You do not get an anti-Bax signal with Bax alone, only if you either load a mixture of Bax and tBID or Bax and compound 106. The loading control (anti-Tom 40) shows that the lack of a bind in lane 2 (Bax only) cannot be explained by a loading error (no sample loaded). You have to therefore assume that your anti-Bax reagent recognizes Bax only in the presence of either tBID or compound 106, presumably due to a conformational change of Bax by the two reagents.
Of course, this cannot be a denaturing SDS gel, more some kind of strange dot-plot arranged to look like a gel, as the complex would not survive denaturing conditions, and the Bax band would be expected to shift on a native gel upon tBid binding.
If you look at:
Nature Reviews Molecular Cell Biology 2014 Vol 15 No 1
Control of apoptosis by the BCL-2 protein family: implications for physiology and therapy
Peter E. Czabotar, Guillaume Lessene, Andreas Strasser & Jerry M. Adams
You will see that the Bax domain has a "latch" helix that is displaced upon binding of BH3-only BCL2 family protein to BAX and therefore becomes accessible for recognition by the antibody. Presumably, compound 106 binds into the same groove and has a similar effect on the BAX structure.
I'm not at all into the topic but it looks as if compound 106 is able to mimick tBid effect in a dose dependent manner (recruiting Bax to mitochondria?).