I think it is very difficult to answer your question without knowing more on the software you are using for PCA.

Finally and maybe more importantly, it is also important to mention that 3D plots are very misleading in 2D representation in scientific papers. There are many authors in data visualization that are highlighting the pitfalls of 3D because of inherent human brain limitations to interpret it (Fundamentals of Data Visualization (clauswilke.com) chapter 26; Chart Dos and Don'ts - Data Visualization - LibGuides at Duke University and many other refs). This is especially true when this is a static figure in a paper vs a gif or an animated figure which allows a better reconstruction of the 3D by the readers.

My advice would be to decompose your 3D in 2 2D graphs or to simply remove the factor 3 which only represent 2% of the variance of your data (very meaningless in PCA for describing your data (at least in my veterinary field). A way to improve readability would be to add dotted lines which relies the points to their (x,y,z) coordinates but at the costs of making the figure confusing.

In the current version of your graph for example it is very impossible to know which 2D ellipses are presented therefore the message you try to pass with the figure is not strairghforward.

My 2 cents on that.