Depending on the circuit configuration, you may want a high-side (pulling up) switch as well as a low-side (pulling down) switch. The low-side switch is generally easy, since you need a gate voltage around 10 V above ground. For the high side switch, you need a voltage around 10 V below the 140 V rail (if you're using a p-FET), or 10 V above that rail (for an n-FET).
That level shift--to provide a gate voltage up near the high voltage rail, and maybe to generate a gate voltage higher than that rail--is usually hard. Typical ways to do that include little pulse transformers, or gate drive ICs.
The most obvious way to drive that piezo would presumably be a full bridge ("H-bridge"), alternately applying +V and -V to the device. This allows you to apply a peak-to-peak swing of 2V with a supply voltage of only V. This requires two high-side switches, and two low-side switches.
A typical gate drive IC accepts logic levels with some reasonable threshold. We could consider the FAN7380, for example:
http://www.fairchildsemi.com/ds/FA/FAN7380.pdf
It accepts two digital inputs, for the high- and low-side switches, with a logic threshold between 0.8 and 2.5 V. So a PIC running at 3.3 or 5 V can drive that with no trouble.
Note that the FAN7380 is designed to work with two n-FETs; so that high-side switch requires a gate drive higher than Vb. So the IC uses a switched capacitor circuit, with Cboot and Dboot, to generate a voltage Vb + (Vcc - 0.6 V). That seems like a nuisance, but p-FETs are very rarely used in power electronics; since the electron mobility in silicon is greater than the hole mobility, a p-FET with the same dimensions as an n-FET will always have worse Rdson.
Another way to drive it might be with a resonant LC circuit: the transducer itself supplies some or all of the C, and you add sufficient inductance for the desired resonance frequency. Circuits of that form could be built with only a low-side switch, in which case you would need only a low-side gate drive IC. That also produces a relatively sinusoidal waveform. But the amplitude at resonance is rather sensitive to the component values, so it's not as clean as square wave (or "modified sine wave", +V, 0, -V, 0, ...) excitation with a full- or half-bridge.