Transcranial direct current stimulation (tDCS) to the dorsolateral prefrontal cortex (dlPFC) can modulate working memory (WM) performance. However, evidence regarding the enhancement of WM training, its sustainability and transferability is ambiguous. Since WM functioning appears to be lateralized in respect to stimulus characteristics, this study examined the difference between task-congruent (spatial-right, verbal-left), task-incongruent (spatial-left, verbal-right) and sham tDCS in regards to the efficacy of WM training. In a randomized, sham-controlled experiment, 71 healthy adults trained on a spatial or verbal adaptive n-back task. After a baseline session, anodal or sham tDCS (1 mA) to the right or left dlPFC was applied during the next three training sessions. Sustainability of training gains and near-transfer (verbal or spatial 3-back task) were tested in a fourth training and a follow-up session. Compared to sham stimulation, we found a steeper learning curve when WM training was combined with task-congruent tDCS. This advantage was also present compared to task-incongruent tDCS. Moreover, these effects lasted for up to nine months and transferred to the respective untrained task. These long-lasting, transferable, task-specific effects demonstrate a behaviorally relevant and sustainable facilitation of neuroplastic processes by tDCS that could be harnessed for the treatment of disorders associated with deficient WM.