There are several differences. But first, you have to differentiate between different experiments with different objectives in different species.
a) Mice are not man: Standard anti-mouse anti-CD3 antibody (hamster-anti-mouse 145-2C11) has intermediate affinity, plate-bound use is necessary to induce strong T cell proliferation. In contrast: standard anti-human CD3 (mouse-anti-human OKT3) has very strong affinity, soluble OKT-3 induce strong T cell receptor stimulation
b) What is the objective? Examples: kinome analysis needs strong but short T cell stimulation (use plate-bound anti-CD3), coculture with regulatory T cells needs suboptimal T cell stimulation (strong signal overcome suppressive activity)
c) Use and concentration of anti-CD3 stimulation is strictly dependent on the individual antibody (affinity, species, cross linking activity)
In addition, it is true that immobilized anti-CD3 (on plate surface or on microbeads) srosslinks the CD3 and thus is a better stimulant than soluble antibody. However, it cannot be consider a replacement for addition of antiCD28. If you would not have the CD28 signal, your T cells would not respond optimally, especially if you purify them prior to culture (i.e. remove monocyte-macrophages from say initial PBMC suspension). Also, if you a re interested in the question of "maximal available response" from your T cells, you would use immobilized antiCD3 plus soluble antiCD28 even if stimulating the PBMC. Technical remark: unless you use the stimulatory antibodies you or your lab already used and titrated, it is prudent to to do a pilot with some combinations of concentrations of both Abs around that recommended (in our hands we had to dilute one of the antiCD3s for immobilization at least twice from the recommended, or we had mostly apoptosis).
Souce: NovoPro 2018-03-13