Syst Evol (2010) 290:85–97, DOI 10.1007/s00606-010-0351-9.

On thepolyploidy-facilitated expansion of plant taxa onto high elevation, I recommend you a paper from the third pole, the Qinghai-Tibetan Plateau, on the Allium przewalskianum diploid/tetraploid complex (WU, L.-L., CUI, X.-K., MILNE, R. I., SUN, Y.-S. and LIU, J.-Q. (2010), Multiple autopolyploidizations and range expansion of Allium przewalskianum Regel. (Alliaceae) in the Qinghai-Tibetan Plateau. Molecular Ecology, 19: 1691–1704. doi: 10.1111/j.1365-294X.2010.04613.x).

First of all, were the two data sets you used of a same taxon-sampling? were the relationships of your interest strongly supported? Haplotype network analysis is sensitive to missing data, and an ITS phylogeny (better generated with likelihood methods as Aaron suggested) may be more reliable in this case to infer species phylogeny from gene phylogenies.

These are plants, I assume? What level of variability are you sampling at? If you are looking at population-level variation, there is no reason to expect phylogenetic analyses of different genes to produce a congruent pattern of relationships among individuals because there may well not be any phylogenetic structure (in fact, phylogenetic analysis of such data may not be appropriate to begin with). As Zheng suggested, if you have sampled the same individuals for these genes, you could always concatenate the data and run as a single data set, if you believe that there should be a single historical pattern of relationships. That would at least give you a starting point of the "weight of the evidence" from which you can begin to ask what is incongruent with what.

Different genes can have different histories because of stochastic sorting, selection or hybridisation. Additionally, there is uncertainty associated with tree construction. With respect to mtDNA, Ne is one quarter that of nuclear genes, and it reflects matriarchial phylogeny alone, so often disagrees with nuclear genes, especially in the case of rapid radiations. Your best estimate of organismal relationships will come from analysing multiple genes and seeing which histories are best supported.

Although more information would be necessary to draw definitive conclusions your results would suggest allopolyploidy rather than autopolyploidy. I think the fact that different markers suggest different parents point in that direction although other explanations are possible (lineage sorting for example). It is my understanding that there are several nuclear and chloroplast PEPC genes, did you sequence all the PEPC genes? Also, when you interpret your results, keep in mind that ITS copies may evolve through concerted evolution.

A main problem is level of resolution/information. using sequences depending on genes rates of variation can be very variable. You may realize that in many publications the defined haplotypes are sustained by no more that 2 or 3 mutations/1000 nucleotides sequenced. Of course, in such cases a single error of lecture/sequencing can change all your results.

ITS is rather variable and most of time will offer you more resolution than most of genes (especially chloroplastic ones), it is not a surprize that you've got different results.

Another point is also the type of algorithms used, for example the software "network" (following Bandelt) is based on a rough parsimony but some more recent softwares (packages under "R") are based on Bayesian statistics.

When you compare types of molecular tools (using the same individuals) the problem of quantity versus quality of information is even more crucial. What is the best approach? To favour quantity of information (but with band patterns: microsat, ISSR, AFLP....) or to favour quality of information (sequences).

We published recently a very preliminary paper (a much more complete one is in progress) comparing level of information depending on markers.

Hoping it helps !!!

Cheers from Luc

2 or 3 in 1000 pb?

I've obtained ITS haplotypes that differ in a large number of nucleotides

Yoannis

Of course, depending on closeness of your species or if you work at the population level. For example on my Mexican Lupins (among around 50 species on ITS2 only) I've got maximum 5 mutations......I've got a friend who is working with Artemisia species and it is almost the same rate.

But I suppose that if you are working within a family this rate will be much higher.

Cheers from Luc

Hi Luc,

you are right, I'm working with populations of three species that are closely related, so the mutation rate is high

cheers

The incongurence between different phylogenies can be explained by hybridisation and allopolyploidy. Both events are quite common in plants. Hybrids and allopolyploids often tend to homogenise ITS repeats towards one parental type as a result of concerted evoluton. As a consequence, cloned ITS from a hybrid will match only to one progenitor sequence but not to the other masking, thus, species history. In contrast, single or low copy sequences that do not undergo homogenisation (or at considerably lower rates) might show haplotypes corresponding to progenitor haplotypes. My suggestion is to check chromosomes of your plants to see if these are diploids or polyploids. You can also use anóther phylogenic marker as suggested by the others. Chloropolast intronic squences work quite well.