I have done some checking on the operation of i-umlaut in Proto-Ossetic;
it seems that it wasn't blocked by consonant clusters. This is clearly
visible in diminutives in *-ika:, such as *pas^m-ika: 'spring wool'
(from *pas^man- < *pek^s-mn., cf. New Persian pas^m, Oss. fæsm 'wool').
In Ossetic, we get *fais^mic^á- > *fais^c^á- > *festá- > Digor festæ,
Iron fist (excuse the accidental pun). Thus, it seems quite possible for
*-afri- (= Gk. -apris) to have developed into *-aifr- --> Slavic *-aipr-
> *-e^pr- (umlaut preceded the apocope of short vowels)
But there's another problem with etymologising Danapris/*dUne^prU as
"river such-and-such". With "river" as the head component the order of
the elements in the compound should be the reverse of what we see. In
Ossetic, at any rate, <-don> would follow, not precede, any descriptive
modifier; cf. also such old names as *saka-da:nu- > *sag-dan- 'the river
of the Sakas', *waru-da:nu- > *var-dan- (Ptolemy's Ouardane:s) 'wide
river/water' (in the Kuban system). If the "-apris" part is related to
*afra- 'deep' (: *afri- 'depth'?), *da:nu-afri- should rather be a
bahuvrihi compound meaning '(having) the depth of *da:nu-', where
*Da:nu- might be some "prototypical" river, perhaps a mythical or
semi-legendary one. [Note that even the Danube's name is an adjectival
derivative, *dah2new-jó- vel sim., as if '*dah2nu-like' or 'relating to
*dah2nu-' (= Tanaïs, if < *da:nav-ya-).] Danastris/Danastius/*dUne^strU
seems to be a parallel formation, though at the moment I can't make
sense of the second element.
Piotr