I think the advice of choroba and Tux to base your solution on Text::CSV is well taken. But I'm a hopeless regex junky, so here's that approach (requires Perl verion 5.10 or greater):

c:\@Work\Perl\monks>perl -wMstrict -le
"use 5.010;
 ;;
 my @records = (
   ',P,Q,R,S',
   'A,,D,E,F',
   'L,T,,G,Q',
   'C,D,E,,L',
   'V,W,X,Y,',
   ' , , , ,',
   ',,,,',
   ',',
   ' , ',
   '',
   ' ',
   );
 ;;
 for my $rec (@records) {
   print qq{'$rec'};
   $rec =~ s{
     (?: \A | ,) \K \s* (?= ,) | (?<= ,) \s* (?= , | \z) | \A \s* \z
     }{NULL}xmsg;
   print qq{'$rec' \n};
   }
"
',P,Q,R,S'
'NULL,P,Q,R,S'

'A,,D,E,F'
'A,NULL,D,E,F'

'L,T,,G,Q'
'L,T,NULL,G,Q'

'C,D,E,,L'
'C,D,E,NULL,L'

'V,W,X,Y,'
'V,W,X,Y,NULL'

' , , , ,'
'NULL,NULL,NULL,NULL,NULL'

',,,,'
'NULL,NULL,NULL,NULL,NULL'

','
'NULL,NULL'

' , '
'NULL,NULL'

''
'NULL'

' '
'NULL'
[download]

The only problematic case is the  '' or  ' ' empty/blank string, but if you don't like how that's handled, it's easy to take the  \A \s* \z term out of the alternation or change it.