If you believe that adipocytes can arise from primitive fibroblastic stromal
cells, then you can use CD34 (QBend 10 clone)on formalin fixed paraffin
embedded muscle tissue- this would give better morphology than frozens nuked
in a microwave. Reference for the stromal cells:  J Cutan Pathol 1997
Sep;24(8):484-93 Fibrohistiocytic differentiation in subcutaneous fatty
tumors. Study of spindle cell, pleomorphic, myxoid, and atypical lipoma and
dedifferentiated liposarcoma cases composed in part of CD34+ fibroblasts and
FXIIIa+ histiocytes. Silverman JS, Tamsen A.

Frozen sections followed by microwave antigen retrieval is very harsh on the
tissues. Frozen sections should eliminate the need for antigen retrieval
anyway though you don't mention what primary antibody you are studying.
Paraffin sections of muscle and/or fat would yield better results, so would
using silane coated or "PLUS" slides. If you need to see the lipids, which
are dissolved out by paraffin processing,  you can post fix formalin fixed
tissue in osmium tetroxide and then do paraffin embedding. CD34 requires no
antigen retrieval even with formalin fixed paraffin embedded sections.  I
don't know what osmication would do to CD34 immunoreactivity though. Hope
this helps some.
Jeff Silverman
Southside Hospital
Bay Shore, NY