Here is official word from DXO tech support:
DxO Optics Pro uses Adobe RGB as its internal working color space.
Thanks for sharing that, it's a much clearer answer than the previous one.
What they are probably referring to is that due to the 'uber' huge Prophoto RGB gamut, the encoding/quantization steps represent pretty large steps in gradients. That's why it is recommended to not use that colorspace with 8-bit/channel images. Using a smaller gamut color space, the quantization steps get smaller, and there is less risk of posterization when applying extreme postprocessing. A 32-b/ch colorspace at least offers a more accurate quantization, a floating-point encoding even more accurate.
I't is a bit puzzling though why they chose to limit themselves to Adobe RGB instead of something with a bit larger gamut like Bruce Lindbloom's Beta RGB
(which can accomodate pretty much all important/common colors, but with relatively small quantization step increments, and without an immediate need for chromatic adaptation).