Underlying class changes between serialisation and deserialisation - python-2.7

Question

There is an object of testclass created at some point in the life of program and then it is serialized in file system using pickle or h5py. Now during this period the definition of testclass is updated. The update is always about adding new fields (and not removing existing ones).
Does deserializing the stored class back and referencing it using updated testclass definition a right approach?
Before Serialization: Class is defined
class testclass:
__fieldA = None
__fieldB = None
__fieldC = 'Hello'
def __init__(self, A, B):
self.__fieldA = A
self.__fieldB = B
def setA(self, A):
self.__fieldA = A
def getA(self):
return self.__fieldA
def setB(self, B):
self.__fieldB = B
def getB(self):
return self.__fieldB
def setC(self, C):
self.__fieldC = C
def getC(self):
return self.__fieldC
>>>orig_obj = testclass('testA', 'testB')
>>>isinstance(orig_obj, testclass)
>>>True
>>>with open('test', 'wb') as handle:
pickle.dump(orig_obj, handle, protocol=pickle.HIGHEST_PROTOCOL)
After Serialization: Class definition is changed
class testclass:
__fieldA = None
__fieldB = None
__fieldC = 'Hello'
__fieldD = 10
def __init__(self, A, B):
self.__fieldA = A
self.__fieldB = B
def setA(self, A):
self.__fieldA = A
def getA(self):
return self.__fieldA
def setB(self, B):
self.__fieldB = B
def getB(self):
return self.__fieldB
def setC(self, C):
self.__fieldC = C
def getC(self):
return self.__fieldC
def setD(self, D):
self.__fieldD = D
def getD(self):
return self.__fieldD
>>>with open('test', 'rb') as handle:
loaded_obj = pickle.load(handle)
>>>isinstance(loaded_obj, testclass)
>>>False
Now loaded_obj is not an instance of testclass. This makes sense and has nothing to do with the fact whether new field __fieldD is introduced or not.
In this scenario, what is the best way to ensure that loaded_obj is an instance of testclass?