Which Regression scoring to use when have small dataset?

I am currently working on a Elastic-net regression assignment. The problem is I only have 40 training examples and I have to use 4 fold Gridsearchcv for hyper-parameter tuning. I am using sklearn for everything. But the problem is I am getting negative R2 values in some cross-validation scores.

I wanted to ask if there are some better scoring functions that I could use as scoring metric in Gridseachcv that would give better results, and better parameter tuning.