implicit differentiation and subscripted variables?

I am in the process of porting over a *lot* of lecture notes from Maple 
-> Maxima. For the most part, this has been a relatively straightforward 
'port'.

But, I am stuck a bit on handling implicit differentiation (which is a 
one-line command in Maple, but requires a couple of steps in Maxima). My 
problem isn't so much >1 steps ( can probably write a wrapper to shield 
students from the multiple steps), but rather, why the depends command 
is blowing up. Take the following example:

mat:matrix([0,s[a]*m[1],s[a]*m[a]],[s[o],0,0],[0,s[a],0]);

cp:charpoly(mat,lambda);

Now, what I want students to do is implicitly differentiate lambda wrt 
to each other variable in the characteristic polynomial. So, for 
example, to implicitly differentiate lambda wrt s[a], I would start with

depends(lambda,s[a]);

But, this throws an error:

depends: argument must be a symbol; found s[a]
  -- an error. To debug this try: debugmode(true);


Guessing that this is because depends doesn't like my attempt to 
'subscript' s[a], I try


mat:matrix([0,sa*m1,sa*ma],[so,0,0],[0,sa,0]);

cp:charpoly(mat,lambda);

depends(lambda,sa);

Now, this works fine (so, my guess was right).

Problem is, using

mat:matrix([0,s[a]*m[1],s[a]*m[a]],[s[o],0,0],[0,s[a],0]);

generates pretty ugly output (not concerned about aesthetics, as much as 
the fact that the lack of explicit subscripting makes the process more 
error prone for students). So, is there a recommended way to allow for 
subscripting of matrix elements (or any variable, for that matter) that 
won't cause depends to have 'issues'?

Thanks in advance.