Comparison of environment by UniFrac

Hello,

I am a bit confused by the use of UniFrac in the OTU-based analysis in the MiSeq SOP (http://www.mothur.org/wiki/MiSeq_SOP).
As I understand, the original definition of UniFrac is based on the phylogenetic tree of OTUs; To compare different samples, a sample name is assigned to different OTUs; To compare different environments or treatments, a certain environment name is assigned to different OTUs.

In the MiSeq SOP, you used the dendrogram of different samples (after calculating Yue & Clayton dissimilarity index), not a phylogenetic tree of OTUs, as the input for the UniFrac to compare environments or treatments (so the tree shows the relationship among samples, not among sequences).
I was searching for a paper which used this strategy to compare environments but I have not been able to find any.

Do you think it is still a valid method to compare different environments (in that case could you introduce some papers which followed this strategy, if you know any)? Or should I rather follow the original definition by assigning a certain environment to each OTU and conduct the phylogeny-based UniFrac? (or alternatively compare each sample by phylogeny-based UniFrac and conduct anosim to see the significance of clustering?)

I have a total of 16 different samples from 4 different environments and would like to see if there is any significant effect of environments.

I would appreciate if you could answer these questions.

regards,
yu62

In the SOP we do it as a statistical test to compare groups based on a tree constructed using samples where the tree is from a beta-diversity index and we also do it to calculate distances based on the sequences in each sample

Pat