Is it plagiarism (or bad practice) to cite reviews instead of source material directly?

Yes and No.

A review is not (or at least should not be) just a list of papers in the field, it is a research paper in its own right. A review should provide some form of new structure or idea that come from reviewing the scientific output the field in question. Thus a review can be useful when either pointing at the results emerging from the review or when you wish to provide an example of where a throrough description of a field is provided. So citing a review should primaruily be for general purposes.

If you need to cite a specific detail provided by one or several references in the review you should turn to the cited articles and check them. There is, first of all, no guarantee that the citations are correct and citing something that is not correct propagates errors in publications. In addition, the cited work may not be all the relevant work supporting a point from your perspective so your sources may become biased. After all your point for writing is most likely not the same as for the one writing the review.

A review therefore has a function on its own for providing overview and synthesis which can be cited but if you need to cite details you must go to the original work and also not limit yourself to the reference list of the review.

EDIT: Just to answer the question in full: plagiarism is only if you copy the text with out quotes. If you copy the references you are not necessarily plagiarizing but you will then give the false impression that you have actually read, in part of fully, the articles concerned and that is in the realm of unethical behaviour. In the best case, it is unwise for the resons given above.

The answer is somewhat field-specific. A mathematician's perspective here.

When citing a result which is uncontroversial and undisputed and unique, you can generally assume that the citation is accurate, and cite the contents of the reference as stated in the review article. So, for instance, if you are reading "Review of geometry by A. Euclid who cites the statement In a right angled triangle, the square of the hypotenuse is equal to the sum of the squares of the other two sides." which he attributes to A theorem about triangles by J. Pythagoras, then you can go ahead and paraphrase the above formulation and cite Pythagoras (not mentioning Euclid).

There is a thin line between results which are "classical" and those which are "modern". If it so happened that Pythagoras proved his theorem long ago that it is taught in graduate-level courses (but you still want to cite it because it is not universally known) then it is OK to give the more up-to-date reference by Euclid. Probably best avoided if Pythagoras is still alive and kicking (and would be glad to get another citation). Very advisable if the original is difficult to get your hands on, and/or written in a foreign language.

In both cases, you should (of course!) look up the original reference and see if Euclid is not making stuff up. If you don't do this, you are putting your reputation in the hands of whoever wrote the survey, so you should ask yourself how much do you trust him. If we are talking about a well-established result, and Euclid is well known to be a serious mathematician, then more likely then not everything will be fine. In all other cases - exercise caution.

Whichever way you go, it is often also a good idea to mention that the reader who wishes to learn more about the topic might want to look up Euclid's survey paper.

It seems to me that you owe a debt to all four sets of authors. How about

"The recent development of methods to probe the physical association of genome regions in a unbiased and genome-wide scale should lead to rapid progress in our still-rudimentary understanding of the functional significance of chromatin loops (Misteli, 2007 based on Simonis et al., 2006; Wurtele and Chartrand, 2006;Zhao et al., 2006)."