Two metacognitive knowledge types in deductive domains are procedural and conditional. This work presents a preliminary study on the impact of metacognitive knowledge and motivation on transfer across two Intelligent Tutoring Systems (ITSs), then two experiments on metacognitive knowledge instruction. Throughout this work, we trained students on a logic ITS that supports a "default" forward-chaining and an alternative backward-chaining (BC) strategy, then a probability ITS that only supports BC. Students were grouped into those with conditional knowledge who know "how" and "when" to use each strategy ("StrBoth"), those with procedural knowledge who know "how only" ("StrHow"), and the rote students who persist in the default strategy and know neither how nor when ("Rote"). The online traces' initial accuracy was used to further split students into high- and low-motivation groups. The preliminary study showed that only high-motivation "StrBoth" students transferred their metacognitive knowledge across the two ITSs. The two experiments provided metacognitive knowledge instruction for "StrHow" and "Rote" students to catch up with their "StrBoth" peers. In "Exp." 1 we utilized prompted nudges to teach when to use BC, and in "Exp." 2 we combined nudges with worked examples to teach how and when to use BC. Based on our findings, we propose a "Metacognitive knowledge, initial Motivation, and instructional Interventions (MMI)" framework for transfer across ITSs. The framework suggests that the key factors for facilitating transfer are the motivation for "StrBoth" students, nudges for their "StrHow" peers, and the combination of worked examples and nudges for "Rote" students.