For the time being PD4ML does not support the feature.
There is a technical problem: for HTML layouting PD4ML needs to have a possibility to measure PDF string widths from Java. In regular mode it is easy to do – it is just a sum of string character widths. The character widths can be taken from TTF fonts metrics or from Acroread built-in fonts metrics (The acroread font metrics are hardcoded to PD4ML lib).
If we rely to the autokerning feature of PDF format – we cannot get anymore the exact width of a particular string, until we re-implemented kerning algorithms of Acroread, for example. We did not address the issue yet.