Historically, they were prepositional constructions that became grammaticalized; now, on verbs of motion, they could be considered productive derivational affixes. For those, X/Y/Z indicates the three spatial dimensions (with some exceptions), while C and P are containment and proximity. The +/- forms could be glossed as left/right, up/down, forward/backward, inside/outside, and near/far, respectively, with "e" forms perhaps translating as "level with" (perhaps "overlapping" for C). Plain "e" for P is spatially undefined and never used in the locational sense, only in a couple of idiomatic derivational constructions to mean something like "in moderation". The e+, e-, +e, -e, +-, and -+ forms show direction of motion, while the e, +, and - forms show location of the action itself. So rasṣútaı 1.move<Xe-> would be "I am moving (to my) right", while rassuaı 1.move<X+> would be "I am moving, while being over to the right (of some salient object)". In practice, the e/+/- forms aren't used in the locational sense much, and it's dealt with more often by actual prepositional phrases. One wouldn't usually say rassuaı, one would say... something like rasú rasaı das; I'm not certain. I have to go back and figure out how similar the modern prepositions actually are to the old ones that got grammaticalized.
But e/+/- forms show up much more often with verb roots that don't deal with motion; in these cases their meanings are highly lexicalized, and typically idiomatic/metaphorical/having nothing at all to do with direction. Only verbs of motion have full axis paradigms; the rest just have a handful of derived forms that use the axes as if they were the other half of a bipartite verb stem (or tripartite, if the verb is already a compound). Every X/Y/Z/C/P marker typically has a couple of different derivational meanings, and which one is relevant is just lexical. As an example, -dí- P- "far from" can attach to some stems with vaguely semantically related results; there's a sense of distance and/or approximation, and also usually a negative connotation:
dlırí - he is a leader > dlırdî - he is a political puppet
dıtatàh - I helped you (directly) > dıtatdîah - I gave you (well-meaning but unhelpful) advice
Ahùraı - you are telling the truth > Ahŭrdıraı - you are telling a partial truth
Ikísí - they are young > Ikísdî - they are naïve
Some, but not all, of these have related C- forms that don't have the negative connotation:
dlırḳŭı - he leads indirectly / from afar
dıtatḳŭaa - I gave you advice
Alternatively, you also get things like kaǎtdıaı "I am illuminating it from a distance" with -dí-, no negative connotations at all, and a clearly spatial sense of P-. There are patterns, but they're not clear-cut.
The R axis is different in that it's productive everywhere; it interacts with the aspect system and with a couple of other morphemes (including the defective E axis paradigm, which I'm still messing with) to create the tense system, where (roughly speaking) R+ is the future and R- in the past. I'm working out the kinks in that, especially given that R behaves slightly differently depending on whether the verb is in a perfective or imperfective aspect. I actually messed with the aspects yesterday as well, and most complex imperfective ones are now more clearly derived from their perfective forms plus the imperfective/durative suffix. Woop-de-doo, diachronics.