Hmmm... depending on how loose you're willing to get with terms like
"content", it could be pushed back to the 1930s, when the "high tech"
way to store large numbers of business records was on punched cards,
where the "content" (the records themselves) was in a form deliberately
arranged for ease of processing (by relay-controlled card sorters and
tabulators), and the records were prepared for human presentation by
having the tabulators generate a report presenting the data in a
specified form, via "style rules" specified by the wiring on a relay
patch panel. (Punched card technology itself dates from earlier than
that, of course, but IBM's machines couldn't deal with alphabetic
data directly until they introduced the 40x series in 1931).
On the other hand, if you're talking about general text (which would
also exclude, say, the 3270 screen layout specs which you find in
CICS applications and the like) the best candidate might be some
early macro package for runoff, the CTSS text formatter, back in the '60s.
Unfortunately, a quick search of the online catalogs for the MIT libraries
and the LCS/MIT-AI reading room doesn't turn up any pertinent citations.
Maybe I'll find something in my CTSS manual this evening, but it's at
home, and right now, I'm not.
I'm not sure that pre-computer typesetting technology would have allowed
for reuse of text in a significantly different format without resetting
it --- if you're changing the line breaks, you want to change the spacing
as well, and a linotype machine casts a line of type (hence the name) as
a single, physical lead slug. You could, I suppose, saw it up at the word
boundaries, but that's enough work that you might as well melt down the
original slugs and reset from scratch. On the other hand, if *all* you
wanted to do was change the fonts of the header text, you might have been
able to reuse some of the body-text slugs...
(I had the privilege of watching a demo of a 1930's-era linotype --- minus
molten lead --- this weekend; it's a really neat piece of machinery.
As another footnote, going back to the relay tabulators, the records on the
cards were generally pretty tightly compressed --- they had only 80 columns,
and tricks like using a single column for multiple purposes were common.
For this reason alone, four-digit years were a real extravagance in a lot
of tab machine applications. And the habits people learned on tab machines
did persist into the computer era --- the RPG programming language was,
IIRC, specificially designed in the 1960s as a programming system for
the IBM 1401 which would be easy to explain to tab-machine jocks. The Y2K
problem is therefore older than electronic computers themselves....).