We show that surface plasmons of a two-dimensional Dirac metal such as graphene can be reflected by line-like perturbations hosting one-dimensional electron states. The reflection originates from a strong enhancement of the local optical conductivity caused by optical transitions involving these bound states. We propose that the bound states can be systematically created, controlled, and liquidated by an ultranarrow electrostatic gate. Using infrared nanoimaging, we obtain experimental evidence for the locally enhanced conductivity of graphene induced by a carbon nanotube gate, which supports this theoretical concept.Plasmon scattering and plasmon losses in Dirac materials, such as graphene and topological insulators, are problems of interest to both fundamental and applied research. It is an outstanding challenge to understand various kinds of interaction (electron-electron, electron-phonon, electron-photon, electron-disorder) responsible for these complex phenomena [1][2][3][4][5] . At the same time, control of plasmon scattering is critical if this class of materials is to become a new platform for nanophotonics [6][7][8][9] .One source of plasmon scattering is long-range inhomogeneity of the electron density, which causes local fluctuations in the plasmon wavelength λ p . If the inhomogeneities are weak, those of size comparable to the average λ p are expected to be the dominant scatterers 10,11 Surprisingly, recent experiments have revealed that one-dimensional (1D) defects of nominally atomic width can act as effective reflectors for plasmons with wavelengths as large as a few hundred nm. Strong plasmon reflection was observed near grain boundaries 12,13 , topological stacking faults 14 , as well as nanometerscale wrinkles and cracks 11,12 in graphene. If this anomalous reflection is indeed an ubiquitous effect largely unrelated to the specific nature of a defect, it calls for a universal explanation. In this Letter we attribute its origin to electron bound states commonly occurring near 1D defects. We show that optical transitions involving the bound states can produce strong dissipation at small distances x from the defect and therefore, alter plasmon dynamics. To support this idea we present a theoretical analysis of an exactly solvable model, which illustrates qualitative and quantitative characteristics of the bound states and predicts how their optical response depends on the tunable parameters of a 1D potential well. We also report an attempt to probe the predicted effects experimentally. Our approach is to employ an ultranarrow electric gate in the form of a carbon nanotube (CNT) to create a precisely tunable 1D barrier in graphene. This device enables a systematic investigation and control of plasmon propagation, including, in principle, an implementation of a plasmon on-off switch (Fig. 1). What we find is that the measured real-space profile of the plasmon amplitude (Fig. 4) cannot be accounted for by a local change in λ p alone. Instead, the data are consistent with the presence of an enhanced ...