You can try different optimisers. nlme takes control argument that can used to set different optimisers:

control=lmeControl(opt = "optim")

Hopefully glmmPQL can pass that argument to nlme.

Thank you very much for your answer. 

I did try this but glmmPQL didn't pass this argument, I called a different package but that overrides the package MASS which I need for GLMM. 

You can test if nlme:lme converge with your data with some different optimers. If so, open option is to create wrapper function which calls   nlme:lme  with a different optimiser and pass other arguments through. Then you need to modify glmmPQL (change mcall[[1L]]

It works. Thank you very much. This is great after so many scripts.

Many thanks