Molten salt has been suggested a fair few times in this thread, and I'd like to throw my own support in for it. I'll also mention that hot oil can be used for the same purpose if you want different recipes to consider.
The key reason I support it it is that molten salt, along with hot oil, lends itself very well to acting as a heat transfer fluid, and I think the fluid system lends itself very well to filling the needs of heat based power generation while providing interesting challenges for it. I've spoken of this before using hot oil for a heat transfer fluid when I proposed solar-termal in the solar power balance thread.   
(Links for those interested in reading up. It's specific to solar-thermal but much of the fluid talk fair bit applies to nuclear-thermal power as well. Everything important there
should be repeated here however.)
A heat transfer fluid lends itself to complex infrastructure, and train based transporation once 1.5 and fluid wagons arrive. You don't even need the cooling tower, just a special "heat exchanger" type of boiler in order to make it work. (A temperature sensor function would also be useful for advanced stuff but it's not necessary.) All the heat exchanger has to do is equalize the temperature of two fluids for so long as they remain inside it, with both sides flowing freely.
Even with just pumping it in the exchange however, you already encounter a new consideration for advanced designers as a co-current heat exchange system only equalizes the temperatures while a counter-current exchange let's them get far more bang for their buck as far as heating the water and steam goes, transferring the vast majority of heat from one fluid to the other and so letting much more work be done.
It's only a small trick, but it's one that will be readily apparent.
It's also worth noting that storage becomes an issue in a way it doesn't with item based heat storage: While just a few chests can store a remarkable large amount of material, fluid storage tanks have a lot more bulk and so require space considerations to a degree that items don't. It's more important for solar-thermal than nuclear-thermal but it does pose a potential problem for players to consider.
Beyond that, there are also potential considerations of pumping capacity that encourage players to design pumping stations to get the fluid to flow faster, and likewise filling the system with various fluids has it's own challenges. Hot water and steam are easy to take care of, but once you start creating fluids like molten salt or hot oil, then you have to ensure that you don't have too much entering the system, as fluids in Factorio require air gaps to flow. And of course, the cooling tower and steam cycle already provides practice for that.
Thinking further, if you want to send the message that nuclear power doesn't necessarily
have to be dangerous or that it's only as dangerous as you make it, another possibility is that steam can have the potential to explode in the buildings if it reaches too high a temperature while heat transfer fluids have no such problem. This provides another reason to move forwards, as moving from steam to a heat transfer fluid increases the safe operating temperature, or safe operating margin if you want to play it safe, of nuclear plants.
If steam also can explode in the pipes, that likewise provides another challenge for designers working at the hottest temperatures, as they need to figure out how to pass that actually use that heat without exploding their system. And here again we have a bit of easy learning and hard mastery, as a co-current exchange evens out temperatures while the more efficient counter-current exchange requires that they find ways to cool down the steam in mid current if they want to run their reactors at high temperature. (It's simple enough actually, they just need to let the steam turbines cool down the steam before heating it up again.)
That said, I'm far more on the fence about exploding steam in general as it lacks a good way to introduce itself to the player.
Ideally, every single part of the system can be presented to the player in isolation, for them to tackle at their own time.
Closed fluid loops? That introduction is already handled by the steam turbine and cooling tower.
Exploding steam? Again it's entirely separate from other issues, and if it works in the pipes as well as the reactors, there may be a way for players to encounter it early on, while not being a factor in the most basic setups. (Possibly a pressure release valve that reduces the temperature and volume of the steam when it hits a critical point. That said, I'm not sold on it.)
Heat exchangers and heat transfer fluid? They're their own thing in just two items. The player already knows how to deal with everything else.
Counter-current exchange: Not necessary, just more efficient, and so can be tackled by players at their leisure. And again, I expect the way the game works already will facilitate it very easily.
The fight between counter-currents and exploding steam at high temperature efficiencies? That's the final step after everything else is done, and it's one that, again, players will have some time working out, effectively making it the final boss of power plant design.
The greatest gulf that we must leap is the gulf between each other's assumptions and conceptions. To argue fairly, we must reach consensus on the meanings and values of basic principles. -Thereisnosaurus