scale function is used to transform the values. values are scaled using mean and standard deviation (link). While clustering, scaling is generally performed to obtain better cluster. clustering would change based on other parameters also such as distance method.
The option that i am using os scale="column" . This tranform my values?
yes, of course. you may try it row for clustering row-wise.
In addition, take a look here: A: cannot replicate the pheatmap scale function