T-20 (enfuvirtide) resistance is caused by the N43D primary resistance mutation at its presumed binding site at the N-terminal heptad repeat (N-HR) of gp41, accompanied by the S138A secondary mutation at the C-terminal HR of gp41 (C-HR). We have discovered that modifying T-20 to include S138A (T-20S138A) allows it to efficiently block wild-type and T20-resistant viruses, by a mechanism that involves improved binding of T-20S138A to the N-HR that contains the N43D primary mutation. To determine how HIV-1 in turn escapes T-20S138A we used a dose escalation method to select T-20S138A-resistant HIV-1 starting with either wild-type (HIV-1WT) or T-20-resistant (HIV-1N43D/S138A) virus. We found that when starting with WT background, I37N and L44M emerged in the N-HR of gp41, and N126K in the C-HR. However, when starting with HIV-1N43D/S138A, L33S and I69L emerged in N-HR, and E137K in C-HR. T-20S138A-resistant recombinant HIV-1 showed cross-resistance to other T-20 derivatives, but not to C34 derivatives, suggesting that T-20S138A suppressed HIV-1 replication by a similar mechanism to T-20. Furthermore, E137K enhanced viral replication kinetics and restored binding affinity with N-HR containing N43D, indicating that it acts as a secondary, compensatory mutation. We therefore introduced E137K into T-20S138A (T-20E137K/S138A) and revealed that T-20E137K/S138A moderately suppressed replication of T-20S138A-resistant HIV-1. T-20E137K/S138A retained activity to HIV-1 without L33S, which seems to be a key mutation for T-20 derivatives. Our data demonstrate that secondary mutations can be consistently used for the design of peptide inhibitors that block replication of HIV resistant to fusion inhibitors.