A bit of digging around finds a position at http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=8477 taken from from one of the 7-piece endgame databases that's claimed as mate in 524 (though I found a few sources that contend that's only the 'time-to-win' and that actual mate could take substantially longer) :
However, this isn't a 'constructed' position; there's no theme to the mating sequence, and the moves are the classic 'no rhyme or reason' dance that shows up so much in these multi-piece endgames. As far as actual chess problems go, the longest I was able to find was a mate in 270, as seen at http://chess.eusa.ed.ac.uk/Chess/Trivia/Longestmate.html :
White starts with 1. Bb1: the core idea here is that White wants to gain access to a6; Black shuttles his king back and forth between a8, b7 and c8, as White plays Ka5-a4-a3-a2-a1, Ba2, Ka1-b1-c1-d1-e1-f1-f2-e1-d1-c1-b1-a1, Bb1, Ka1-a2-a3-a4-a5. Since White has 'lost a move' with the e1-f1-f2-e1 triangulation, Black now has to move a pawn in lieu of moving his King and giving White access to a6. Black's available pawn moves are Pf6-f5-f4, Pf7-f6-f5, Ph4-h3-h2, Ph7-h6-h5-h4-h3 before he finally runs out of pawn moves and must allow White access to a6 to queen his lead b-pawn.
(I should note that it's not clear to me that the opening move is unique and that White can't play Ka5-a4-a3 and then Bb1; I presume there's some rationale that prevents this cook, but I don't immediately see it.)